A Genetic Algorithm for Autonomous Navigation Using Variable-Monotone Paths.

This paper demonstrates a new genetic algorithm approach for autonomous robot navigation that is based on a variable-monotone path representation. Many contemporary genetic algorithm methods assume that, navigation follows a single-monotone path, where either the x or y coordinate of the path is always non-decreasing. While this approach allows for simple path representation structures, it can limit the success of finding a complete path through complex obstacle navigation environments. To address this limitation, this paper describes the development; of a variable-monotone chromosome structure that allows for more flexibility in the navigation path, while still retaining the basic principles of the existing monotone chromosome structures. A genetic algorithm was developed using this new chromosome structure and tested on ten navigation spaces with varying obstacle complexity. The results show that the variable-monotone approach significantly outperformed the previous single-monotone approaches for navigating complex obstacle spaces.