Collision Avoidance Simulation Using Voronoi Diagrams in a Centralized System of Holonomic Multi-agents

Abstract

AbstractThis work solves the Collision Avoidance problem in a simulation of a centralized system of holonomic multi-agents in a two dimensional space free of static obstacles. For this, we propose an implementation of three modules in an architecture: Threat Assessment Strategy (TAS), Path Planning Strategy (PPS), and Path Tracking Strategy (PTS). The Buffered Voronoi Cells represent the TAS. The PPS modules use two algorithms: the Analytical Geometric Algorithm (AGA) and the Receding Horizons Control (RHC) based on Quadratic Programming (QP) Algorithm. Finally, PTS controls the tracking according to fixed distance magnitudes in each iteration. The analysis of the results considers the computational execution time, the number of steps until convergence, and the calculation of optimal values. Also, these results are compared with the Optimal Reciprocal Collision Avoidance (ORCA) algorithm. In this way, our proposal successfully addresses and solves the collision avoidance problem but takes more execution time and number of steps compared with the ORCA algorithm. Besides, the number of steps of AGA is closer to ORCA, producing promising results with an accuracy of 95%.KeywordsCollision avoidanceVoronoi diagramsConvex optimizationQuadratic programmingPath planningSimulation