Improved Genetic Algorithms for Mobile Robot Path Planning

Abstract

An improved genetic algorithm (IGA) is proposed to address the problems of the simple genetic algorithm (SGA) in the path planning of mobile robots, such as the paths not being smooth enough, being prone to fall into local optima, and having an unstable algorithm. Firstly, an intermediate value insertion method is adopted to improve the initial population quality. Secondly, in addition to the path length function, a turn angle penalty function is also introduced in the fitness function to achieve the purpose of smoothing the path. Then, a hybrid selection strategy combining roulette and elite retention is adopted to ensure population diversity and avoid the loss of optimal solutions. Meanwhile, a single-point crossover operator for high and low populations is designed, and a strong variation operator is improved to enhance the search efficiency of the algorithm. An adaptive strategy for crossover and variation probabilities with the number of iterations is designed. Finally, an optimization operator generates a quadratic optimization of the optimal path. The simulation results show that these improvements can significantly improve the convergence speed and performance of the algorithm.