 Research
 Open access
 Published:
A spatiotemporal specification language and its completeness & decidability
Journal of Cloud Computing volume 9, Article number: 65 (2020)
Abstract
In the past few years, significant progress has been made on spatiotemporal cyberphysical systems in achieving spatiotemporal properties on several longstanding tasks. With the broader specification of spatiotemporal properties on various applications, the concerns over their spatiotemporal logics have been raised in public, especially after the widely reported safetycritical systems involving selfdriving cars, intelligent transportation system, image processing. In this paper, we present a spatiotemporal specification language, STSL _{PC}, by combining Signal Temporal Logic (STL) with a spatial logic S4 _{u}, to characterize spatiotemporal dynamic behaviors of cyberphysical systems. This language is highly expressive: it allows the description of quantitative signals, by expressing spatiotemporal traces over real valued signals in dense time, and Boolean signals, by constraining values of spatial objects across threshold predicates. STSL _{PC} combines the power of temporal modalities and spatial operators, and enjoys important properties such as finite model property. We provide a Hilbertstyle axiomatization for the proposed STSL _{PC} and prove the soundness and completeness by the spatiotemporal extension of maximal consistent set and canonical model. Further, we demonstrate the decidability of STSL _{PC} and analyze the complexity of STSL _{PC}. Besides, we generalize STSL to the evolution of spatial objects over time, called STSL _{OC}, and provide the proof of its axiomatization system and decidability.
Introduction
It is a challenging work to model cyberphysical systems, not only because cyberphysical systems integrate cyber systems, physical environment and the interactive part of them, but also because cyberphysical systems combine temporal and spatial aspects, discrete and continuous behaviors, and nondeterministic models [1]. Describing spatiotemporal aspects is one of the important areas in cyberphysical systems. Many works have been done with concurrent [2], hybrid [3–5] and stochastic [6, 7] behaviors of motionbased spatially distributed systems [8], but fewer researchers concentrate on spatiotemporal aspects. The major problem is multidimensional expressiveness and expensive verifiability for modeling and analysis of the spatiotemporal behaviors of cyberphysical systems.
This work aims at building a spatiotemporal specification language by solving spatiotemporal constraints concerning dense time and realvalued variables, because an intelligent object in physical environment is equipped with changes in specified space and continuous time. More specifically, we confine ourselves to the combination of topometric space [9] and time constraints with realvalued interval, which is a halfopen and halfclosed interval in a time flow of a strict partial ordering of time points. We adopt the modal spatial logic S4 _{u} to express topometric constraints, which is one of the most influential form and the most expressiveness for topometric relations. For signal temporal logic (STL) [10, 11], there are two interpretations for signals, quantitative semantics and Boolean semantics. Quantitative semantics obtains realvalued signals from satisfaction degree of a trace over realvalued interval. While Boolean semantics evaluates Boolean signals from a trace which can be booleanized through a set of threshold predicates.
Because the changes of spatial entities and the flows of time are not independent, the combination between modal spatial and temporal logic is divided into two forms: STSL _{PC} and STSL _{OC}. STSL _{PC} means the changes of spatial propositions over time, while STSL _{OC} represents the changes or evolution of spatial objects over time. Each form is equipped with different expressiveness, so Boolean semantics and quantitative semantics for the two forms need to be provided respectively. Combining spatiotemporal constraints from temporal logic and modal spatial logic is a very important problem. Given a topometric temporal model \(\mathfrak {M}\) and an STSL _{PC} formula φ, the satisfiability problem of the formula φ is to check whether φ is satisfiable against model \(\mathfrak {M}\) or not. We present the semantics for the proposed language according to the satisfaction relations.
Many works have been done on the axiomatization and completeness of combined logics. The completeness of normal modal logic is present through maximal consistent set and canonical model [12]. J.M. Davoren [13] proposes topological semantics for intuitionistic tense logics and multimodal logic and provide the Hilbertstyle axiomatization and the completeness result. F.D. David [14] proves the absolute completeness of S4 _{u} for its measuretheoretic semantics. Based on our previous work [15], in this paper, we present an axiomatization system for STSL _{PC} and prove the soundness and completeness result of the axiomatization system based on maximal consistent set and canonical model. Further, the notation of finite model property provides a basis for the decidability of logics. The filtration, which is similar to bisimulation quotients with respect to equivalences generated by sets of formulae [16], can serve as an approach to achieve the finite model property. By way of the finite frame property, decidability can be proved by applying subframe transformations and a variant of the filtration technique [17]. The decidability of STSL _{PC} is proved according to the finite model property. For the decidable fragment, we present the complexity for the satisfiability problem and the decision procedure.
Compared with the another work [18] reasoning cyberphysical systems, DTL defines the trace to uniform discrete jumps and continuous evolution. Control operations in discrete jumps can control the continuous evolution along differential equations, which are interpreted in hybrid trace and verified through the differential invariant. Our proposed STSL _{PC} is interpreted on the sampling trace of statebased cyberphysical systems, where the formulas are verified through monitoring partial trace, instead of classical model checking, which needs to achieve all the behaviors of the systems. Further, A time scale is defined as arbitrary nonempty closed subset of the real numbers [19]. The continuity is defined according to the density of time scale. A differential equations employ the notation of differentiation and density of a time scale by delta derivative of a function at time t, while STSL _{PC} applies the time interval to express a duration. When STSL involves in dense time, we can use the notation of “time scale”, but we never mention the notation of “differentiation”. This paper is an expanded version of the SEKE 2019 conference paper [15], and includes all the notions required for the construction for proving the decidability. We reorganize the work to make the idea more clear for readers. Also, we provide incomplete and undecidable result of STSL _{OC} and prove the result.
In this work, there are three contributions:

1
We propose a spatiotemporal specification language STSL _{PC}, based on STL and S4 _{u}, to specify the changes in topometric space and dense time. We present STSL _{PC}, and provide syntax, Boolean semantics and quantitative semantics,

2
We present an axiomatization system and prove the completeness and decidability for the proposed STSL _{PC}.

3
We extend the expressiveness of STSL _{PC}, called STSL _{OC}, to the changes or evolution of spatial objects over time, and prove the incompleteness and undecidability.
The next section introduces temporal logic STL and modal spatial logic S4 _{u}. “Spatiotemporal specification language” section presents the spatiotemporal specification language STSL _{PC}, and completeness of axiomatization system and decidability of STSL _{PC} is proved in “Completeness and decidability of STSL _{PC}” section. In “STSL _{OC}” section, we present STSL _{OC} through extending the expressiveness of STSL _{PC}, and the completeness and complexity are provided. “Case study” section presents a case study about train collision avoidance system. “Related work” section compares the related works. We conclude the work and talk about the future work in “Conclusion and future work” section.
Preliminary
The section provides the background to the proposed spatiotemporal logic, including signal temporal logic (STL) and spatial logic S4 _{u}.
Spatial logic: S4 _{u}
S4 [20] is a proposition modal logic and τ is a spatial term under the interpretation of topological space. In the absence of ambiguity, the terminology a spatial term denotes a spatial object. According to the observation by [21], S4 is a logic of topological spaces, and the propositional variable is interpreted as an element of a subset of the topological space. From the perspective of the topometric space [9], propositional variables of S4 will be understood as spatial variables [22]. In this paper, we restrict the formula of S4 to topometric space. The syntax of S4 can be defined on the topometric space as follows:
where p is a spatial variable on topometric space and \(\overline \tau \) is the complementary of τ,τ_{1}⊓τ_{2} the intersection operation of τ_{1} and τ_{2}. \(\mathbb {I}\) is an interior operator under the topometric space interpretation. The union and closure operator can be defined by:
\(\mathbb {C}_{\tau }\) refers to the closure of a spatial object τ. The 1dimensional interpretation is shown in Fig. 1.
Let \(\mathfrak {L} = (M, d)\) is a metric space, where M is a nonempty set denoting the universe of the space, and d is the metric operator on the elements of M, i.e., a function \(d:M\times M\rightarrow \mathbb {R}\) such that for any spatial objects x,y,z∈M, the equations follow d(x,y)=0⇒x=y,d(x,y)=d(y,x) and d(x,z)≤d(x,y)±d(y,z). A metric model is a pair of the form \(\mathfrak {M} = (\mathfrak {L}, \mathfrak {V}(d))\), where \(\mathfrak {V}(d)\subseteq M\), denotes a set of valuations on the metric of spatial variables. A topometric space is a tuple \((M, \mathbb {I}_{d})\), where \(\mathbb {I}_{d}\) is an interior operator on M induced by the metric space (M,d), and \(\forall X\subseteq M, \mathbb {I}_{d}(X)=\left \{x\in X \mid \exists a >0 \; \forall y\; \left (d(x,y)< a \rightarrow y\in X\right)\right \}\). The topometric model is defined as \(\mathfrak {M}=\left (M, d, \mathbb {I}_{d}, P_{1}^{\mathfrak {M}}, P_{2}^{\mathfrak {M}}...\right)\), where \(\mathfrak {M}=\left (M, d, P_{1}^{\mathfrak {M}}, P_{2}^{\mathfrak {M}}...\right)\) is a metric model and \(\mathbb {I}_{d}\) is the interior operator induced by (M,d). Therefore, we get the valuation of other spatial formulas as follows:
The interpretation of regionbased spatial entities and their relations between them in 2dimensional space can be seen in Fig. 1. In the figure, a spatial entity τ is present and the spatial complementary, interior, closure operators is defined on τ, where the spatial term U means the universal set. Meanwhile, the spatial union and intersection on spatial terms τ_{1} and τ_{2} are present.
A spatial logic is a formal language interpreted over a class of structures featuring geometrical entities and relations. Among the wellknown spatial logics such as RCC8 [23, 24], BRCC8 and S4 _{u}, the most expressive spatial form is S4 _{u} [25]. S4 _{u} extends S4 with the universal and existential quantifiers , based on a spatial term τ. refers to that there is at least one element in space τ, and means that all elements in the space belong to τ. The formula φ is defined in the form of BNF:
where, ¬φ is the negation of φ and φ_{1}∧φ_{2} the conjunction of φ_{1} and φ_{2}.
Correspondingly, the disjunction and existential operators and the spatial subset \(\sqsubseteq \) can be derived by:
The axiomatization system for S4 _{u} includes the classical propositional logic in topology \(\mathfrak {L}\), the modal logic S4 and extended universal operator and inference rules.

CP
Axioms of classical propositional logic in \(\mathfrak {L}\)
and the inference rules
The soundness and completeness of S4 _{u} is given in [26]. It is worth noting that the current axiomatization is considerably different from the version in [26]. For one thing, the current logic S4 _{u} can be used to interpret spatial logic, while Shehtman’s work only employs it in modal logic. The interpretation in spatial domain makes S4 _{u} enjoy more meanings. For another, we want to use soundness and completeness result of the axiomatization as a basic to consider quantitative axioms temporal axioms.
Signal temporal logic
LTL is proposed by Pnueli [27] to specify sequential and parallel programs. The logic is built on a finite set P of propositional letters. The Boolean connectives and temporal operators are defined based on the propositional letters. When the domain of time is extended from discrete time to dense time, the logic metric interval temporal logic (MITL) [28] emerges. While STL extends the signals of MITL [29] from Boolean value to real value. An STL signal [10, 30] is defined on densetime domain \(\mathbb {T}\), which depends on the sampling times and frequency within an interval. A \(signal\; function \varepsilon \colon \mathbb {T}\rightarrow \mathbb {E}\) associates a set of time domain with a set of signals. Signals with \(\mathbb {E} = \mathbb {B} = \{0, 1\}\) are called Boolean signals, while these with \(\mathbb {E} = \mathbb {R}^{+}\) are called realvalued or quantitative signals. A Boolean signal, transformed from realvalued one through a set of predicates, can be represented by MITL [31].
An execution trace w is a set of realvalued signals \({x^{w}_{1},..., x^{w}_{k}}\) bound in some interval I of \(\mathbb {R}^{+}\), which is called the time domain of w [11]. We constrain such an interval \(I \subseteq \mathbb {R}^{+}\) to be halfclosed and halfopen [t_{1},t_{2}). The syntax of STL is given by
where ap∈AP is a atomic predicate and AP a finite set of atomic predicates {x_{i}⋈c⋈∈{<,≤,≥,>}} whose truth value is determined by the sign of an evaluation based on the signal x_{i}. Let y_{i}=x_{i}−c, an atomic predicate with the format x_{i}≥c can be translated as y_{i}≥0. The Boolean operators ¬ and ∧ are negation and conjunction, respectively. The temporal bounded until operator \(\mathcal {U}_{I}\) is defined on the time interval I. The bounded temporal operators □_{I} and ♢_{I}, and binary disjunction ∨ can be derived as follows:
Formula ♢_{I}φ indicates that φ is eventually satisfiable within the time interval I, while □_{I}φ denotes that φ is always satisfiable.
The fundamental composition of the axiomatization system for LTL contains all the tautologies like atomic proposition, Boolean operators in firstorder logic. Temporal expressiveness and inference rules are shown as follows:

