MIT Researchers Develop System To Reduce Robot Congestion

Researchers from MIT and Symbotic developed a hybrid deep reinforcement learning and planning system that prioritizes and reroutes hundreds of warehouse robots to avoid congestion, reported in the Journal of Artificial Intelligence Research. In simulations based on real e-commerce layouts, the approach increased throughput by about 25 percent compared with traditional algorithms and adapted quickly to different layouts and robot densities. The team plans to extend the method to task assignment and larger warehouses.