Research Article

A Deep -Network-Based Collaborative Control Research for Smart Ammunition Formation

Table 2

Parameter set for the DQN.

ParameterValueParameterValue

Inner radius (m)600Total episodes
Outer radius (m)1000Each episode’s time (s)120
Adjust factor 0.05Time step (s)1.0
Return discount factor 0.95Episode number to calculate the average total reward100
Update period of the target network 1000’s mean value and variance (, 0.8)
Exploration probability 0.1’s mean value and variance (0.0, 1.0)
Number of followers2’s mean value and variance (0.0, 1.0)
Capacity of experience replay pool 105’s mean value and variance (0.0, 1.0)
Mini-batch size 32’s mean value and variance (0.0, 1.0)
Learning rate 0.01’s mean value and variance (0.0, 1.0)