Self play with parameter sharing in n-player mixed competitive-cooperative games
Georgios Marios Skaltsis, Hyo‐Sang Shin, Antonios Tsourdos · AIAA SCITECH 2022 Forum · 2022
View Video Presentation: https://doi.org/10.2514/6.2022-2498.vid We introduce some parameter sharing multi-agent reinforcement learning schemes, combined with self-play for n-players mixed competitive-cooperative games. Except for the pure self-play scheme, an another one using the best policy, outperform the pure parameter sharing baseline, leading to better exploration of the state space and protecting from the performance deterioration observed with the baseline.