A0
All classical tautologies of firstorder logic

A1
□(ϕ→φ)→(□ϕ→□φ)

A2
¬○ϕ⇔○¬ϕ

A3
○(ϕ→φ)→(○ϕ→○φ)

A4
□(ϕ→○φ)→(ϕ→□φ)

A5
(ϕUφ)⇔φ∨○(ϕUφ)

A6
(ϕUφ)→♢φ
and the inference rules:

MP \(\frac {\phi \quad \phi \rightarrow \varphi }{\varphi }\)

N_{□} \(\frac {\phi }{\vdash \Box \phi }\)

N_{○} \(\frac {\phi }{\vdash \bigcirc \phi }\)
Gabbay et al. [32] present the completeness of the deductive systems of LTL. Also, Lichtenstein and Pnueli [33] prove the complete system of LTL from three parts: the general part, domain part and program part. The temporal operator next expresses dynamic behaviors in discrete time, so it is unsuitable to express dense time. So the axioms A2A6 and the inference rule N_{○} will be ignored for the axiomatization of STL.
Spatiotemporal specification language
STL provides an approach that combines the truth value and quantitative value of general signals. But it is inadequate to represent the changes of a spatial entity and the binary relation between spatial entities and temporal aspects. We propose the spatiotemporal specification language that combines STL and S4 _{u} to describe the evolution in spatial and temporal domain.
Spatiotemporal signals
A spatiotemporal signal is defined with continuous time and topometric space [29, 34]. A realvalued interval [0,t) is defined in the dense time domain \(\mathbb {T}\), where \(t\in \mathbb {R}_{\geq 0}\). Because the time domain inherits STL, it will keep consistency with the interval in STL. The signal function is extended to spatialtemporal domain with \(\varepsilon \colon \mathbb {T}\times \mathbb {L} \rightarrow \mathbb {E}\), where \(\mathbb {L}\) denotes the topometric space. Firstly, an elementary signal evaluates a spatial entity or the connections between spatial entities. That is, the elementary signal is quantitative signals Secondly, a Boolean signal can be transformed from a quantitative signal by the threshold predicate x_{i}≥0. The signals stem from topometric space, therefore, the Boolean and quantitative signals are extended from the domain of STL signals to topometric space.
An atomic spatial entity enjoys two meanings: pointbased interpretation, point setbased interpretation and regionbased interpretation. The points interpret discrete location coordinates. It is consistent with your comment. But, the point setbased interpretation and regionbased interpretation denote the discrete and continuous space respectively. When involving in changes of spatial entities, we are talking about motion. The motion of spatial entities can be discrete or dense time. We define a spatiotemporal trace w as the changes of spatial objects over time. Formally, a spatiotemporal trace assigns \(\mathbb {T}\times \mathbb {L}\) to a multidimensional signal \(\mathbb {R}^{n}\), where n refers to the number of variables.
A spatiotemporal trace provides a notation about execution sequence of temporal and spatial domain.
Definition 3.1.
(Spatiotemporal signal) A spatiotemporal signal ε is an evaluation of spatial entities in a trace w. A Boolean signal \(\mu _{i}^{w(t)} (i\in \mathbb {N})\) is an evaluation of an atomic proposition transferred from quantitative signals \(\mathbf {x}_{i}^{w(t)}\) by atomic predicate \(\mu _{i}^{w(t)} = \left (\mathbf {x}_{i}^{w(t)}\geq 0\right)\) in the trace.
where \(\mathbb {B}\) refers to the domain of Boolean signals and \(\mathbb {R}_{\geq 0}\) quantitative signals.
For a spatiotemporal trace w, there are two different interpretations:

A trace represents a sequence of spatial objects and time point and each point in the trace evaluates a pair of spatial objects and time.

Another interpretation means that a spatiotemporal trace takes spatial objects as the basic entities and spatiotemporal primitive relations could be obtained by the changes of ontology of space over time.
In this work, we treat the spatiotemporal trace as the second interpretation. The changes of spatial objects are influenced by the flow of time.
The interpretation of the combined logic
It is essential that a combined spatiotemporal form should be provided with enough expressiveness to contain the three parameters [22]:

1
the expressiveness of the spatial component;

2
the expressiveness of the temporal component;

3
the interaction between the two components allowed in the combined logic.
Based on the principle of PC [22], which expresses that the language should be able to express changes over time of the truthvalues of purely spatial propositions. We interpret an STSL _{PC} formula based on the topometric temporal model, which is defined on topometric space and temporal interval structure in strict partial ordering with a set of sampling time point. The model can be treated as a set of sampling trace monitoring from statebased cyberphysical systems, rather than differential equations [18]. Firstly, the scene snapshot of a system is abstracted to be a topometric model. As the system executes, the system is sampled as a sequence of traces at dense time. At each time instant, the spatial structure denotes a topometric model. And the topometric temporal model in dense time and topometric space can be a sequence of sampling traces. Formally, a topometric temporal model is defined as a triple \(\mathfrak {M} = (\mathfrak {T}, \mathfrak {L}, \mathfrak {V})\), where

\(\mathfrak {T}\) is an interval structure \((\mathcal {T}, \mathcal {I}(\mathcal {T}))\), where \(\mathcal {T} = (\mathbb {T},<)\) is strict partial ordering with a set of time point \(\mathbb {T}\) and < an irreflexive, transitive and asymmetric relation on \(\mathbb {T}\) with a linear strict time flow, and \(\mathcal {I}(\mathcal {T})\) is a set of intervals,

\(\mathfrak {L}\) is a topometric space with the definition of \((M, \mathbb {I}_{d})\) in which M is a nonempty set, the universe of the space, and \(\mathbb {I}_{d}\) is the interior operator on M induced by the metric space (M,d), which satisfies the standard Kuratowski axioms [35]: \(\forall X, Y\subseteq M, \mathbb {I}(X\cap Y) = \mathbb {I}X\cap \mathbb {I}Y, \mathbb {I}X\subseteq \mathbb {II}X\) and \(\mathbb {I}(M) = M\),

\(\mathfrak {V}\) is a valuation on the time domain \(\mathfrak {T}\) and the spatial term set \(\mathbb {L}\), i.e., \(\forall \tau \in \mathbb {L}\), and \(t \in \mathbb {T}\). Formally, \(\mathfrak {V}(\tau, t) = \left \{\mu _{i} \mid \forall i\in \mathbb {N}, x_{i}\geq 0\right \}\) means the space occupied by a spatial term τ at time point t. As for the spatial term τ, the valuation can be defined as: \(\mathfrak {V}(\overline {\tau }, t) = \neg \mathfrak {V}(\tau, t), \mathfrak {V}(\tau _{1}\sqcap \tau _{2}, t) = \mathfrak {V}(\tau _{1},t)\cap \mathfrak {V}(\tau _{2},t), \mathfrak {V}(\mathbb {I}\tau, t) =\mathbb {I}\mathfrak {V}(\tau,t)\).
The ontology of space includes the static spatial entities and dynamic spatial entities. We describe the ontology of static spatial entities with S4 _{u} atomic spatial terms and spatial operators like complementary, intersection and union, interior and closure. And the spatial until operator is employed to represent the dynamic spatial ontology. However, the changes of ontology of space means that the evolution of spatial entities can be changed by external event or the arrival of a time slice. The changes of spatial terms in topometric model over time can be shown in Fig. 2.
A topometric temporal model is an abstraction of a cyberphysical system, while a spatiotemporal trace is an execution of topometric temporal model. Generally, classical model checking provides an approach to verify whether a topometric temporal model, i.e., all the traces, satisfies spatiotemporal properties. However, the state space explosion makes it difficult for model checking to verify the reliability and security of real cyberphysical systems. Monitoring verifies whether the spatiotemporal signals on one execution of the system hold the specified spatiotemporal properties. Especially, online monitoring can provide the verification results to help analyze the potential hazards, which can avoid unnecessary loss.
Example 3.1.
One point represents a spatial entity and an edge between two points means the connection between spatial entities. The weight on the edge denotes the metric between spatial entities. At any time instant, the points, edges and weight on the edges consist a undirected weighted graph, which denotes the topological metric models. As time goes, the topometric model leads to the topometric temporal model. In this figure, a spatial term τ is characterized as a 2dimensional space and the spatial terms change over time in the model \(\mathfrak {M}\).
Example 3.2.
The changes of spatial terms in topometric model over time can be shown in Fig. 3. In this figure, a spatial term p is characterized as a 2dimensional space and the spatial terms change over time in the model \(\mathfrak {M}\). The model describes the changes of spatial relations. The y axis describe the spatial relations. In the x axis, the time instant samples the spatial relations between the spatial entities red, blue and green. The black time instants describe the tangential proper part (TPP) relations between three regionbased spatial entities green and red. While the red time instants express the partial overlap (PO) relations between green and red. The spatial relations are expressed by S4 _{u} terms and the spatial relations are sampled with dense time and represented in STL.
Firstly, a cyberphysical system involves in discrete and continuous time. The cyber system describes the execution of an actual system, and the signal is sampled in discrete time. The physical system generally expresses the continuous changes of a spatial entity. A topometric temporal model is an abstract of a cyberphysical system. A spatiotemporal trace is sampled from the topometric temporal model. And the trace is an execution of a cyberphysical system. Further, when the trace describes spatial relations among spatial entities and is sampled in dense time, it will be able to be monitored by STSL formulas.
The concurrency can be interpreted in topometric model. Mainly, we interpret the region through borrowing the notation from Milner’s Bigraph [36], which presents the agent and the communication between different agents. We treat a spatial entity as an agent. Meanwhile, we define the concurrence of topometric model as the concurrency between spatial entities. We present the topometric model in regionbased and pointbased interpretation. So, we will illustrate the model from two perspectives.
Firstly, the pointbased topometric model can consider a list of regions. For instance, S4 _{u} terms can specify the properties that the region made by spatial entities a_{6},a_{7} and a_{8} exists in the region occupied by the union of the spatial entities a_{5},a_{6},a_{7},a_{8} and a_{9} as:
Another properties can be that the spatial entity a_{5} belongs to the intersection of the region that made up of spatial entities a_{1},a_{2},a_{3},a_{5} and a_{6}, and the region that made up of spatial entities a_{5},a_{6},a_{7},a_{8} and a_{9}. The property can be specified with S4 _{u} as:
From the perspective of agent, the spatial entities are hierarchical, which lead a treelike structure. The treelike structure is a place graph in Bigraph. Here, we didn’t define the interfaces or names between spatial entities for communication. Instead, we define the metric between the spatial entities. If we ignore the interfaces or names, the graph in Fig. 4 can be treated as a link graph. The place graph and link graph from topometric model compose the Bigraph. So, the topometric model is able to express concurrency.
Secondly, regionbased topometric model is relatively concise with spatial complementary, intersection, union, interior, closure and until operators. So it is easier to be interpreted. Figure 5 shows the spatial relations between the spatial entities green, red and blue.
Based on the notation of Bigraph with sharing [37], the overlapping part between two spatial entities can be treated as the sharing part. The hierarchical relations between regionbased topometric model are more clear. For instance, we can say that the spatial entity blue and green are concurrent, and their parent spatial entity is concurrent with the red entity.
The spatiotemporal signals are divided into Boolean and quantitative signals. According to the category of spatiotemporal signals, we will present syntax and semantics for the proposed spatiotemporal specification language from two sides:

The Boolean semantics returns true or false depending on whether the trace of topometric temporal model satisfies the properties or not.

