Knowledge Incorporation into ACO-Based Autonomous Mobile Robot Navigation

摘要

A novel Ant Colony Optimization (ACO) strategy with an external memory containing horizontal and vertical trunks from previously promising paths is introduced for the solution of wall-following robot problem. Ants construct their navigations by retrieving linear path segments, called trunks, from the external memory. Selection of trunks from lists of available candidates is made using a Greedy Randomized Adaptive Search Procedure (GRASP) instead of pure Greedy heuristic as used in traditional ACO algorithms. The proposed algorithm is tested for several arbitrary rectilinearly shaped room environments with random initial direction and position settings. It is experimentally shown that this novel approach leads to good navigations within reasonable computation times.