AUTHOR=Zhang Jie , Kang Man , Li Xiaojuan , Liu Geng-yang TITLE=Bio-Inspired Genetic Algorithms with Formalized Crossover Operators for Robotic Applications JOURNAL=Frontiers in Neurorobotics VOLUME=11 YEAR=2017 URL=https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2017.00056 DOI=10.3389/fnbot.2017.00056 ISSN=1662-5218 ABSTRACT=

Genetic algorithms are widely adopted to solve optimization problems in robotic applications. In such safety-critical systems, it is vitally important to formally prove the correctness when genetic algorithms are applied. This paper focuses on formal modeling of crossover operations that are one of most important operations in genetic algorithms. Specially, we for the first time formalize crossover operations with higher-order logic based on HOL4 that is easy to be deployed with its user-friendly programing environment. With correctness-guaranteed formalized crossover operations, we can safely apply them in robotic applications. We implement our technique to solve a path planning problem using a genetic algorithm with our formalized crossover operations, and the results show the effectiveness of our technique.