CoffeeddCat/Multiagent_chainMDP

A simple&naive platform for multiagent chainMDP problem.

Multiagent_chainMDP

• It's a simple&naive platform for multiagent chainMDP problem.

• The algorithm now used: DDQN+reward updated.

• Environment needed:

  python3.5, Tensorflow1.4.0, Numpy1.13.3, matplotlib2.1.0, Tensorboard0.4.0rc(unnecessary)
  

• Both GPU or CPU for tensorflow are surpported.

• References:

  • https://arxiv.org/abs/1507.00814
  • http://papers.nips.cc/paper/6500-deep-exploration-via-bootstrapped-dqn
  • ...

• Need to be done:

  • Multi thread to speed up
  • More algorithm should be applied
  • A more efficient exploration method to solve it.