A reinforcement learing environment for robotic mobile fulfilment system (RMFS)
We provide a reinforcement learning environment for RMFS and provide an improved DQN algorithm to realize the control of AGV.
For more information, you can search my paper: Astar guiding DQN algorithm for AGV pathfinding problem of robotic mobile fulfillment systems
1.Run the program by running the run.py
2.Choose how to control the AGV by setting the parameter control_mode
3.Various system layouts can be generated by setting the parameters shown in the figure. It should be noted that each storage station (red square) must be connected to a road.
4.EXAMPLE: The following set of animations shows the different learning stages of the neural network.
The first picture is the early stage of training, the AGV under the control of the neural network does not know how to act at all.
After training for a period of time, the AGV was able to complete a small number of tasks, but the neural network still made a lot of wrong decisions.
After a period of training, AGV can complete a large number of tasks, but the neural network still occasionally makes wrong decisions.
After the neural network is trained, it can control the AGV to complete all tasks.