Signal A retailer's fleet foundation model is reported to raise robot travel efficiency by 10 percent

Summary

Amazon says a learned coordination model has raised the travel efficiency of its warehouse robots by 10 percent. CB Insights cites chief executive Andy Jassy on the company's second quarter 2025 earnings call describing the model, called DeepFleet. Jassy said that at Amazon's scale such a gain is significant. DeepFleet works like a traffic management system, coordinating robot movements to find the best paths and cut bottlenecks. It learns from billions of hours of robot data to predict flows, a shift away from classical rule-based scheduling algorithms. Amazon has deployed around one million robots, so learned coordination is being applied at very large scale.