【摘要】 针对仿人机器人攀爬的实时运动生成问题,提出一种基于非线性模型预测控制(nonlinear model predictive control,NMPC)方法,能够综合优化路径和肢体运动。该方法将攀爬任务视为一个机械约束的NMPC问题,并使用了基于墙体图的状态相关权重和势函数。在每个采样时间点根据墙体信息和机器人的状态进行计算获得控制输入。此外,还提出了为NMPC在线配置性能指标的视距评估方法。研究结果表明:随着视距的减小,控制输入的计算时间也随之减少,有效降低了计算成本;与将墙体上的所有支撑都纳入视距范围的情况相比,攀爬时间最多能减少36.4%,有效适应了复杂的墙体模型。更多还原

【Abstract】 Aiming at the problem of real-time motion generation of humanoid robot climbing, a control method based on nonlinear model predictive control(nonlinear model predictive control, NMPC) is proposed, which can comprehensively optimize the path and limb motion. In this method, the climbing task is regarded as a mechanically constrained NMPC problem, and the state-dependent weight and potential function based on wall graph are used. At each sampling time point, the control input is calculated according to the wall information and the state of the robot. In addition, a line-of-sight evaluation method for configuring performance indicators for NMPC is proposed. The results show that:with the decrease of the line-of-sight distance, the calculation time of the control input is also reduced,which effectively reduces the computational cost;compared with the case where all the supports on the wall are included in the line-of-sight range, the climbing time can be reduced by up to 36.4%, effectively adapting to complex wall models.更多还原