Dependent Task-Offloading Strategy Based on Deep Reinforcement Learning in Mobile Edge Computing

Table 1

Experimental parameters.

Parameter	Value
Computing resources of the device	[0.1, 0.5] G cycles/s
Total edge server computing resources	{18, 15, and 12} G cycles/s
Power at the device idle	[0.004, 0.04] W
The data size of the task i	[30, 50] MB
The computational complexity of tasks	600cycles/Kb
The device wireless data transmission rate	[0.1, 10] MB/s
Data sending power of the device	0.1 W
The local calculated power of the device	10 W
Delay threshold	200
Energy consumption threshold	200
Weighting factor	0.5
Number of DNNs	3
The memory size of experience replay	1024
Batch training size	128
Learning rate	0.01
Number of training rounds	4000
Threshold value	512
Number of IoT devices	10
Number of neurons in the first hidden layer	120
Number of neurons in the second hidden layer	80