CS-DLMA is a carrier sense multiple access (CSMA) protocol for heterogeneous wirless networking. The underpinning technique of CS-DLMA is deep Q-network (DQN). However, conventional DQN algorithms cannot fit in with the specicial issues in CSMA protocols design. We put forth a new DQN algorithm to address these issuses.

There are three main files in each folder, simulating the interactions of CS-DLMA node with other nodes.

• run_DQN.py is the main framework, and you can run this file to start the simulation.
• environment.py simulates the interaction between the CS-DLMA node and the WiFi node. It is similar as OpenAI gym environment.
• DQN_brain.py builds the CS-DLMA node/agent. The new DQN algorithm is here.