The quantitative semantics returns a real value in different time that can be interpreted as an evaluation of satisfaction.
The Boolean semantics of the spatiotemporal specification language interprets that an STSL _{PC} formula over spatiotemporal traces returns true or false, so it is able to express purely spatial propositions’ changes with the truthvalues. Meanwhile, the \(\mathcal {U}_{I}\) and □_{I} operators of the quantitative semantics of STSL _{PC} are able to express satisfaction degree of spatial entities over some fixed finite periods and the whole duration of time, respectively.
The syntax of STSL _{PC}
As usual, we define the realvalue interval in temporal domain. Formally, we confine the temporal interval I to be leftclosed rightopen \([t, t'), \forall t, t'\in \mathbb {T}\) and t<t^{′}. The STSL _{PC} fuses the temporal logic STL and modal spatial logic S4 _{u} so that the language can express the changes of purely spatial propositions over time. Specifically, the language is defined on spatial terms τ and spatial operators complementary, intersection, union, interior and closure, atomic predicates, Boolean connectives and temporal operators globally, finally and until over the temporal interval I. There are two kinds of atomic predicates: the binary subset operator of two spatial terms \(\tau _{1}\sqsubseteq \tau _{2}\) and the threshold predicates on a signal x_{i}≥0. The spatial subset relation describes the relations between regionbased spatial entities. Specifically, the binary subset can be derived by the unary spatial operator universal with the form , where ⊤ denotes the spatial universal set. The quantitative signals evaluate the spatial entities. The Boolean signals can be achieved from the quantitative signals by the threshold predicates. The atomic predicate \(\tau _{1}\sqsubseteq \tau _{2}\) means that the elements in the spatial term τ_{1} must belong to τ_{2}. And x_{i}≥0 is a threshold predicate, which transfers general realvalued signals to Boolean value. The syntax of STSL _{PC} is given by:
\(\tau ::= p \mid \overline {\tau } \mid \tau _{1}\sqcap \tau _{2} \mid \mathbb {I}\tau \)
\(\varphi ::= \tau _{1}\sqsubseteq \tau _{2} \mid x_{i}\geq 0 \mid \neg \varphi \mid \varphi _{1}\wedge \varphi _{2} \mid \varphi _{1}\mathcal {U}_{I}\varphi _{2}\)

τ is a spatial term,

p is an atomic spatial variable,

\(\overline {\tau }\) is the complementary of τ,

τ_{1}⊓τ_{2} is the intersection of τ_{1} and τ_{2},

\(\mathbb {I}\) is the interior operator under the topometric space interpretation. Moreover, the dual operator of \(\mathbb {I}\) is the closure operator \(\mathbb {C}\), which means possible or consistent,

\(\tau _{1}\sqsubseteq \tau _{2}\) implies the spatial subset relations, which means that for all points p,p∈τ_{1} implies p∈τ_{2},

x_{i}≥0 is an atomic predicate,

¬,∨ and ∧ are the Boolean operators,

\(\mathcal {U}_{I}\) is the until operator.
We can define equivalence of operators as syntactic abbreviations:
Atomic predicates, Boolean operators, and the temporal bounded until operator \(\mathcal {U}_{I}\) are from STL. The new spatial operators are the interior operator \(\mathbb {I}\) and the closure operator \(\mathbb {C}\) with reference to S4 _{u}. The ♢_{I} and □_{I} operators are derived unary operators. □_{I}φ denotes that φ holds within the whole interval I, and ♢_{I}φ means that φ holds in at least one time point of the interval I.
Example 3.3.
In the mobile blocking mode, the protection point is located behind the forward train to protect the position uncertainty and the back boundary. The protection point is the zero speed limit point, which is a limit that is absolutely not allowed to be crossed by the signal system control. The core of the train’s automatic protection is the safe braking model of the train. It describes how to calculate the emergency braking curve and the normal braking curve of the train in Fig. 6. The emergency braking curve considers the emergency braking deceleration of the train protection, the current protection point calculated by the trackside ATP, the most restrictive speed curve and the slope section of the line. The normal braking curve takes into account the ATP response delay time and the cutoff traction after emergency braking. The ATP onboard computer unit dynamically calculates and continuously monitors the normal braking curve. The safe braking model ensures that the train will not exceed the most restricted speed and the train will stop in front of the protection point. Formally, the location of end of mobility authority is marked with EoA, and that of the end of protection is marked with EoP. The distance from the location of the following train to EoA denotes τ_{ma}. The distance from the location of the beginning of following train to EoP means τ_{p}. Also, the braking distance in normal and emergent braking mode is represented as τ_{nb} and τ_{eb}. A spatiotemporal property can be expressed as After receiving the signals, the train brakes in the normal braking mode and it keeps running in the region of τ_{ma}. While the train keeps running without braking in emergent braking model within 10 seconds. After the train brakes in the emergent braking mode, there exists a moment that the velocity of normal braking is larger than that of the emergent braking mode. And the train keeps running in the emergent braking mode in the region of τ_{p} within 40 seconds. The property can be specified with STSL _{PC} formula as:
where a_{nb} and a_{eb} denote the acceleration in normal and emergent braking mode, respectively. v_{nb} and v_{eb} represent the velocity in normal and emergent braking mode, respectively.
Example 3.4.
In the example 3.1, we can employ an STSL _{PC} formula to specify the properties that the region made by spatial entities a_{6},a_{7} and a_{8} exists in the region occupied by the union of the spatial entities a_{5},a_{6},a_{7},a_{8} and a_{9}, until the region made by spatial entities a_{3},a_{6} and a_{8} exists in the region occupied by the union of the spatial entities a_{3},a_{5},a_{6},a_{7} and a_{8} as
Example 3.5.
In Fig. 3 of example 3.2, we can specify the property that for any execution within 4 seconds, there is an inclusion between two spatial entities τ_{blue} and τ_{grey}, and it will follow an overlapping between them within 1.5 seconds. The property can be specified with STSL _{PC} formula as
The semantics of STSL _{PC}
The semantics of STSL _{PC} is divided into Boolean semantics and quantitative semantics, which return the truth value of purely spatial propositions and realvalued spatial objects. We define μ_{i} as a predicate because the spatial entities are a discrete set and each signal is evaluated by a threshold predicate μ_{i}. The quantitative semantics can be transformed to Boolean semantics by a predicate μ_{i}.
The spatial element of the spatiotemporal specification language exists in spatial entity τ. The value of the spatial entity τ can be achieved by the definition \(\mathfrak {V}(\tau, w, t)\). That is, we interpret a spatiotemporal formula in simulated trace rather than topometric temporal model. The satisfaction relation for an STSL _{PC} formula φ over a topometric temporal model \(\mathfrak {M}\) is given by:

\((w, t)\models \tau _{1}\sqsubseteq \tau _{2} \; \Leftrightarrow \; \mathfrak {V}(\tau _{1}, w, t) \leq \mathfrak {V}(\tau _{2}, w, t)\)

\((w, t)\models x_{i}\geq 0 \; \Leftrightarrow \; \mathfrak {V}(x_{i}, w, t)\geq 0\)

(w,t)⊧¬φ ⇔ (w,t)⊮φ

(w,t)⊧φ_{1}∧φ_{2} ⇔ (w,t)⊧φ_{1} and(w,t)⊧φ_{2}

\( (w, t)\models \varphi _{1}\mathcal {U}_{I}\varphi _{2} \; \Leftrightarrow \; \exists t'\in t + I \text { s.t.} (w, t') \models \varphi _{2} \text { and} \forall t''\in [t, t'], (w, t'')\models \varphi _{1}\)
A trace w satisfies an STSL _{PC} formula φ at t, denoted by (w,t)⊧φ. For the satisfaction relation, the “ ⇒” answers whether the implementation procedure of computation holds the specification relation. The “ ⇐” can be achieved from the definition of satisfaction relation of an STSL _{PC} formula φ.
For a given formula φ and execution trace w, we define the satisfaction signal χ(φ,w,t) over a trace w(t,l):
where ⊤ and ⊥ respectively denote Boolean value true and false. Therefore, \(\chi (\tau _{1}\sqsubseteq \tau _{2}, w, t)\) returns true if spatial subset relation \(\tau _{1}\sqsubseteq \tau _{2}\) holds over the model. For a general signal x_{i},χ(x_{i},w,t) returns true if x_{i}≥0 in the trace w(t). χ(□_{I}φ,w,t) means that for all t∈I,φ always returns trues in the interval I over the model. While χ(♢_{I}φ,w,t) denotes that there exists t∈I,φ returns trues in the interval I over the model.
In order to compute the satisfaction of a formula φ, we divide the formula φ into each subformula ϕ_{i} until atomic formula so that formula φ can be computed through the subformula and atomic formulas instead of the entire satisfaction signal χ(φ,w,t). The procedure can be treated as a hierarchical structure from the full formula φ down to each atomic formula.
We define ρ to quantify the satisfaction degree of the property φ over the trace w(t), and it returns a real value ρ(φ,w,t). For an atomic spatial formula , the satisfaction degree can be evaluated as \(\mathcal {V}(\tau, w, t)\). And for a atomic predicate x_{i}≥0 can be evaluated as \(x_{i}^{w(t)}\). The quantitative satisfaction relation for a formula φ over a spatiotemporal trace w at the time t by the notation of satisfaction degree is given by:

.

\(\rho (x_{i} \geq 0, w, t) \; = \; x_{i}^{w(t)}\)

ρ(¬φ,w,t) = −ρ(φ,w,t)

ρ(φ_{1}∧φ_{2},w,t) = min{ρ(φ_{1},w,t),ρ(φ_{2},w,t)}

ρ(φ_{1}∨φ_{2},w,t) = max{ρ(φ_{1},w,t),ρ(φ_{2},w,t)}

ρ(□_{I}φ,w,t) = inft^{′}∈t+I{ρ(φ,w,t^{′})}

ρ(♢_{I}φ,w,t) = supt^{′}∈t+I{ρ(φ,w,t^{′})}

\(\rho (\varphi _{1} \mathcal {U}_{I}\varphi _{2}, w, t) = \sup _{t'\in t + I}(\min \{\rho (\varphi _{2}, w, t'), \inf _{t''\in [t,t']}(\rho (\varphi _{1}, w, t'')\})\)
The negation of a formula is evaluated as the negative of its satisfaction degree. The conjunction and disjunction of two formulas are evaluated as the minimum and maximum of the satisfaction degree of the two formulas. ρ(□_{I}φ,w,t) refers to that the infimum of ρ(φ,w,t^{′}),∀t^{′}∈t+I is always true in the interval I over the trace. Similar to ρ(□_{I}φ,w,t),ρ(♢_{I}φ,w,t) returns the truth value of the supremum of ρ(φ,w,t^{′}),∀t^{′}∈t+I. The satisfaction degree of until formula \(\varphi _{1} \mathcal {U}_{I}\varphi _{2}\) is evaluated complexly. Firstly, we achieve the satisfaction degree of the formula φ_{2} in the time t^{′}, which belongs to the interval t+I. Secondly, the infimum of the formula φ_{1} is evaluated in the interval of [t,t^{′}). Thirdly, the minimum of the result of first steps is achieved. At last, the satisfaction degree of until formula \(\varphi _{1} \mathcal {U}_{I}\varphi _{2}\) is evaluated as supremum of the minimum in the third step. It is worthy noting that the infimum of the satisfaction degree of a formula means that the minimum value of the signals within temporal interval I. Similarly, the supremum of the satisfaction degree of a formula denotes the maximum value of signals with a temporal interval I.
The connection between Boolean and quantitative signals is built by the way of predicate x_{i}≥0 and obtain the satisfaction signal χ(x_{i}≥0,w,t), which returns a real value of the quantitative signals x_{i} representing the distance to satisfaction. Specifically, the satisfaction degree of quantitative signal can be derived from the Lemire’s algorithm [38] through MAXMIN filter of a running sequence.
Example 3.6.
In Fig. 3, we can specify the property that for any execution within 4 seconds, there is an inclusion between two spatial entities τ_{blue} and τ_{grey}, and it will follow an overlapping between them within 1.5 seconds. The property can be specified as
Completeness and decidability of STSL _{PC}
The language STSL _{PC} describes the spatial changes in dense time, which is used to monitor the execution of continuous systems at run time. So we ignore the temporal next operator that describes the next step of the discrete systems.
An axiomatization system of STSL _{PC}
We will present a Hilbertstyle proof system for STSL _{PC} according to expressiveness of the proposed language. STSL _{PC} combines temporal logic and spatial logic, so the proof system will be introduced from spatial and temporal part. Further, the temporal operators □_{I},♢_{I} and \(\mathcal {U}_{I}\) are defined in an interval I in dense time, so we will add the quantitative part for the proof systems.
Spatial part
The spatial part of the axiomatization presents the subset relation of spatial terms with complementary, intersection and union operators:

S0
\(\mathbb {I}\tau \sqsubseteq \tau \)

S1
\(\tau \sqsubseteq \mathbb {C}\tau \)

S2
\(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \overline {\tau _{2}}\sqsubseteq \overline {\tau _{1}}\)

S3
\(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \mathbb {C}\tau _{1}\sqsubseteq \tau _{2}\)

S4
\(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\)

S5
\(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \mathbb {I}\tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\)

S6
\(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \mathbb {C}\tau _{1}\sqsubseteq \mathbb {C}\tau _{2}\)

S7
\(\tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\rightarrow \tau _{1}\sqsubseteq \mathbb {C}\tau _{2}\)

S8
\((\tau _{1}\sqcup \tau _{2})\sqsubseteq \tau _{3}\rightarrow (\tau _{1}\sqsubseteq \tau _{3})\vee (\tau _{2}\sqsubseteq \tau _{3})\)

S9
\(\tau _{1}\sqsubseteq (\tau _{2}\sqcap \tau _{3})\rightarrow (\tau _{1}\sqsubseteq \tau _{2})\wedge (\tau _{1}\sqsubseteq \tau _{3})\)
 S10
 S11
 S12
And the inference rules:

\(N_{\sqsubseteq }\) \(\frac {\tau _{1}\sqsubseteq \tau _{2} \quad \tau _{2}\sqsubseteq \tau _{3}}{\vdash \tau _{1}\sqsubseteq \tau _{3}}\)
Temporal part
The temporal part of STSL _{PC} in real time with interval implies that the temporal next operator is forbidden.

