Multi-Agent Reinforcement Learning on Flatland 3 challenge

Through this project, we aim to test the applicability of Multi-Agent methods to a setting similar to a real-world scenario, in particular to the efficient automated scheduling of trains. The Flatland 3 challenge provides a two-dimensional grid environment perfect for enabling agent learning and testing the performance of proposed Multi-Agent solutions. We will tackle the Flatland 3 challenge by implementing different Multi- Agent Reinforcement Learning algorithms and comparing their perfor- mance on the test benchmark.