MIT AI System Optimizes Warehouse Robot Traffic for Smoother Operations

Researchers at MIT and the technology firm Symbotic have developed a hybrid system that coordinates traffic among hundreds of robots inside large autonomous warehouses. The method uses deep reinforcement learning to determine which robots should receive priority when congestion forms and then applies a planning algorithm to reroute them before bottlenecks occur. The approach enables robots to respond quickly as new tasks arrive and conditions change on the warehouse floor. In simulations based on actual e-commerce warehouse layouts, the system achieved about 25 percent greater throughput than traditional algorithms or a random search method, measured by packages delivered per robot. The trained model adapts to warehouses with different numbers of robots or varied layouts without additional retraining. Researchers designed their own simulation environments because existing industrial tools proved too slow for the problem. Han Zheng, a graduate student in MIT's Laboratory for Information and Decision Systems and lead author of the paper, said the method can achieve super-human performance on complex decision-making tasks in manufacturing and logistics. The research appears in the Journal of Artificial Intelligence Research. It was funded by Symbotic.