Skip to main content

Advances, Systems and Applications

Table 4 Simulation parameters

From: A deep reinforcement learning assisted task offloading and resource allocation approach towards self-driving object detection

Parameter

Meaning

Value

B

The bandwidth

11 MB/s

\(\tau _H\)

The execution delay tolerance of high-priority tasks

0.01 s

\(\tau _L\)

The execution delay tolerance of low-priority tasks

0.02 s

\(\Upsilon _H\)

The utility constant for high-priority tasks

0.2

\(\Upsilon _L\)

The utility constant for low-priority tasks

0.1

\(N_0\)

The noise power

\(10^{-8}\)

P

The data transmission power

6 W

\(\Upsilon\)

The computation time of the object detection task on the edge server

0.001s

\(F_L\)

The computing power of the vehicle terminal

\(1.08\times 10^6\) bytes/s

\(C_{i,v}^j\)

The local computation size for task \(S_i^j\) at partition point v

\(\left[ 0,2000\right]\) bytes

\(M_{i,v}^j\)

The data size of task \(S_i^j\) for offloading at partition point v

[0,5] MB

E

The potential for the object detection algorithm to suffer from detection error

0.02

\(P_{i,x,y}^j\)

The probability that the object x belongs to category y when the object detection algorithm achieves correct detection

[0.85,1]

\(\overline{P}_{i,x,y}^j\)

The probability that the object x belongs to category y when the object detection algorithm suffers from detection error

[0,0.85]