T0
All classical tautologies of propositional logic

T1
♢_{I}ϕ⇔¬□_{I}¬ϕ

T2
□_{I}(ϕ∧φ)→(□_{I}ϕ∧□_{I}φ)

T3
♢_{I}(ϕ∨φ)→(♢_{I}ϕ∨♢_{I}φ)

T4
ϕU_{I}φ→♢_{I}φ
And the inference rules:

MP \(\frac {\phi \quad \phi \rightarrow \varphi }{\varphi }\)

N_{□} \(\frac {\phi }{\vdash \Box _{I}\phi }\)
Axiom (T0) and Modus Ponens (MP) are from the Hilbertstyle axiomatization of propositional logic. Axioms (T1T4) are achieved from by Manna and Pnueli’s temporal logic [39]. A complete proof system for quantitative version is proposed in [40].
Quantitative part
The quantitative part of spatiotemporal logic involves the execution of system in dense time, so the quantitative axioms need to be provided. We follow the way of [41] to present the quantitative axioms for STSL _{PC}. However, we forbid the appearance of the punctuality in interval as the metric logic MITL and ban the temporal next operator because of the continuous time. The quantitative axioms characterize the translation from the intersection, union of intervals into conjunction, disjunction of temporal operators with interval. Specifically, the intersection of two intervals is bounded in finally operator with the form ♢_{I∪J}φ, and it implies the disjunction of finally operator bounded with their respective interval. Conversely, the disjunction of finally operator bounded with their respective interval implies the finally operator bounded with the union of the two intervals. However, the globally operator bounded with the intersection of the two intervals and the conjunction of globally operator bounded with their respective interval imply each other. For until operator, the axiom generalizes ϕU_{I}φ→♢_{I}φ with the union of two intervals I∪J.

Q0
♢_{I∪J}φ⇔♢_{I}φ∨♢_{J}φ

Q1
□_{I∩J}φ⇔□_{I}φ∧□_{J}φ

Q2
ϕU_{I∪J}φ→♢_{I∪J}φ
Soundness and completeness of the axiomatization system
Once an axiomatization system is present, the soundness and completeness of the axiomatization system need to be proved, including spatial, temporal and quantitative axioms. Soundness refers to that all the theorems in STSL _{PC} are logically valid. Equivalently, a spatiotemporal logic is sound with respect to topometric temporal model if for all the formulas \(\phi, \vdash _{\mathcal {ST}\phi }\) implies \(\Vdash \phi \). Let \(\mathcal {ST}\) be a class of topometric temporal modelred. A spatiotemporal logic is strongly complete in \(\mathcal {ST}\) if for any set of formulas Γ∪{ϕ}, if \(\Gamma \Vdash _{\mathcal {ST}} \phi \) then \(\Gamma \vdash _{\mathcal {ST}}\phi \). If the semantics of Γ satisfies ϕ on \(\mathcal {ST}\) then ϕ is deducible from Γ.
Theorem 4.1.
The above axiomatization is sound for topometric temporal model, i.e., for any ϕ∈STSL_{PC}, if ⊩_{Γ}ϕ implies \(\Vdash _{\mathcal {ST}}\phi \)
Proof
The soundness theorem proof need guarantee each axiom is sound and the inference rules preserve soundness from the sound hypothesis. This follows the fact that all axioms are valid and all rules preserve validity. We provide the proof in Appendix A. □
It is wellknown that weak completeness plus compactness implies strong completeness [42]. The lexicographic products of modal logic with linear temporal logic are sound and complete [43]. MTL, which empowers more expressiveness in punctuality operator than MITL, is complete in twosorted model [44]. But temporal logic in the flow of real time has weak completeness [45], which proposes finitely complete and expressively complete, but fails compactness theorem. These conclusions contribute to the result of weak completeness.
Theorem 4.2.
The system for STSL _{PC} is weakly complete with respect to topometric temporal model, i.e., for every STSL _{PC} formulas, \(\Vdash _{\mathcal {ST}}\phi \) implies ⊩_{Γ}ϕ.
Before presenting the complete proof of the axiomatization system of STSL _{PC}, we will introduce the notation of maximal consistent set [46].
The axiomatization system of STSL _{PC} is a logical system. A proof in STSL _{PC} is a sequence of finite formula: A_{0},A_{1},...,A_{n}, where each of them is an axiom, or there exists j,k<i, such that A_{i} is the conclusion derived from A_{j} and A_{k} using MP inference rule. The last term A_{n} is a theorem in STSL _{PC}, using the sign ⊩A_{n}, where n is the length of proof.
The concepts of deducibility and consistency from [12, 47] are fundamental to deduce the logic system STSL _{PC}. A formula A is deducible from a set of formulas Γ in a system \(\mathcal {ST}\), written \(\Gamma \Vdash _{\mathcal {ST}}A\), if and only if \(\mathcal {ST}\) contains a theorem of the form (A_{1}∧...∧A_{n})→A, where the conjunctions A_{i}(i=1,...,n) of the antecedent are formulas in Γ. A set of formulas Γ is consistent in \(\mathcal {ST}\), written \({Con}_{\mathcal {ST}}\Gamma \), just in case the formula ⊥ is not \(\mathcal {ST}\)deducible from Γ.
Definition 4.1.
(\(\mathcal {ST}\)MCS) A set of formulas Γ is maximal \(\mathcal {ST}\)consistent iff

(i)
Γ is \(\mathcal {ST}\)consistent, and

(ii)
for every formula A, if Γ∪{A} is \(\mathcal {ST}\)consistent, then A∈Γ.
If Γ is a maximal \(\mathcal {ST}\)consistent set of formulas then we say it is an \(\mathcal {ST}\)MCS. The (ii) condition refers to that any set of formulas properly containing Γ is \(\mathcal {ST}\)inconsistent.
The canonical model is defined in [47] to induce the soundness and completeness of modal logics. We extend the notation of canonical model to spatiotemporal systems for completeness of STSL _{PC}.
Definition 4.2.
(\(\mathcal {ST}\)canonical Model) The \(\mathcal {ST}\)canonical model \(\mathfrak {M}^{\Gamma }\) for a spatiotemporal logic is a triple (W^{Γ},R^{Γ},V^{Γ}) where:

(i)
W^{Γ} is the set of all ΓMCSs;

(ii)
R^{Γ} is the metric relation on topometric space over a quasiorder on time. It is the canonical binary relation on W^{Γ} defined by \(s R_{i}^{\Gamma }s'\) over state s and s^{′} if for all formulas ϕ,ϕ∈s implies ϕ∈s^{′}.

(iii)
V^{Γ} is the valuation defined by V^{Γ}(p)={s∈W^{Γ}∣p∈s}. V^{Γ} is called the canonical valuation.
Lemma 4.3.
(Truth Lemma) Let \(\mathcal {ST}\)canonical model be a class of topometric temporal model. For all \(\phi \in \mathcal {ST}\)MCS, \(\mathcal {ST}\Vdash \phi \) iff \(\phi \in \mathcal {ST}\)MCS.
Proof
The proof is by induction on the structure of ϕ.
Base case: Suppose ϕ is a spatial formula or an atomic predicate x_{i}≥0.
\((\mathcal {ST}, s)\) ,
\((\mathcal {ST}, s)\Vdash x_{i}\geq 0 \Leftrightarrow V^{\Gamma }(x_{i}\geq 0,s)=x_{i} \Leftrightarrow x_{i}\geq 0\in s\).
Inductive step: Suppose ϕ is an atomic predicate \(\neg \phi, \phi _{1}\wedge \phi _{2}, \phi _{1}\vee \phi _{2}, \Box _{I}\phi, \Diamond _{I}\phi, \phi \mathcal {U}_{I}\varphi \). We show the proof of the case □_{I}ϕ, and leave the others to reader. We have \((\mathcal {ST}, s)\Vdash \Box _{I}\phi \Leftrightarrow \Box _{I}\phi \in s\) (assuming the inductive hypothesis).
\((\mathcal {ST}, s)\Vdash \Box _{I}\phi \)
\(\Leftrightarrow \forall s', sR^{\Gamma }s'\Rightarrow \mathcal {ST}, s'\Vdash \phi \)
⇔∀,sR^{Γ}s^{′}⇒ϕ∈s^{′}
we need to show that □_{I}ϕ∈s⇔∀s^{′},sR^{Γ}s^{′}⇒ϕ∈s^{′}.
⇒ follows immediately from the Definition 4.2.
As for ⇐: suppose □_{I}ϕ∉s. We need to show
∃s^{′},sR^{Γ}s^{′} and ϕ∉s^{′}
⇔∃s^{′},sR^{Γ}s^{′} and ¬ϕ∈s^{′}
⇔∃s^{′},{φ∣□_{I}φ∈s}⊆s^{′} and ¬ϕ∈s^{′}
⇔∃s^{′},{φ∣□_{I}φ∈s}∪{¬ϕ}⊆s^{′}
It is easy to show that {φ∣□_{I}φ∈s}∪{¬ϕ} is \(\mathcal {ST}\)consistent. Suppose not, i.e., {φ∣□_{I}φ∈s}∪{¬ϕ} is \(\mathcal {ST}\)inconsistent. Then \(\vdash _{\mathcal {ST}}(\varphi _{1}\wedge...\wedge \varphi _{n})\rightarrow \phi \) for some {□_{I}φ_{1},...□_{I}φ_{n}}⊆s. But \(\mathcal {ST}\) is canonical and s is \(\mathcal {ST}\)MCS, so s must contain (□_{I}φ_{1}∧...∧φ_{n})→□_{I}ϕ. From □_{I}φ_{i}∈s, it follows □_{I}ϕ∈s. This contradicts the hypothesis that □_{I}ϕ∉s □
The proof of the weakly complete system of STSL _{PC} is immediately the result of Lemma 4.3.
Decidability of of STSL _{PC}
We present the decidability of STSL _{PC} based on the finite model property [12]. A decision procedure for the decidable fragment will be present.
Definition 4.3.
(Filtration) Let \(\mathfrak {M}\) be the topometric temporal model and φ subformula closed set of formulas. ≈ is an equivalence relation on the states of \(\mathfrak {M}\) defined by:
(t,l)≈(t^{′},l^{′}) iff for all ϕ in φ: \((\mathfrak {M},t,l)\Vdash \phi \) iff \((\mathfrak {M},t',l')\Vdash \phi \).
We denote the equivalence class of a state with respect to ≈ by (t,l). Let \((\mathbb {T},\mathrm {U})=\{(t,l)\mid (t,l)\in (\mathbb {T},\mathrm {U})\}\). Suppose \(\mathfrak {M}^{f}\) is any model \((\mathfrak {T}^{f}, \mathfrak {L}^{f}, \mathfrak {V}^{f})\) such that:

\((\mathfrak {T}^{f},\mathfrak {L}^{f})=(\mathfrak {T}, \mathfrak {L})\).

if (t,l)≈(t^{f},l^{f}) then (t,l)≈(t^{f},l^{f}).

if (t,l)≈(t^{f},l^{f}) then for all ♢ϕ∈φ, if \((\mathfrak {M},t,l)\Vdash \phi \) then \((\mathfrak {M},t',l')\Vdash \Diamond \phi \).

