Selected impact

Improving robot fleet routing in production

Routing improvements at Walmart Advanced Systems & Robotics increased pick rates and reduced long-tail task times in production.

Challenge

An automated pickup and delivery system depends on robots sharing space and completing tasks together. Routing decisions affect pick rates and the time needed to complete dispense and product tasks. Improvements need to be evaluated across the fleet, including the tasks that take longest to complete.

Approach and contribution

At Walmart Advanced Systems & Robotics, I proposed and prototyped routing improvements and worked with software engineers to deploy them. The evaluation covered pick rates, long-tail dispense and product task times, search effort, and excessive robot movements.

Results

The deployed routing improvement achieved 10% higher pick rates and a 30% reduction in long-tail dispense and product task times. It also reduced state expansions by 20% and excessive robot movements by 10%.

A separate deadlock-free multi-agent routing prototype demonstrated a further 50% reduction in long-tail task times and a 25% improvement in pick rates relative to the already improved routing layer. These additional gains were measured in the prototype comparison.