AI system learns to keep warehouse robot traffic running smoothly

Researchers from MIT and the tech firm Symbotic have developed a method that uses artificial intelligence to keep a fleet of robots moving smoothly in large autonomous warehouses. The system learns which robots should go first at each moment based on how congestion is forming. It adapts to prioritize robots that are about to get stuck and reroutes them in advance to avoid bottlenecks. The hybrid system utilizes deep reinforcement learning to figure out which robots should be prioritized. A fast and reliable planning algorithm then feeds instructions to the robots. This setup enables them to respond rapidly in constantly changing conditions. In simulations inspired by actual e-commerce warehouse layouts, the approach achieved about a 25 percent gain in throughput over other methods. The system can quickly adapt to new environments with different quantities of robots or varied warehouse layouts. Lead author Han Zheng said the work demonstrates that deep reinforcement learning can achieve super-human performance, and the research appears in the Journal of Artificial Intelligence Research.