An Exoatmospheric Homing Guidance Law Based on Deep Q Network

Table 3

DQN algorithm hyperparameter.

Hyperparameter	Parameter value
Maximum iterations	3000
Discount factor	0.996
Q network learning rate	0.001
Capacity of experience replay memory	100000
Minibatch size	64
Target network update rate	10
Initial exploration	0.8
Final exploration	0.01
Reward coefficient	0.05