基于GA优化RBF神经网络的机器人轨迹规划

针对机器人在不确定环境下末端执行器运动轨迹的准确性及平稳性问题,采用基于遗传算法(GA)优化径向基函数(RBF)神经网络的轨迹规划方法对Kinova Mico2机器人进行轨迹规划研究。介绍了机器人的相关参数及坐标系、建立了D-H矩阵和运动学模型。提取机器人实际抓取物品的直线轨迹并等分插补,用GA优化并实时在线更新RBF神经网络的权值,以更优的权值参数建立新的RBF网络。研究结果表明:相比优化前,基于GA优化RBF的规划轨迹逼近误差小且平滑稳定,仿真结果较为稳定,轨迹规划的可行性满足机器人实际抓取工作的需要。

Robot Trajectory Planning Based on GA Optimized RBF Neural Network

Aiming at the accuracy and stability of the end effector's motion trajectory in uncertain environment,the trajectory planning method based on neural network of optimized radial basis function(RBF)with genetic algorithm(GA)is used to study the trajectory planning of Kinova Mico2 robot.The related parameters and coordinate system of the robot are introduced,and the D-H matrix and kinematics model are established.Extract the linear trajectory of the robot to actually grab the item and divide it equally,use GA to optimize and update the weight of the RBF neural network online,and establish a new RBF network with better weight parameters.The research results show that the planning trajectory approximation error based on GA optimization RBF is small and smooth and stable,and the simulation results are stable.The feasibility of trajectory planning can meet the needs of the actual robot grabing work.