基于混合微粒群算法的智能水下机器人模糊神经网络控制

Fuzzy neural network control of AUV based on HPSO algorithm

为减少因水下机器人模糊神经网络控制器参数较多、手工调整困难及主观不确定性因素的影响,提出一种基于免疫理论和惯性权值非线性递减策略的混合微粒群算法.该算法在保持基本微粒群算法处理多峰和多维问题能力的基础上,根据粒子浓度和适应度来动态调整约束因子,同时结合惯性权值非线性递减策略来抑制算法早熟收敛,平衡全局和局部搜索能力.在与GAI、GA及基本微粒群算法的仿真比较试验中,该算法搜索到最佳近优解,且其收敛速度最快.在水下机器人仿真平台上的控制试验表明,基于混合微粒群算法的控制器性能良好,具有较强的抗海流干扰能力.仿真结果证明了该算法的可行性.

To reduce the numerous work and subjective uncertainty in manual adjustments for underwater vehicles, a hybrid particle swarm optimization （HPSO） algorithm based on immune theory and nonlinear decreasing inertia weight （NDIW） strategy was developed. Particle＇s concentration was introduced into the algorithm from the immune theory, and the restraint factor is dynamically tuned according to particle＇s concentration and fitness. Owing to the restraint factor and NDIW strategy, HPSO algorithm can effectively prevents premature convergence and keeps balance between global and local searching ability. Meanwhile, the algorithm maintains the ability of handling multimodal and multidimensional problems. HPSO algorithm has the fastest convergence velocity and finds the best solutions compared with genetic algorithm （GA）, immune genetic algorithm（IGA）, and basic particle swarm optimization （PSO） algorithm in simulation experiments. The experimental results on the autonomous underwater vehicle（AUV） simulation platform show that HPSO-based controllers perform well and have strong abilities against current disturbance. Simulation results verify the feasibility in application to AUV.