\(\mathfrak {V}^{f}(p)=\{(t,l)\mid (\mathfrak {M},t,l)\Vdash p\}\) for all proposition letters p in φ.
Then \(\mathfrak {M}^{f}\) is a filtration of \(\mathfrak {M}\) through φ.
Proposition 4.1.
Let φ subformula closed set of STSL _{PC} formulas. For any model \(\mathfrak {M}\) if \(\mathfrak {M}^{f}\) is a filtration of \(\mathfrak {M}\) through a subformula closed set φ, then \(\mathfrak {M}^{f}\) contains at most 2^{n} nodes (where n denotes the size of φ).
Theorem 4.4.
(Filtration Theorem) Let \(\mathfrak {M}^{f} = (\mathfrak {T}^{f}, \mathfrak {L}^{f}, \mathfrak {V}^{f})\) be a filtration of \(\mathfrak {M}\) through a subformula closed set φ. Then for all formulas ϕ∈φ, and all nodes (t,l) in \(\mathfrak {M}\), we have \((\mathfrak {M},t,l)\Vdash \phi \) iff \((\mathfrak {M}^{f},(t,l))\Vdash \phi \)
Theorem 4.5.
(Finite Model Property) If ϕ is satisfiable, then it is satisfiable on a finite model. Indeed, it is satisfiable on a finite model containing at most z^{n}, where n is the number of subformulas of ϕ.
Proof
If ϕ is satisfiable in the filtration is immediate from Theorem 4.4, and the bound od size of the filtration is immediate from Proposition 4.1. It is wellknown in standard case [12, 16]. □
The satisfiability means for all STSL _{PC} formula ϕ there is a spatiotemporal trace in topometric temporal model \(\mathfrak {M}\) such that \((\mathfrak {M}, t,l)\Vdash \phi \). Theorem 4.5 shows the searching is finite in the nodes of topometric temporal model. So it follows Theorem 4.6.
Theorem 4.6.
The satisfiability problem for STSL _{PC} against topometric temporal model is decidable.
Further, we have the complexity of the decidable STSL _{PC}.
Theorem 4.7.
The satisfiability problem for STSL _{PC} against topometric temporal models \(\mathfrak {M}\) based on \((\mathbb {N},<), (\mathbb {Z}, <)\) is EXPSPACEcomplete.
Proof
Recall that STSL _{PC} refers to the changes over time of truthvalues of purely spatial propositions. The interaction between spatial and temporal components of STSL _{PC} is very restricted to the elementary unit: purely spatial propositions. The proof will be carried through construction step, reduction step and a decision procedure.
Construction step: For every STSL _{PC}formula φ, one can construct an STLformula ϕ by replacing every occurrence of a spatial proposition \(\tau _{1}\sqsubseteq \tau _{2}\) and atomic predicate x_{i}≥0 in φ with a fresh propositional variable. Specifically, given an STLmodel \(\mathfrak {M} = (\mathfrak {T},\mathfrak {V})\) and an STLformula ϕ with a time point t in \(\mathfrak {T}\), we construct the set
\(\phi _{t} = \{ \tau _{1}\sqsubseteq \tau _{2}\mid (\mathfrak {M},t)\models \varphi \}\cup \{ x_{i}\geq 0\mid (\mathfrak {M},t)\models \varphi \}\).
of spatial formulas as a collection of STL proposition variables. If ϕ_{t} is satisfiable for every t in \(\mathfrak {T}\), then there is a topometric temporal model \(\mathfrak {M}\) satisfying φ and based on the flow \(\mathfrak {T}\).
Reduction step: Definitely, one can obtain the formulas □_{I}φ and ♢_{I}φ from φ_{1}U_{I}φ_{2}, so checking whether an STL formula ϕ is satisfiable or not depends on the complexity of computing a formula with U_{I} operators. In fact, the time domain of STL is bound in the interval I, and an STL formula in Boolean signal returns true or false for a given STL model \(\mathfrak {M}\). The bound STL only with Boolean value has the same expressiveness with MITL [29]. It suffices to reduce the satisfiability problem of STL to the satisfiability problem for MITL [48] over infinite trace with interval to check satisfiability of ϕ_{t}. Further, STL is interpreted in a temporal trace, while MITL is interpreted in timed automata, which capture all trace. The Boolean semantics of STL returns Boolean value, which is equal to MITL. However, STL can be interpreted in quantitative semantics, which returns real value. It makes STL enjoy more powerful expressiveness. STL and MITL have the same expressiveness in temporal interval because the interval in both of them is dense time.
Decision Procedure: We divide the procedure for deciding the satisfiability of STSL _{PC} into two steps: the first is deciding spatial formula ϕ_{t} in PC, the second is dealing with STL. According to the observation [29], an STL formula in Boolean semantics is as expressive as an MITL formula, so an Satisfiability Modulo Theories (SMT) [49]based decision procedure for an MITL is suitable to decide an STL formula in Boolean semantics. We present a decision procedure to satisfiability checking of MITL, which is similar to the approach in [50]. The difference is that we restrict our MITL formula without past tense and the counting modality. We propose the encoding of MITL to Constraint LTL over clock (CLTLoc) [51], which is an extension of Constraint LTL [52] with clocks. □
Lemma 4.8.
([50]) Let M be a signal, and ϕ be an MITL formula. For any (π,σ)∈r_{sub(ϕ)}(M), we have \((\pi,\sigma),0\models \bigwedge _{\theta \in sub(\phi)}{ck}_{\theta } \wedge \bigwedge _{\theta =F_{(a,b)}(\gamma)} {auxck}_{\theta }\) and for all \(k\in \mathbb {N}, \theta \in sub(\phi)\), we have (π,σ),k⊧m(θ).Conversely, if \((\pi,\sigma),0\models \bigwedge _{\theta \in sub(\phi)}({ck}_{\theta }\wedge G(m(\theta))) \wedge \bigwedge _{\theta =F_{(a,b)}(\gamma)} {auxck}_{\theta }\), then there is a signal M such that (π,σ)∈r_{sub(ϕ)}(M).
The proof can be found in [50]. Let θ be the subformula of MITL formula ϕ, then θ is one of the form \(\neg \phi, \phi \wedge \varphi, \phi \mathcal {U}_{I}\varphi, \Diamond _{I}\phi \). The function m(θ) is defined to describe the translation of subformula θ of an MITL formula to a corresponding CLTLoc formula, which is the form:
The transformation from MITL to CLTLoc is implemented by qtlsolver [53]. The decision procedure of the CLTLoc formula is described in [54], which relies on the Zot toolkit [53]. In [50], it shows the satisfiability of an CLTLoc formula is PSPACE in the size of the formula and in the binary encoding of the constants, the decision procedure induced the encoding is in EXPSPACE.
STSL _{OC}
The interaction between the spatial and temporal components should comply with the principle of PC and OC [22], which is used to evaluate the interaction:

STSL _{PC}: the language should be able to express changes over time of the truthvalues of purely spatial propositions.

