Ant-ViBRA: A Swarm Intelligence Approach to Learn Task Coordination

Situated Learning on FPGA for Superscalar Microprocessor Design Education

Reinaldo A. C. BianchiAnna H. Reali Costa

In this work we propose the Ant-ViBRA system, which uses a Swarm Intelligence Algorithm that combines a Reinforcement Learning (RL) approach with Heuristic Search in order to coordinate agent actions in a Multi Agent System. The goal of Ant-ViBRA is to create plans that minimize the execution time of assembly tasks. To achieve this goal, a swarm algorithm called the Ant Colony System algorithm (ACS) was modified to be able to cope with planning when several agents are involved in a combinatorial optimization problem where interleaved execution is needed. Aiming at the reduction of the learning time, Ant-ViBRA uses a priori domain knowledge to decompose the assembly problem into subtasks and to define the relationship between actions and states based on the interactions among subtasks. Ant-ViBRA was applied to the domain of visually guided assembly tasks performed by a manipulator working in an assembly cell. Results acquired using Ant-ViBRA are encouraging and show that the combination of RL, Heuristic Search and the use of explicit domain knowledge presents better results than any of the techniques alone.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: