Real-Time Motion Planning in Autonomous Vehicles: A Hybrid Approach

Abstract

In this paper a multi-agent architecture of an Autonomous Robot Navigator for a vehicle that operates in dynamic real-world environments is presented. The vehicle is capable of executing different navigation missions while smoothly avoiding static obstacles in its path as well as moving objects. The navigator architecture is part of a general multi-agent cognitive framework, which is organised into three non-hierarchical components characterised by the type of knowledge they deal with: a symbolic component, handling a declarative explicit propositional formalism, a diagrammatic component, dealing with analogical, iconic representations, and a reactive behaviour based component. The navigator is distributed in all three components combining motion planning on a topological graph with reactive motion planning techniques. It is on these aspects that the paper focuses. Experimental results with our mobile robot will also be provided.