Deep Multi-agent Reinforcement Learning for Highway On-Ramp Me | Data Science Digest

Deep Multi-agent Reinforcement Learning for Highway On-Ramp Merging in Mixed Traffic

In this paper, Dong Chen et al. formulate the mixed-traffic highway on-ramp merging problem as a multi-agent reinforcement learning (MARL) problem. They propose an efficient MARL framework that can be used in dynamic traffic. A novel safety supervisor is developed to significantly reduce collision rate and greatly expedite the training process.

Paper — https://bit.ly/2QSyPvE
Code — https://bit.ly/34lLJoN
Code — https://bit.ly/2SnRiRd