Multiobjective hybrid evolutionary path planning with adaptive pareto ranking of variable-length chromosomes

This paper presents a novel approach to multiobjective evolutionary robot path planning. The paths are designed for bi-dimensional continuous working scenes with disjoint, nonconvex obstacles. A new ranking procedure ensures a self-adapted progressive articulation between decision and search, performed in relation to the landscape of the objective space. The ranks are assigned by means of dominance analysis, while making use of adaptive grouping of individuals and adaptive control of diversity. The algorithm also accepts variable-length chromosomes and uses a new compatible crossover, which avoids the production of offspring longer than their parents. The unfeasible paths are repaired and shortened by means of a new corrective algorithm. This correction works only once on a specific chromosome and its main role is to guide the genetic search towards the feasible regions of the search space. The experiments demonstrate the effectiveness of the suggested techniques on several working scenes with different maps of obstacles.