Junfeng Xin, Jiabao Zhong, Jinlu Sheng, Penghao Li and Ying Cui
Improved genetic algorithm, path planning, Monte-Carlo simulation, unmanned surface vehicle
The genetic algorithm is an effective method to solve the path-planning problem. However, it has certain disadvantages, such as slow convergence speed, low computational efficiency and premature convergence due to the subjectivity of determining crossover and mutation probabilities. Hence, this paper proposes two data-driven algorithms whose crossover probabilities are adjusted self-adaptively and dynamically with iterations. Monte-Carlo simulations and the application tests to an unmanned surface vehicle are conducted to investigate their path-planning performance. Results indicate that the proposed algorithms have undeniable advantages over the conventional algorithms, such as effective avoiding from the local optimal, keeping higher population diversity during later iterations, reducing the average optimal distance and having better robustness. Meanwhile, it is observed that the reverse linear changing parameters genetic algorithm (RLCPGA) has the superior performance in optimizing the path distance for the unmanned surface vehicle.
Important Links:
Go Back