STSL _{OC}: the language should be able to express changes or evolution of spatial objects over time.
STSL _{PC} expresses the change of truthvalue of proposition and it is the elementary requirement for a combined spatiotemporal logic. For STSL _{OC}, spatiotemporal properties are specified about the changes of spatial objects over dense time with interval through extending the temporal globally, eventually, until operators to spatial terms of STSL _{PC}. The spatiotemporal trace comply the finite variability. The trace are divided into some intervals. The STSL _{OC} formulas with these intervals are also verified at runtime.
The difference between STSL _{PC} and STSL _{OC} exists that STSL _{PC} involves in the change of truthvalues of propositions, while STSL _{OC} describes the change of extensions of predicates. Specifically, STSL _{PC} expresses static spatial terms in topometric space over dense time through spatial complementary, intersection and union operators. The dynamic evolution of STSL _{O}C is achieved by admitting spatial until operators with interval [l_{1},l_{2}) to spatial terms in topometric temporal model.
The spatial until \(\tau _{1}\mathcal {U}_{[l_{1},l_{2})}\tau _{2}\) at spatial location l_{1} consists of those points x of the topometric space for which there is l_{2}>l_{1} such that x belongs to τ_{2} at moment l_{2} and x is in τ_{1} at all l whenever l_{1}<l<l_{2}. The difference between spatial until and temporal until exists in that spatial until is interpreted as spatial interval [l_{1},l_{2}), and temporal until is defined in a temporal interval [t_{1},t_{2}). Further, spatial until operates two spatial terms τ_{1} and τ_{2}, while temporal until operates two formulas φ_{1} and φ_{2}. The syntax of STSL _{OC} is given by:
\(\tau ::= p \mid \overline {\tau } \mid \tau _{1}\sqcap \tau _{2} \mid \mathbb {I}\tau \mid \tau _{1}\mathcal {U}_{[l_{1},l_{2})}\tau _{2}\)
\(\varphi ::= \tau _{1}\sqsubseteq \tau _{2} \mid x_{i}\geq 0 \mid \neg \varphi \mid \varphi _{1}\wedge \varphi _{2} \mid \varphi _{1}\mathcal {U}_{[t_{1},t_{2})}\varphi _{2}\)
Similar to the atomic formulas, the unary operators □_{I} and ♢ of spatiotemporal term can also be derived from the operator \(\mathcal {U}_{I}\). The occupied spaces of term □_{I}τ and ♢_{I}τ at moment w are interpreted as the intersection and union of all spatial extensions of τ at moments v>w, respectively.
Example 5.1.
In the Example 3.3, what we can also guarantee is that After the train brakes in the emergent braking mode, there exists a moment that the velocity of normal braking is larger than that of the emergent braking mode within 40 s. And the train keeps running in the normal braking mode in the region of τ_{nb} until it runs in the region of τ_{eb} within the spatial interval [p_{nb},EoA). The specification can be expressed as
where v_{eb} and v_{nb} mean the velocity of the emergent and normal braking mode.
Example 5.2.
The railway traffic with sensors, present by Liu et.al [2], provides a good perspective to discuss the proposed spatiotemporal specification language. In their example, a train and a control zone can be treated as a regionbased spatial terms. Obviously, the spatial region of control zone seems bigger than the region of train. Therefore, the spatial relation between control zone and the train can be characterized as S4 _{u} formulas or RCC8 relations, i.e., disconnected (DC), externally connected (EC), partial overlap (PO), equal (EQ), tangential proper part (TPP) and tangential proper part inverse (TPPI), and nontangential proper part (NTPP) and nontangential proper part inverse (NTPPI). The changes of spatial relations over dense time can be specified by STSL _{PC}. The specification a train intersects with the control zone until the train leaves within 10 min can be expressed as
where τ_{cz} means the region of control zone. The predicate PO(τ_{train},τ_{cz}) is equal to the STSL _{PC} formula (Iτ_{train}⊓Iτ_{cz}∧¬ \((\tau _{train}\sqsubseteq \tau _{cz})\wedge \neg \) \((\tau _{cz}\sqsubseteq \tau _{train})\), and the predicate DC(τ_{train},τ_{cz}) can be expressed by the STSL _{PC} formula .
However, it is not enough to specify spatiotemporal properties with STSL _{OC} because there is no evolution of spatial entities. Further, it can not even be specified with the spatial relation between pointbased spatial terms in topometric space.
Example 5.3.
In example 3.1, the spatiotemporal properties that the spatial entity a_{1} reaches a_{9} within 20, until the distance from the spatial entity a_{1} to a_{9} is reduced 18 within 8 min can be specified with STSL _{OC} as
The production of temporal logic and spatial logic expresses the spatial evolution in real time with spatial until. The axioms for STSL _{OC} will add the spatial until. The added operator has no influence on STSL _{OC}. So the STSL _{OC} enjoys the same soundness as STSL _{PC}.
Theorem 5.1.
STSL _{OC} isn’t complete with respect to topometric temporal model.
Proof
The incomplete STSL _{OC} can be proved by the counter example: \(\Box _{I}(\Diamond _{I}(\tau _{1}\sqcup \tau _{2})\sqsubseteq \Diamond _{I}\tau _{1}\sqcup \Diamond _{I}\tau _{2})\rightarrow \Box _{I}\Diamond _{I}(\tau _{1}\sqcup \tau _{2})\sqsubseteq \Box _{I}(\Diamond _{I}\tau _{1}\sqcup \Diamond _{I}\tau _{2})\). Although the axioms T2 can be applied into temporal deduction, the formula □_{I} on the spatial terms returns the quantitative space. □
A spatiotemporal trace w is a sequence over signals ε. The Boolean satisfaction relation and satisfaction degree for an STSL _{OC} formula φ over a spatiotemporal trace w are similar to that of STSL _{PC} formula.
The decidability of STSL _{OC} will follow:
Theorem 5.2.
The satisfiability problem for STSL _{OC} formulas based on \((\mathbb {N},<), (\mathbb {Z}, <)\) is undecidable.
Before providing the proof of Theorem 5.2, we present the lemma 5.3:
Lemma 5.3.
There exists a natural number \(\mathbb {N}\geq 1\) and a sequence i_{1},...,i_{N} of indices such that \(v_{i_{1}},...,v_{i_{N}}= w_{i_{1}},...,w_{i_{N}}\), then the satisfiability problem for STSL _{OC} formulas based on \((\mathbb {N},<), (\mathbb {Z}, <)\) is undecidable.
Proof
As we all know, Post’s correspondence problem is undecidable [55]. Given a finite alphabet A={a_{1},...,a_{m}} and a finite set P of pairs (v_{1},w_{1}),...,(v_{k},w_{k}) of nonempty finite sequences (words) v_{i},w_{i} over A, decide whether there exists an N≥1 and a sequence i_{1},...,i_{N} of indices such that \(v_{i_{1}},...,v_{i_{N}}= w_{i_{1}},...,w_{i_{N}}\).
An execution trace w is a set of STL signals \(\left \{x_{1}^{w},..., x_{k}^{w}\right \}\) defined over some interval D of \(\mathbb {R}^{+}\) [11]. Assume the realvalued signals are finite variability. The interval \([t_{i}, t_{i+1})_{i\in \mathbb {N}}\) and threshold predicates divide the execution trace to be piecewise Boolean signals.
We encode the satisfiability problem of STSL _{OC} formulas to Post’s correspondence problem as the way of [22]. The decidability with less expressive language initially is proved in [56]. We construct a formula φ_{A,P} which is STSL _{OC}satisfiable iff for each 1≤i≤k, let l_{i} and r_{i} be the length of words v_{i} and w_{i}, respectively, and let
The formula φ_{A,P} is construct as:
where
φ_{range}=range∧♢_{I}¬range∧□_{I}(¬range→□_{I}¬range)
\(\varphi _{pair} = \Box _{I}(\Diamond _{I} range\rightarrow \bigvee _{1\leq i \leq k} {pair}_{i} \wedge \bigwedge _{1\leq i< j\leq k} \neg ({pair}_{i}\wedge {pair}_{j}))\)
φ_{stripe}=□_{I} \((stripe \sqsubseteq \Box stripe)\wedge \Box _{I}\) \((\overline {stripe}\sqsubseteq \Box \overline {stripe}))\)
\(\varphi _{eq} = \Diamond _{I}(range \wedge \bigwedge _{a\in A}\) (left_{a}≡left_{b}))
where left_{a} and right_{a}(a∈A), left, right and stripe are spatial variables, for every pair (v_{i},w_{i})(1≤i≤k),pair_{i} are propositional variables. The variable range is required to ‘relativise’ temporal operators □_{I} and ♢_{I} in order to ensure that we can construct a model based on a finite flow of time.
ψ_{left} is a conjunction of (12)(18), for all i in 1≤i≤k, and for all j<l_{i},
The conjunct ψ_{right} is defined by replacing in ψ_{left} all occurrences of left with right, left _{a} with right _{a} (for a∈A) l_{i} with r_{i} and the sequence of \(left_{b_{j}^{i}}\)(for 1≤j≤l_{i}) with \(right_{c_{j}^{i}}\)(for 1≤j≤r_{i}). (Note that pair _{i} occurs in both ψ_{left} and ψ_{right}.)
Because of the difference between spatiotemporal logics and the corresponding model from [57], we change the Aleksandrov ttmodel \(\mathfrak {M} = ((\mathbb {N},<), \mathfrak {G},\mathfrak {V})\) with \(\mathfrak {G} = (V,R)\) to topometric temporal model \(\mathfrak {M} = (\mathfrak {T}, \mathfrak {L}, \mathfrak {V})\) with \(\mathfrak {L} = (M,\mathbb {I}_{d})\).
Since stripe holds in \(\mathfrak {M}\) at 0, we have, for every \(y\in M, \mathfrak {M}, (0,y)\models stripe\) iff \(\mathfrak {M}, (j,y)\models stripe\) for all j, 1≤j≤N. The transitive binary relation R_{s} on V defined by taking xR_{s}y will be modified to the distance relation d on the topometric space M defined by taking d(x,y). The condition will be changed to that if there is z∈M such that ∃ε,d(x,z)<ε and d(z,y)<ε, and (M,(0,x))⊧stripe holds iff (M,(0,z))⊮stripe. □
Theorem 5.2 is the immediate result of Lemma 5.3.
Case study
The example of the train collision avoidance system can be treated as a cyberphysical system. In cyber system, the electrical signals are discrete, and the system clock ticks discretely. However, in physical environment, the running of the train follows kinetic equation, which is continuously changing. The execution of the train collision avoidance system is characterized a sequence of trace, which describes the evolution of spatial region.
The region of movement authority is defined as a special rail sections from the position that an train is authorized to enter to the end of movement authority (EoA), which is the position that the train reaches the safety margin of the leading train. A train must send a movement authority request and receive a permission before the train can enter the next section. In case of emergency, the following train must brake without the permission. In Fig. 7, \(p_{ma\_req} \) and p_{brake} refer to the position that the following train sends a movement authority request and the braking position, respectively. The braking distance τ_{bd} occupies the section from the braking position p_{brake} to EoA. If the region of movement authority τ_{ma} is treated as the universal set, we can describe the spatial relation as
And \(\overline {\tau _{bd}}\) refers to the distance that the movement authority τ_{ma} minus the braking distance τ_{bd}, i.e., \(\mathfrak {V}(\overline {\tau _{bd}}) = \mathfrak {V}(\tau _{ma})  \mathfrak {V}(\tau _{bd})\), where \(\mathfrak {V}(\tau _{ma}), \mathfrak {V}(\overline {\tau _{bd}})\) and \(\mathfrak {V}(\tau _{bd})\) denote the value of the corresponding spatial terms \(\tau _{ma}, \overline {\tau _{bd}}\) and τ_{bd}.
Also, before the following train receives a movement authority, the position of the train is less than p_{brake}, i.e., the spatial entities of the train τ_{train} and the braking distance τ_{bd} doesn’t overlap. The spatial relation can be specified as
There are two kinds of braking modes: service braking (SB) and emergency braking (EB). Service braking refers to that a train decelerates until it stops in EoA. Emergency braking means the maximal acceleration a_{max} of a train to EoA if its velocity is greater than some critical value v_{c}. In order to ensure collision avoidance, after receiving movement authority, the following train decelerates by a given velocity v until it stops within the time t. Its formalization in STL is straightforward by formula 21:
where v, a and s is the velocity, acceleration and position of the following train.
Suppose the time from sending movement authority to stopping of the following train is t_{0}, and the braking time of the following train is t_{1}. In order to ensure collision avoidance, after sending movement authority at time 0, the train doesn’t overlap with the braking distance, until the train stops within t_{0}. The specification will be expressed as an STSL _{PC} formula
