Reducing Entropy Overestimation in Soft Actor Critic Using Dual Policy Network

Table 2

Max evaluation reward achieved during learning.

Environment	Algorithm	Max value	S.D
Ant-v2	SAC	4536.18	1425.31
	TD3	4360.79	2081.05
	Ours	5151.27	1600.29
HalfCheetah-v2	SAC	8308.59	4298.95
	TD3	-1.65	0
	Ours	9158.21	4414.13
Hopper-v2	SAC	2905.15	1388.34
	TD3	2622.66	1245.89
	Ours	2812.76	1311.20
Walker2d-v2	SAC	3611.99	1670.66
	TD3	3513.84	1635.82
	Ours	4357.33	2090.74