Multirobot Coverage Path Planning Based on Deep Q-Network in Unknown Environment

Table 2

The list of parameters of training.

Parameters	Values
Replay memory size	100000
Discount factor	0.99
The initial value of the greedy exploration	1
The final value of the greedy exploration	0.25
Robot's maximum number of steps in each episode	180
Learning rate	0.0005
Target network update frequency	50 (episodes)