EURASIP Journal on Wireless Communications and Networking

Table 3 Comparison of testing results with w_m=0.8 for all mobile users

From: Decentralized computation offloading for multi-user mobile edge computing: a deep reinforcement learning approach

	Average reward			Average power (watt)			Average delay (ms)
	*User 1*	*User 2*	*User 3*	*User 1*	*User 2*	*User 3*	*User 1*	*User 2*	*User 3*
DDPG	− 1.919	− 6.366	− 16.164	0.162	0.602	1.674	3.114	7.752	13.861
DQN	− 2.780	− 8.915	− 18.675	0.284	0.915	1.954	2.539	7.973	15.216
GD-Offload	− 3.417	− 10.893	− 26.334	0.402	1.309	3.143	1.007	2.103	5.951
GD-Local	− 1.949	− 13.870	− 28.470	0.216	1.678	3.407	1.106	2.228	6.072

Back to article page