基于PSO-BP算法的水下机器人运动模型辨识

Kinematic Model Identification of Underwater Vehicle Based on PSO-BP Algorithm

将粒子群算法(PSO)与误差反传算法(BP)相结合,利用粒子群算法的全局突变性,使BP算法避免在神经网络权值寻优过程中陷入局部极小值。对Elman神经网络结构进行调整,并将PSO-BP算法用于改进后的Elman网络的权值修改。最后,对比了3种不同算法、结构的神经网络对水下机器人运动学模型的辨识结果,证明了基于PSO-BP算法的改进Elman神经网络对水下机器人运动模型,有较高的辨识精度。

Combine particle swarm optimization （PSO） and back - propagation algorithm （BP）, which take advantages of the global mutation of PSO, to prevent BP algorithm getting stuck in local minima during the weights optimization of neural network. Make use of a modified Elman neural net-work whose weights were optimized by the PSO-BP algorithm, to identify the kinematic model of underwater vehicles and the simulation experi-ments, comparison between the modified Elman network based on PSO and another two different neural network, showed that the modified Elman network based on PSO - BP had a better accuracy in the kinematic model identification of underwater vehicles.