关于四足机器人足端行走轨迹优化设计仿真

Optimal design and Simulation of Quadruped Robot Foot Walking Trajectory

四足机器人足端行走轨迹目前主要依赖运动学模型关系实现,针对实际行走轨迹与规划轨迹存在较大偏差,造成偏差的原因具有多样性,难以建立准确模型对偏差进行处理的问题,利用自适应神经模糊推理系统在Matalb中经过仿真建立四足机器人的足端轨迹坐标点和关节变量的新的关系模型,该模型中采用反向误差传播算法和最小二乘法结合的混合算法来优化模型参数,模型的建立依赖于原始数据的训练,原始数据包含了偏差因素影响。以一种自行开发的四足机器人样机为平台进行现场验证,引入ANFIS模型的直线轨迹相比较初始的直线轨迹偏差减小,更加接近规划直线。上述方法能运用在不同结构的四足机器人行走轨迹优化中。

Quadruped robot foot walking trajectory currently relies mainly on the kinematics model of relationship, for the large deviations between the planning trajectory and actual trajectory, the cause of the deviation is diversity, it is difficult to establish accurate model to process the deviation problem. The simulation of adaptive neural fuzzy inference system in Matlab is used to establish a new relationship model of quadruped robot foot trajectory coordinates and the joint variables, the hybrid algorithm model which combined BP algorithm with least square method to optimize the parameters of the model, the model depends on the training of original data, the original data contains the deviation factor. With a self - developed prototype of quadruped robot as a platform for on - site verification, the ANFIS model reduced the linear trajectory deviation compared to the initial model, closer to the planning straight line. The method can be used in the quadruped robot walking trajectory optimization of different structure.