This article applies high-order differential estimation to the motion planning of humanoid robots for the first time. A multiobjective optimization model and the corresponding optimal policy are designed from the perspective of solving time-varying linear equations. This method can avoid the calculation of the Jacobian matrix pseudo-inverse and its derivative, reduce energy consumption, and achieve smooth human-like robot motions. High-order differential estimation is realized by cascading multiple integration-enhanced differentiators, which estimate the first derivative based on hybrid error and quasi-sliding mode techniques. The merits of the differentiator include high accuracy in estimating high-order derivatives and the elimination of chattering. Theoretical analyses verify that the proposed differentiator and the differentiator-based solver have asymptotic convergence. Simulations prove that the integration-enhanced differentiator and the differentiator-based method have excellent performance. Experiments illustrate that the designed solver for the motion planning of a humanoid upper-body robot can track desired trajectories and perform carrying tasks.
This paper presents an angular acceleration planning for joint configuration spaces of vision-based kinematically redundant robot manipulators to achieve real-time tracking tasks in dynamic environment. In accordance with the proved proposition, the joint velocity formed by the inverse kinematics mapping of the defined pose error is regarded as the system error, and utilized to further deduce the planned joint angular acceleration and with kinematic redundancy for obstacle avoidance simultaneously. Highly-complex nonlinearities in the angular acceleration are equivalent to a system perturbation represented by a synthetic form, thereby simplifying the tedious calculations of the angular acceleration theoretically and converting the planning problem into a control problem. Hyperbolic tangent-based super twisting algorithm, as the control input of manipulator system, is designed to resist the synthetic perturbation. The collision-free movement of redundant manipulators for tracking tasks is thus achieved through the integration of the control input in real time. An error-based S-function is proposed as the internal parameter of the planned joint acceleration to prevent the integral saturation. Lyapunov theoretical analysis proves the stability of the proposed super twisting algorithm. Simulation and contrast experiments in dynamic-obstacles environments indicate that the proposed joint angular acceleration planning for kinematically redundant manipulators owns feasibility, smoothness and practicability.
1 January 2025 in IEEE Transactions on Instrumentation and Measurement