基于改进hopfiled网络的机器人路径优化控制

Robot path optimization control based on improved hopfiled network

摘要:

针对现有移动机器人路径优化算法存在的迭代效率低、路径规划能力差等问题,提出一种基于改进hopfiled神经网络的机器路径优化算法。首先,在世界坐标系内构建移动机器人空间运动模型,掌握移动机器人不同时刻的位置信息和移动信息；其次,构建hopfiled神经网络模型,并利用BP网络优化hopfiled神经网络模型的结构,提升其数据训练能力；同时利用LSTM网络的门控结构替代原网络隐含层的神经元,引入遗忘门、输入门和输出门,提升hopfiled神经网络的泛化学习能力和样本容纳能力；最后引入路径评价函数,评价局部区域内的碰撞风险以降低移动机器人之间的碰撞概率。实验测试结果显示：提出的改进hopfiled神经网络模型路径规划均值为104.3m,耗时均值为122.1s,随机提取采样点的方差值仅为0.01,显著低于其他的传统路径优化算法。

Abstract:

Aiming at the problems of the existing mobile robot path optimization algorithms, such as low iteration efficiency and poor path planning ability, a machine path planning algorithm based on hopfiled neural network is proposed. Firstly, the space motion model of the robot is constructed in the world coordinate system to grasp the position information of the mobile robot at different times. Secondly, the hopfiled neural network model is constructed, and the structure of hopfiled neural network model is optimized by BP network to improve its data training ability. LSTM network structure is used to replace the hidden layer neurons, and forgetting gates, input gates and output gates are introduced to improve the generalization learning ability and sample holding ability of hopfiled neural network. Finally, the path evaluation function is introduced to evaluate the collision risk in the local area to reduce the collision probability between mobile robots. The test results show that the path planning average of the proposed improved hopfiled neural network model is 104.3m, the time is 122.1s, and the variance value of random sampling points is only 0.01, which is significantly lower than other traditional algorithms.