Adaptive Path Planning for Collaborative Robots in Dynamic Warehouse Environments

Abstract

Collaborative robots deployed in warehouse fulfillment operations must navigate alongside human workers and shifting inventory layouts, conditions that challenge path planning algorithms designed for static or fully autonomous environments. This paper presents an adaptive path planning approach that continuously updates local trajectory segments in response to detected human proximity and temporary obstacles, while preserving global route efficiency computed at the task allocation level. The method combines a sampling-based global planner with a reactive local layer that adjusts robot velocity and heading based on predicted pedestrian trajectories rather than reactive stopping alone, aiming to reduce disruptive halting behavior that slows overall throughput. We evaluate the approach in a simulated warehouse environment modeled on real facility layouts, comparing task completion time, near-miss frequency, and worker-reported comfort against a conventional reactive avoidance baseline. Results show that the adaptive approach reduces unnecessary full stops while maintaining comparable safety margins, and worker feedback indicates smoother, more predictable robot behavior around shared aisles. We discuss practical considerations for deploying the approach on physical robot fleets, including sensor latency and computational budget constraints relevant to real-time warehouse operations.