3 ms·
MARO is a new multi-agent reinforcement learning playground in the resource optimization domain.
by arthursjiang 6y ago
MARO is a new multi-agent reinforcement learning playground in the resource optimization domain.
- MasterScrat 6y agoThis looks great! happy to see more activity in the MARL space for real-world problems. We at AIcrowd are working on a similar project: Flatland (http://flatland.aicrowd.com/ http://flatland.aicrowd.com/), a procedurally-generated MARL environment focused on managing traffic on complex railway networks. It's a collaboration with SBB (the Swiss national railway company) and Deutsche Bahn (the German one). We're currently running a NeurIPS challenge pitting OR solutions against RL ones: https://www.aicrowd.com/challenges/neurips-2020-flatland-challenge https://www.aicrowd.com/challenges/neurips-2020-flatland-cha... We will look into MARO. We should talk at some point!