PyTorch implementation
Originally taken with first commit from NIPS 2020 sumbision of paper "Bayesian Multi-type Mean Field Multi-agent Imitation Learning"
buildILBerlinGAIL: the codes for the transportation environment (Berlin) buildILGymEnvsRT: the codes for the Rover Tower environment
collect_expert: the code to collect expert samples BM3IL: Bayesian multi-type mean field multi-agent imitation learning MA-DAAC: multi-agent discriminator-attention-actor-criti MA-GAIL: multi-agent generative adversarial imitation learning MTMFIL: the MA-GAIL plus the existing multi-type mean field approximation Bayesian-2-MA-DAAC: Bayesian approach for MA-DAAC with 2 samples of the discriminator parameters per agent