双足机器人节能步态规划算法

Energy-efficient gait-planning algorithm for biped robots

为提高双足机器人的步行性能,提出基于五质心倒立摆模型的节能步态规划算法。算法包括步态参数优化算法和步态合成算法。步态参数优化算法允许身体做三维运动,以有限阶傅里叶级数的系数表征特定步长下机器人身体的运动空间。通过离散化这些系数,使运动空间网格化。进而对网格交点进行逆动力学计算,划分出满足允许零力矩点区域要求的种子集合。算法以电机的负荷转矩和角速度的乘积为能耗指标函数,在每个种子的邻域迭代计算。按照最大梯度原则逐次逼近函数极小值,此时的电机角度序列作为对应步长下的解,存入数据库。步态合成算法按照步行距离,规划由起始步、中间步和停止步构成的完整行走轨迹。按照行走步长,从数据库读取腿关节电机的角度序列,并依据双足机器人行走中反馈的零力矩点,对序列进行修改。为验证算法有效性,进行了动态仿真实验和现实环境中双足步行实验。实验结果与固定身体高度或允许身体垂直运动的算法对比,证明步态算法具有明显的节能效果。该算法实现低能耗和高鲁棒性的折中,较好地解决具有高度非线性特征的双足机器人行走问题,为煤矿救援机器人的开发开辟一种新途径。

An energy-efficient gait-planning algorithm based on five-centroid inverted pendulum model is presented in the paper for improving the walking performance of biped robot.The gait-planning algorithm is divided into gait parameter optimization(GPO)algorithm and gait synthesis(GSYN)algorithm.GPO algorithm allows the body to perform three-dimensional motion,and represents the motion space of the robot body in a given step length with coefficients of the finite-order Fourier series.By discretizing these coefficients,the motion space is gridded.Then inverse-kinemics calculations of grid intersection points are carried out,and the seed set satisfying the requirement of the allowable zero moment point region is divided.The algorithm takes the product of the load torque and angular velocity of motors as the energy-consumption index function,and calculates iteratively it in the neighborhood of each seed.According to the principle of maximum gradient,the minimum value of the function is approximated successively,the current angle sequences of motors are stored in the database as the solution under the corresponding step length.Given a distance to be traveled,GSYN algorithm plans a complete walking trajectory,i.e.,two starting steps,multiple cyclic steps,and two stopping steps.Based on the walking step length,angle sequences in leg joints is read from the database,and these sequences are modified in lignt of the zero-moment-point feedback during the biped robot walking.In order to identify effectiveness of the proposed algorithm,the dynamic simulation experiment and the biped walking experiment in the real environment were conducted.The experimental results show that the gait algorithm in the paper has a significant energy-saving effect compared with the algorithm of the fixed body height or allowed body vertical motion.The algorithm can optimize the tradeoff between low energy consumption and high robustness and solve the problem of walking in biped robots with highly nonlinear characteristics,which opens up a new way for the development of coal mine rescue robots.