/Cooperative-Search

Multi-agent RL algorithms are applied to large-scale cooperative target searching mission

Primary LanguagePython

Cooperative-Search

Multi-agent RL algorithms are applied to large-scale cooperative target searching mission

Env:
  1. flight_easy: Continuous search environment

image image

  1. flight: Add probability graph to flight_easy
  2. search_env: Discrete search environment
  3. simple_spread: A test env
Algorithm:
  1. QMIX
  2. DOP
  3. VDN
  4. Reinforce
Result:

env = flight_easy target_num = 15

  1. agent_num = 1, agent_mode = 0:

image image

  1. agent_num = 3, agent_mode = 0:

image image

  1. agent_num = 5, agent_mode = 0:

image image

...

Partial code Reference:
  1. https://github.com/starry-sky6688/StarCraft
  2. https://github.com/TonghanWang/DOP
  3. https://github.com/openai/multiagent-particle-envs