1. Introduction
Purely global planners such as A* cannot react to obstacles unknown at planning time, while purely local reactive planners such as the dynamic window approach can become trapped in local minima in cluttered environments, motivating hybrid architectures that combine the strengths of both.
2. Methodology
A hybrid planner was implemented in which A* computed an initial global path over a static occupancy grid representation of a 2,000 square metre simulated warehouse, with the dynamic window approach used to locally adjust velocity commands in real time to avoid up to 15 simultaneously moving dynamic obstacles representing personnel and other robots, re-planning the global path when local deviation exceeded a threshold.
3. Results
Across 500 simulated trial runs with randomised obstacle trajectories, the hybrid planner achieved a 99.1 percent task completion rate with zero collisions, compared with 94.3 percent completion and a 2.4 percent collision rate for a DWA-only reactive baseline. Average realised path length was within 8 percent of the theoretical shortest path computed with perfect foreknowledge of obstacle trajectories.
4. Conclusion
Combining global A* planning with DWA-based local reactivity provides robust, near-collision-free navigation in dynamic warehouse settings without significant path-length penalty. Future work will evaluate the planner on physical robot hardware under real sensor noise conditions.
References
[1] Hart P. E. et al., A formal basis for the heuristic determination of minimum cost paths, IEEE Transactions on Systems Science and Cybernetics, 1968. [2] Fox D. et al., The dynamic window approach to collision avoidance, IEEE Robotics and Automation Magazine, 1997.