PWL Approximation for Dense Mapping and Associated Hybrid PSO-Dijkstra Processes for Path Planning
Karime Pereida, José Guivant, Anton Lohr · 2014
This work proposes and illustrates an ecient path planning algorithm for nonholonomic platforms in dense contexts. The aim is to compute optimal paths in a map described by dense properties and including the robot’s constraints. The proposed algorithm integrates a low dimensional Dijsktra cost-to-go function, calculated over a QT-PWL approximation, with Particle Swarm Optimisation (PSO) minimisation, performed in the higher dimensional conguration space. The methodology includes adaptations to reduce the processing cost and increase the optimality of the planned path. Experimental results show the performance of the algorithm.