Two-Loop Acceleration Autopilot Design and Analysis Based on TD3 Strategy

Table 2

TD3 algorithm hyperparameters.

Parameter	Value
Maximum permissible episodes	1600
Maximum permissible steps of each episode	572
Actor-network learning rate	10^-4
Critic-network learning rate	10^-4
Regularization constant
Discounting factor	0.99
Sampling size	256
Variance for the initial exploration noise	0.25
Variance fading factor	0.005
Variance for random noise	0.2
Upper limit of the random noise	0.25