Multi-agent path planning with nonlinear restrictions
Autor: | Juliana Fonseca de Lima, Vinícius Abrão da Silva Marques, Eduardo Morgado Belo |
---|---|
Rok vydání: | 2021 |
Předmět: |
Mathematical optimization
Computer science Cognitive Neuroscience Particle swarm optimization 020206 networking & telecommunications 02 engineering and technology Algebraic equation Nonlinear system ALGORITMOS GENÉTICOS Mathematics (miscellaneous) Artificial Intelligence Differential evolution Genetic algorithm Convergence (routing) 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Computer Vision and Pattern Recognition Motion planning Heuristics |
Zdroj: | Repositório Institucional da USP (Biblioteca Digital da Produção Intelectual) Universidade de São Paulo (USP) instacron:USP |
ISSN: | 1864-5917 1864-5909 |
Popis: | This paper presents a novel simplified mathematical modeling that can solve a path planning problem comprising three agents in a triangular formation. The problem is modeled to optimize the trajectories of the agents and to minimize the distances traveled. The trajectories are obtained in a way that deviates them from fixed obstacles whose dimensions are known. Furthermore, geometry and orientation constraints of the formation of the multi-agent are imposed. Heuristics methods such as the Genetic Algorithm, Differential Evolution and Particle Swarm Optimization are applied to solve the non-linear algebraic equations which represent the system. Comparisons are presented among the results of the different methods highlighting the processing time and the convergence to the global minimum solution. The results prove that all the algorithms can be applied to the path planning problem with constraints. The Particle Swarm Optimization method presents the lowest processing time for all simulated cases. However, the Differential Evolution method is more robust than the others when searching for the global minimum solution. |
Databáze: | OpenAIRE |
Externí odkaz: |