After sending the movement authority, the following train runs to the braking region. And after receiving movement authority, the train and the braking distance don’t overlap until it stops from t_{0} to t_{1}. Its formalization in an STSL _{OC} formula is straightforward the formula
Related work
In order to specify spatiotemporal properties of cyberphysical systems, the spatiotemporal logics, which enjoy discrete or dense time domain and Boolean or quantitative value, attract some researchers’ attention in Table 1. Generally, there are two kinds of logicbased approaches to specify spatiotemporal properties: the extension of temporal logic with spatial modalities and the combinations of spatial logic and temporal logic.
Some spatiotemporal logics are extensions of temporal logic with spatial modality. In [66], temporal modalities are extended with spatial directions to reason reaction diffusion systems. SSTL [60] is presented to combine the temporal modality until with two spatial modalities, so that one can express that something is true somewhere nearby and being surrounded by a region that satisfies a given spatiotemporal property. STREL [61] extends SSTL with spatial reachability, escape operators to describe the mobile and spatially distributed cyberphysical systems. Balbiani [67] explores the 2dimensional space in multiagent systems through extending dynamic logic with formulas representing the agents’ positions and programs moving from one position to another position. But these works face the problem of the expressiveness of discrete spatial representation, rather than more complex continuous space. Andreas [68] et al. present Shape Calculus based on Duration Calculus extended bounded polyhedron for the ndimensional space for the specification and verification of mobile realtime systems. Mardare [64] presents Dynamic Spatial Logic, \(\mathcal {L}_{DS}\), as an extension of HennessyMilner logic with parallel operator to distinguish processes. MLSL [63] is a twodimensional extension of spatial interval temporal logic, where one dimension is characterized by a continuous space to describe the position in each lane and the other denotes a discrete space to count the number of the lane.
The combinations of temporal logic and spatial logic inherit the expressiveness of the two kinds of logics. LTL and CTL imply discrete time in temporal part. Bennett et al. [65] construct a multidimensional modal logic named PSTL through the Cartesian product of the temporal logic PTL and the modal logic S4 _{u} to specify the discrete time and “general” topological space. Gabelaia et al. [22] present the principles for the requirements of a combined spatiotemporal form, and apply properly those principles and propose the combined spatiotemporal logic between PTL and some fragments of modal spatial logic S4 _{u}. They prove that the complexity of combination of PTL with S4 _{u} is PSPACEcomplete. Kremer and Mints [62] provide dynamic topological logic (DTL) as a combination of LTL and S4 _{u} to describe the dynamic changes of spatial objects over time. Shao et al. [69] also consider the combination of proposition temporal logic PTL and S4 _{u} and apply it to several classical properties of train control systems. Ciancia et.al [58] present STLCS through enhancing SLCS with temporal operators that features the CTL path quantifiers ∀ (for all paths) and ∃ (there exists a path). All these work are trying to answer how to specify spatiotemporal properties in discrete time, rather than dense time.
In order to express spatial changes in dense time, MTSL [31] is proposed to specify spatiotemporal properties of cyberphysical systems by integrating MTL with S4 _{u}. They follow the traditional S4 _{u} with truth value to specify spatial changes and extend the domain of time to realvalued interval within bounded time in the principle of PC and OC.
The above proposed language are employing classical model checking to verify the system from the specified properties. The approach achieves the model of a system to check whether the model satisfies the properties specified by the proposed language. However, we may need an approach to get the satisfaction degree, rather than satisfaction or violation.
STL expresses the changes of realvalued signals in dense time. The system can be verified at run time to monitor the satisfaction degree of the signals from an STL formula [11]. But it is not enough to specify spatial properties using an STL formula. Haghighi et al. [59] present SpaTeL as a combination of signal temporal logic (STL) and tree spatial superposition logic (TSSL) in networked systems. While, TSSL is a discrete structure to describe spatial static relations. To specify the spatial terms with changeable shape over dense time, we propose STSL _{PC} through integrating STL with S4 _{u}, to describe the spatiotemporal properties of cyberphysical systems with dense time and realvalued variables. STSL _{PC} interprets spatial subset relation and threshold predicate ad atomic proposition, and returns Boolean value to satisfaction or violation and the satisfaction degree of signals and spatial terms according to Boolean and quantitative interpretation. We extend STSL _{PC} to express the spatial evolution over dense time through extending the interpretation temporal globally, eventually and until bounded with interval as spatial operators. We assume that the systems satisfy finite variability so that the STSL _{PC} formulas are verified at runtime. The decidability and complexity of the two formalisms are analyzed, and the soundness and completeness of their axiomatization are proved.
Conclusion and future work
In this paper, we build STSL _{PC}, a spatiotemporal specification language by combining STL with spatial logic S4 _{u}, specifically containing dense time and topometric space. We provide the syntax and semantics of the proposed language, and guarantee the seamless integration of spatial logic with temporal aspect from the perspective of the changes of purely spatial proposition in STSL _{PC} and spatial objects in STSL _{OC} over time. A Hilbertstyle proof axiomatization system and the soundness and completeness results show that the completeness of STSL _{PC} and the incompleteness of STSL _{OC}. The decidability indicates the undecidable STSL _{OC} and the decidable STSL _{PC}. Further, we present that the complexity for the STSL _{PC} is EXPSPACEcomplete and a decision procedure is present for the decidable fragment. Currently, the proposed STSL has a powerful expressiveness.
However, the insufficiency of the paper exists in that there is no concrete monitoring technique, like other works [60, 70] have done. In order to verify a cyberphysical system, it is not the situation that we can achieve all the model of the system. However, monitoring provides an approach to verify a trace of the system to guarantee the reliability of the current execution. We have already presents the feasibility of the proposed language in the semantics. One can implement the language and present the monitoring algorithm to verify the spatiotemporal specification language. The reason why we are trying to monitor an STSL formula exists in the interpretation of STSL on spatiotemporal traces. This makes traditional model checking insufficient to verify the spatiotemporal properties specified with STSL. The spatiotemporal traces for monitoring an STSL specification can be automata, petri nets [2], process algebra, neural networks, differential equations et al. We are developing the monitoring tool. Firstly, we are developing the offline monitoring through sampling spatiotemporal traces for a simulated system against the spatiotemporal specification language. Secondly, the runtime verification like online monitoring will be interesting to achieve spatiotemporal traces from the executing systems to verify an STSL formula. Thirdly, the more applications, like mobile systems [71] and cloud servicebased systems, will be developed.
Appendix A: Proof of soundness of STSL _{PC}
The proof of Theorem 4.1:
S0 and S1 can be immediately got from the definition.
S2: \(\tau _{1}\sqsubseteq \tau _{2}\leftrightarrow \overline {\tau _{2}}\sqsubseteq \overline {\tau _{1}}\)
Proof
\(\tau _{1}\sqsubseteq \tau _{2}\), ⇔ If x∈τ_{1}, then x∈τ_{2}, ⇔ If x∉τ_{2}, then x∉τ_{1},\(\Leftrightarrow \overline {\tau _{2}}\sqsubseteq \overline {\tau _{1}}\). □
S3 \(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \mathbb {C}\tau _{1}\sqsubseteq \tau _{2}\)
Proof
\(\tau _{1}\sqsubseteq \tau _{2}\),\(\mathfrak {V}(\tau) = \mathfrak {V}(\mathbb {C}\tau)\),\(\Leftrightarrow \mathbb {C}\tau _{1}\sqsubseteq \tau _{2}\). □
S4 \(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\)
Proof
\(\tau _{1}\sqsubseteq \tau _{2}\), ⇔, if x∈τ_{1}, then x∈τ_{2},and \(\mathfrak {V}(\tau _{1})\neq \mathfrak {V}(\tau _{2})\),\(\Leftrightarrow, \tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\), □
S5 \(\tau _{1}\sqsubseteq \tau _{2} \rightarrow \mathbb {I}\tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\)
Proof
\(\tau _{1}\sqsubseteq \tau _{2}\),\(\Leftrightarrow, \tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\),\(\mathbb {I}\tau _{1}\sqsubseteq \tau _{1}\),\(\Rightarrow, \mathbb {I}\tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\). □
S6 \(\tau _{1}\sqsubseteq \tau _{2} \leftrightarrow \mathbb {C}\tau _{1}\sqsubseteq \mathbb {C}\tau _{2}\)
Proof
\(\tau _{1}\sqsubseteq \tau _{2}\),\(\Leftrightarrow,\mathbb {C}\tau _{1}\sqsubseteq \tau _{2}\),\(\tau _{2}\sqsubseteq \mathbb {C}\tau _{2}\),\(\Rightarrow, \mathbb {C}\tau _{1}\sqsubseteq \mathbb {C}\tau _{2}\). □
S7 \(\tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\rightarrow \tau _{1}\sqsubseteq \mathbb {C}\tau _{2}\)
Proof
\(\tau _{1}\sqsubseteq \mathbb {I}\tau _{2}\),\(\Leftrightarrow, \mathbb {I}\tau _{2}\sqsubseteq \tau _{2}\),\(\tau _{2}\sqsubseteq \mathbb {C}\tau _{2}\),\(\Rightarrow, \tau _{1}\sqsubseteq \mathbb {C}\tau _{2}\). □
S8 \((\tau _{1}\sqcup \tau _{2})\sqsubseteq \tau _{3}\rightarrow (\tau _{1}\sqsubseteq \tau _{3})\vee (\tau _{2}\sqsubseteq \tau _{3})\)
Proof
\((\tau _{1}\sqcup \tau _{2})\sqsubseteq \tau _{3}\), ⇔, if x∈τ_{1} or x∈τ_{2}, then x∈τ_{3}, ⇒, if x∈τ_{1}, then x∈τ_{3} or x∈τ_{2}, then x∈τ_{3},\(\Leftrightarrow, \tau _{1}\sqsubseteq \tau _{3}\), or \(\tau _{2}\sqsubseteq \tau _{3}\),\(\Leftrightarrow, (\tau _{1}\sqsubseteq \tau _{3})\vee (\tau _{2}\sqsubseteq \tau _{3})\) □
S9 \(\tau _{1}\sqsubseteq (\tau _{2}\sqcap \tau _{3})\rightarrow (\tau _{1}\sqsubseteq \tau _{2})\wedge (\tau _{1}\sqsubseteq \tau _{3})\)
Proof
\(\tau _{1}\sqsubseteq (\tau _{2}\sqcap \tau _{3})\), ⇔, if x∈τ_{1}, then x∈τ_{2} and x∈τ_{3}, ⇒, if x∈τ_{1}, then x∈τ_{2} and x∈τ_{1}, then x∈τ_{3},\(\Leftrightarrow, \tau _{1}\sqsubseteq \tau _{2}\), and \(\tau _{1}\sqsubseteq \tau _{3}\),\(\Leftrightarrow, (\tau _{1}\sqsubseteq \tau _{2})\wedge (\tau _{1}\sqsubseteq \tau _{3})\). □
S10:
Proof
,\(\Leftrightarrow, \exists x\in \mathcal {L}, x\in \tau \),\(\Leftrightarrow, \forall y\notin \mathcal {L}, y\notin \tau \),
□
S11:
Proof
,\(\Leftrightarrow, \forall x\in \mathcal {L}, x\in \tau _{1}\) and x∈τ_{2},\(\Rightarrow \forall x\in \mathcal {L}, x\in \tau _{1}\) and \(\forall x\in \mathcal {L}, x\in \tau _{2}\),
□
S12:
Proof
,\(\Leftrightarrow \exists x\in \mathcal {L}, x\in \tau _{1}\) or x∈τ_{2},\(\Rightarrow \exists x\in \mathcal {L}, x\in \tau _{1}\) or \(\exists x\in \mathcal {L}, x\in \tau _{2}\),
. □
T1: ♢_{I}ϕ⇔¬□_{I}¬ϕ
Proof
♢_{I}ϕ, ⇔∃t∈I,ϕ holds, ⇔ The proposition ∀t∈I,ϕ doesn’t hold is false, ⇔¬□_{I}¬ϕ. □
T2: □_{I}(ϕ∧φ)→(□_{I}ϕ∧□_{I}φ)
Proof
□_{I}(ϕ∧φ), ⇔,∀t∈I,ϕ and φ hold, ⇒,∀t∈I,ϕ holds and ∀t∈I,φ holds, ⇔,□_{I}ϕ∧□_{I}φ □
T3 ♢_{I}(ϕ∨φ)→(♢_{I}ϕ∨♢_{I}φ)
Proof
♢_{I}(ϕ∨φ), ⇔,∃t∈I,ϕ or φ hold, ⇒,∃t∈I,ϕ holds or ∃t∈I,φ holds, ⇔,♢_{I}ϕ∨♢_{I}φ □
T4: ϕU_{I}φ→♢_{I}φ
Proof
ϕU_{I}φ, ⇔∃t^{′}∈[t+I],φ holds, and ∀t^{″}∈[t,t^{′}],ϕ holds, ⇒∃t^{′}∈[t+I],φ holds, ⇔♢_{I}φ □
Q0: ♢_{I∪J}φ⇔♢_{I}φ∨♢_{J}φ
Proof
♢_{I∪J}φ, ⇔∃t∈I or J,φ holds, ⇔∃t∈I,φ holds, or ∃t∈J,φ holds, ⇔♢_{I}φ∨♢_{J}φ. □
Q1: □_{I∩J}φ⇔□_{I}φ∧□_{J}φ
Proof
□_{I∩J}φ, ⇔∀t∈I and J,φ holds, ⇔∀t∈I,φ holds, and ∀t∈J,φ holds, ⇔Box_{I}φ∧□_{J}φ. □
Q2: ϕU_{I∪J}φ→♢_{I∪J}φ
Proof
ϕU_{I∪J}φ, ⇔∃t^{′}∈[t+I∪J],φ holds, and ∀t^{″}∈[t,t^{′}],ϕ holds, ⇒∃t^{′}∈[t+I∪J],φ holds, ⇔♢_{I∪J}φ □
Availability of data and materials
Not applicable.
Abbreviations
 STL:

