CGPlan: a scalable constructive path planning for mobile agents based on the compact genetic algorithm
Nenhuma Miniatura disponível
Data
2017-02-16
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
between desired points. These optimal paths can be understood as trajectories that best achieves an
objective, e.g. minimizing the distance travelled or the time spent. Most of usual path planning techniques
assumes a complete and accurate environment model to generate optimal paths. But many of the real world
problems are in the scope of Local Path Planning, i.e. working with partially known or unknown
environments. Therefore, these applications are usually restricted to sub-optimal approaches which plan an
initial path based on known information and then modifying the path locally or re-planning the entire path
as the agent discovers new obstacles or environment features. Even though traditional path planning
strategies have been widely used in partially known environments, their sub-optimal solutions becomes
even worse when the size or resolution of the environment's representation scale up.
Thus, in this work we present the CGPlan (Constructive Genetic Planning), a new evolutionary approach
based on the Compact Genetic Algorithm (cGA) that pursue efficient path planning in known and unknown
environments. The CGPlan was evaluated in simulated environments with increasing complexity and
compared with common techniques used for path planning, such as the A*, the BUG2 algorithm, the RRT
(Rapidly-Exploring Random Tree) and the evolutionary path planning based on classic Genetic Algorithm.
The results shown a great efficient of the proposal and thus indicate a new reliable approach for path
planning of mobile agents with limited computational power and real-time constraints on on-board
hardware.
Descrição
Citação
ASSIS, L. S. CGPlan: a scalable constructive path planning for mobile agents based on the compact genetic algorithm. 2017. 74 f. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Goiás, Goiânia, 2017.