An Evaluation of Path Planners for Guidance With Vision Based Simultaneous Localization and Mapping
Holly Borowski, Eric W. Frew · 2012
This paper discusses implementation and compares results from three dierent path planning methods for guidance through a cluttered environment using vision-based simultaneous localization and mapping (SLAM). The planners are implemented using sequential quadratic programming (SQP), a genetic algorithm, and a rapidly-exploring random tree (RRT). Various planning horizon lengths are explored. The planners are compared for resulting path length from a start to a goal location and back, obstacle avoidance, and computation time. A short planning horizon SQP method resulted in shortest path lengths, a medium horizon genetic algorithm produced the best obstacle avoiding trajectories, and the short horizon SQP planner required the least planning time.