Signal temporal logic
 STSL:

Spatiotemporal specification language
 LTL:

Linear temporal logic
 MITL:

Metric interval temporal logic
 CP:

Classical propositional logic
 BNF:

BackusNaur form
 PST:

Propositional spatiotemporal logic
 SMT:

Satisfiability modulo theories
References
Lee EA, Seshia SA (2016) Introduction to Embedded Systems: A Cyberphysical Systems Approach. MIT Press, California.
Liu G, Jiang C, Zhou M (2018) Timesoundness of time Petri nets modelling timecritical systems. ACM Trans Cyber Phys Syst 2(2):1–27.
Fan C, Qi B, Mitra S, Viswanathan M, Duggirala PS (2016) Automatic reachability analysis for nonlinear hybrid models with C2E2 In: International Conference on Computer Aided Verification, 531–538, Springer.
Gao H, Liu C, Li Y, Yang X (2020) V2VR: reliable hybridnetworkoriented V2V data transmission and routing considering RSUs and connectivity probability. IEEE Trans Intell Transp Syst 21:1–14. https://doi.org/10.1109/TITS.2020.2983835.
Liu J, Li T, Ding Z, Qian Y, Sun H, He J (2019) AADL+: a simulationbased methodology for cyberphysical systems. Front Comput Sci 13(3):1–23.
Gao H, Chu D, Duan Y, Yin Y (2017) Probabilistic model checkingbased service selection method for business process modeling. Int J Softw Eng Knowl Eng 27(6):897–923.
An D, Liu J, Chen X, Li T, Yin L (2019) A Modeling Framework of CyberPhysicalSocial Systems with Human Behavior Classification Based on Machine Learning In: 21st International Conference on Formal Engineering Methods, 522–525, Springer.
Gao H, Kuang L, Yin Y, Guo B, Dou K (2020) ’Mining consuming behaviors with temporal evolution for personalized recommendation in mobile marketing Apps. Mob Netw Appl (MONET) 25(4):1233–1248.
Wolter F, Zakharyaschev M (2005) A logic for metric and topology. J Symb Log 70(3):795–828.
Raman V, Donzé A, Sadigh D, Murray RM, Seshia SA (2015) Reactive synthesis from signal temporal logic specifications In: Proceedings of the 18th International Conference on Hybrid Systems: Computation and Control, 239–248, ACM.
Donzé A, Ferrere T, Maler O (2013) Efficient robust monitoring for STL In: International Conference on Computer Aided Verification, 264–279, Springer.
Blackburn P, De Rijke M, Venema Y (2002) Modal Logic: Graph. Darst, Vol. 53. Cambridge University Press, Dallas, America.
Davoren JM (2007) Topological semantics and bisimulations for intuitionistic modal logics and their classical companion logics In: International Symposium on Logical Foundations of Computer Science, 162–179, Springer.
FernándezDuque D (2010) Absolute completeness of S4u for its measuretheoretic semantics. Adv Modal Log 8:100–119.
Li T, Jing L, An D, Sun H (2019) A Sound and Complete Axiomatisation for SpatioTemporal Specification Language In: The 31st International Conference on Software Engineering & Knowledge Engineering, 153–204, KSI.
Demri S, Goranko V, Lange M (2016) Temporal Logics in Computer Science: Finitestate Systems, Vol. 58. Cambridge University Press, Cambridge, United Kingdom.
Zhang Y, Li K (2015) Decidability of logics based on an indeterministic metric tense logic. Stud Logica 103(6):1123–1162.
Platzer A (2018) Logical foundations of cyberphysical systems. Springer, Gewerbestrasse, Switzerland.
Bohner M, Peterson A (2012) Dynamic equations on time scales: An introduction with applications. Birkhäuser Boston, Washington D.C., USA.
Ladner RE (1977) The computational complexity of provability in systems of modal propositional logic. SIAM J Comput 6(3):467–480.
McKinsey JCC (1941) A solution of the decision problem for the Lewis systems S2 and S4, with an application to topology. J Symb Log 6(4):117–124.
Gabelaia D, Kontchakov R, Kurucz A, Wolter F, Zakharyaschev M (2005) Combining spatial and temporal logics: expressiveness vs. complexity. J Artif Intell Res 23:167–243.
Randell DA, Cui Z, Cohn AG (1992) A spatial logic based on regions and connection In: Proceedings of the 3rd International Conference on Principles of Knowledge Representation and Reasoning, 165–176, Morgan.
Liu W, Li S, Renz J (2009) Combining RCC8 with Qualitative Direction Calculi: Algorithms and Complexity In: Proceedings of the 21st International Joint Conference on Artificial Intelligence, 854–859, Morgan Kaufmann.
Kontchakov R, Kurucz A, Wolter F, Zakharyaschev M (2007) Spatial logic+ temporal logic=? In: Handbook of Spatial Logics, 497–564, Springer.
Shehtman V (1999) Everywhere and here. J Appl NonClass Log 9(23):369–379.
Pnueli A (1977) The temporal logic of programs In: 18th Annual Symposium on Foundations of Computer Science, 46–57, IEEE.
Pradella M, Morzenti A, Pietro PS (2013) Bounded satisfiability checking of metric temporal logic specifications. ACM Trans Softw Eng Methodol (TOSEM) 22(3):1–54.
Maler O, Nickovic D (2004) Monitoring temporal properties of continuous signals In: Formal Techniques, Modeling and Analysis of Timed and FaultTolerant Systems, 152–166, Springer.
Donzé A, Maler O (2010) Robust satisfaction of temporal logic over realvalued signals In: International Conference on Formal Modeling and Analysis of Timed Systems, 92–106, Springer.
Sun H, Liu J, Chen X, Du D (2015) Specifying cyber physical system safety properties with metric temporal spatial logic In: 2015 AsiaPacific Software Engineering Conference (APSEC), 254–260, IEEE.
Gabbay D, Pnueli A, Shelah S, Stavi J (1980) On the temporal analysis of fairness In: Proceedings of the 7th ACM SIGPLANSIGACT Symposium on Principles of Programming Languages, 163–173, ACM.
Lichtenstein O, Pnueli A (1985) Checking that finite state concurrent programs satisfy their linear specification In: Proceedings of the 12th ACM SIGACTSIGPLAN Symposium on Principles of Programming Languages, 97–107, ACM.
Nenzi L, Bortolussi L, Ciancia V, Loreti M, Massink M (2015) Qualitative and quantitative monitoring of spatiotemporal properties In: Runtime Verification, 21–37, Springer.
Kuratowski K (2014) Topology, Vol. 1. Elsevier Science, London, England.
Milner R (2001) Bigraphical reactive systems In: International Conference on Concurrency Theory, 16–35.
Sevegnani M, Calder M (2015) Bigraphs with sharing. Theor Comput Sci 577:43–73.
Lemire D (2007) Streaming maximumminimum filter using no more than three comparisons per element. Nordic J Comput 13(4):328–339.
Pnueli A (1981) The temporal semantics of concurrent programs. Theor Comput Sci 13(1):45–60.
Kesten Y, Pnueli A (2002) Complete proof system for QPTL. J Log Comput 12(5):701–745.
Schobbens PY, Raskin JF, Henzinger TA (2002) Axioms for realtime logics. Theor Comput Sci 274(12):151–182.
Jacquette D (2002) A Companion to Philosophical Logic. Wiley Online Library, Viotoria, Australia.
Balbiani P, FernándezDuque D (2016) Axiomatizing the lexicographic products of modal logics with linear temporal logic In: International Conference on Advances in Modal Logic, 78–96.
Montanaria A, de Rijkeb M (1997) Twosorted metric temporal logics. Theor Comput Sci 183(2):187–214.
Gabbay DM, Hodkinson IM (1990) An axiomatization of the temporal logic with until and since over the real numbers. J Log Comput 1(2):229–259.
Kojima K, Igarashi A (2011) Constructive lineartime temporal logic: Proof systems and Kripke semantics. Inf Comput 209(12):1491–1503.
Chellas BF (1980) Modal Logic: An Introduction. Cambridge university press, New York, USA.
Alur R, Feder T, Henzinger TA (1996) The benefits of relaxing punctuality. J ACM 43(1):116–146.
Barrett C, Tinelli C (2018) Satisfiability modulo theories In: Handbook of Model Checking, 305–343.. Springer, Cham, Switzerland.
Bersani MM, Rossi M, San Pietro P (2015) An SMTbased approach to satisfiability checking of MITL. Inf Comput 245:72–97.
Bersani MM, Rossi M, San Pietro P (2013) Deciding continuoustime metric temporal logic with counting modalities In: International Workshop on Reachability Problems, 70–82, Springer.
Demri S, D’Souza D (2007) An automatatheoretic approach to constraint LTL. Inf Comput 205(3):380–415.
Bersani MM, Rossi M, Pietro PS (2013) Deciding the satisfiability of MITL specifications In: 4th International Symposium on Games, Automata, Logics and Formal Verification, 64–78.
Bersani MM, Rossi MG, San Pietro P (2014) On the satisfiability of metric temporal logics over the reals. Electron Commun EASST 66:1–15.
Hopcroft JE, Motwani R, Ullman JD (2001) Introduction to automata theory, languages, and computation. Acm Sigact News 32(1):60–65.
Gabbay DM, Kurucz A, Wolter F, Zakharyaschev M (2003) Manydimensional modal logics: theory and applications. Elsevier North Holland, London, United Kingdom.
Gabelaia D, Kontchakov R, Kurucz A, Wolter F, Zakharyaschev M (2003) On the Computational Complexity of SpatioTemporal Logics In: FLAIRS Conference, 460–464.
Ciancia V, Gilmore S, Grilletti G, Latella D, Loreti M, Massink M (2018) Spatiotemporal model checking of vehicular movement in public transport systems. Int J Softw Tools Technol Transfer 20(3):289–311.
Haghighi I, Jones A, Kong Z, Bartocci E, Gros R, Belta C (2015) SpaTeL: a novel spatialtemporal logic and its applications to networked systems In: Proceedings of the 18th International Conference on Hybrid Systems: Computation and Control, 189–198, ACM.
Nenzi L, Bortolussi L, Ciancia V, Loreti M, Massink M (2017) Qualitative and quantitative monitoring of spatiotemporal properties with SSTL. Log Methods Comput Sci 14(4):1–38.
Bartocci E, Bortolussi L, Loreti M, Nenzi L (2017) Monitoring mobile and spatially distributed cyberphysical systems In: Proceedings of the 15th ACMIEEE International Conference on Formal Methods and Models for System Design, 146–155, ACM.
Kremer P, Mints G (2005) Dynamic topological logic. Ann Pure Appl Log 131(1):133–158.
Xu B, Li Q (2016) A spatial logic for modeling and verification of collisionfree control of vehicles In: 2016 21st International Conference on Engineering of Complex Computer Systems (ICECCS), 33–42, IEEE.
Mardare R (2006) Logical analysis of complex systems: Dynamic epistemic spatial logics. PhD thesis, University of Trento.
Bennett B, Cohn AG, Wolter F, Zakharyaschev M (2002) Multidimensional modal logic as a framework for spatiotemporal reasoning. Appl Intell 17(3):239–251.
Bartocci E, Gol EA, Haghighi I, Belta C (2018) A formal methods approach to pattern recognition and synthesis in reaction diffusion networks. IEEE Trans Control Netw Syst 5(1):308–320.
Balbiani P, FernándezDuque D, Lorini E (2017) Exploring the bidimensional space: a dynamic logic point of view In: The 16th Conference on Autonomous Agents and MultiAgent Systems, 132–140, Springer.
Schäfer A (2004) A calculus for shapes in time and space In: International Colloquium on Theoretical Aspects of Computing, 463–477, Springer.
Shao Z, Liu J, Ding Z, Chen M, Jiang N (2013) Spatiotemporal properties analysis for cyberphysical systems In: 2013 18th International Conference on Engineering of Complex Computer Systems, 101–110, IEEE.
Gao H, Xu Y, Yin Y, Zhang W, Li R, Wang X (2020) Contextaware QoS prediction with neural collaborative filtering for InternetofThings services. IEEE Internet Things J 7(5):4532–4542.
Gao H, Huang W, Duan Y (2020) The cloudedge based dynamic reconfiguration to service workflow for mobile ecommerce environments: A QoS prediction perspective. ACM Trans Internet Technol. https://doi.org/10.1145/3391198.
Acknowledgments
We would like to thank the anonymous reviewers for their valuable comments and suggestions, which helped improve the quality of this paper. Also, we want to express our heartfelt gratitude to the authors of the literature cited in this paper for contributing useful ideas to this study.
Funding
This document is the results of the research project funded by the National Key Research and Development Project 2017YFB1001800, NSFC 61972150 and Shanghai Knowledge Service Platform Project ZF1213.
Author information
Authors and Affiliations
Contributions
Authors’ contributions
Tengfei Li, Jing Liu and Haiying Sun have written this paper and have done the research which supports it. Tengfei Li writes the paper and is responsible to prove the completeness and decidability of the proposed language. Jing Liu is the lead of the laboratory. She is responsible for obtaining the funding of the work. Also, she contributes to the fundamental idea of the work. Haiying Sun is responsible for maneging the schedule and provides feasible plan. She provides the initial idea of the proposed language and takes participate in the proof work. Further, she proofreads the paper to improve the English writing and has a great influence on the revision process. So it is bettwe for her to be the corresponding author. Both Xiang Chen and Lipeng Zhang contribute to the applications and case study of our proposed spatiotemporal specification language. Xiang Chen mainly considers the train collision avoidance system and provides the spatiotemporal properties according to the specification. Lipeng Zhang contributes to several related work and discusses the difference from other applications. Also, Lipeng provides a perspective to interpret the concurrency for the topometric model. Junfeng Sun has collaborated in the examples of this paper as an assistant president of CASCO Signal Ltd. The authors read and approved the final manuscript.
Authors’ information
Tengfei Li received the B.S. degree in mathematics from the School of Mathematics and Statistics in 2014. He is currently pursuing the Ph.D. degree in Software Engineering at School of Software Engineering, East China Normal University, Shanghai, China. From April to October 2018, he visited INRIA Sophia Antipolis, Nice, France. His research interests are in the area of spatiotemporal logics, safetycritical cyber physical systems, formal verification and hybrid systems. Jing Liu is currently a professor of computer science with East China Normal University, China. In recent years, she has been involved in the area of modeldriven architecture. She currently focuses on the design of realtime embedded systems and cyberphysical systems. Haiying Sun received a Ph.D. degree in East China Normal University, China, her research interests include formal method, system simulation and modeldriven engineering. Xiang Chen is engineer of Urban Train Control System Development Department in CASCO, he has rich experience in safetyrelated software development, and leads a team to develop automatic train control(ATC) system for urban rail transit., he has done a lot of research on safety software technology, and introduces formalmethod/semimethod and Model Based Software Engineering(MBSE) for ATC software. Lipeng Zhang is currently a Manager of the Safetd platform department, CASCO Signal Ltd. In recent years, he has been involved in the area of safety critical software verification and validation. Currently, he focuses on the design of CBTC(Communication Based Train Control) system based on vehiclevehicle communication. JunFeng Sun is currently an assistant president of CASCO Signal Ltd and a vice director of R&D Institute of CASCO Signal Ltd. In recent years, he is working on the area of train control system design. Now his work focuses on the design of railway signal security computer platform.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests regarding the publication of this manuscript.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visithttp://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Li, T., Liu, J., Sun, H. et al. A spatiotemporal specification language and its completeness & decidability. J Cloud Comp 9, 65 (2020). https://doi.org/10.1186/s13677020002093
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13677020002093