论文链接：https://ieeexplore.ieee.org/document/10839230

B站链接: https://www.bilibili.com/video/BV1YKfnYiExd

现有的不确定环境下机器人-环境交互力控制方法在交互建模、控制算法等方面存在缺陷。交互建模方面：（1）大多仅考虑环境位置、刚度、阻尼等结构化不确定性，忽视了交互过程中机器人底层位置跟踪误差、建模误差、未知外部干扰等非结构化不确定性的存在；（2）模型相对阶为零，难以应用现有控制理论严格分析系统稳定性和性能。控制算法方面：（1）基于交互模型在线调节阻抗/导纳参考轨迹控制策略严重依赖交互模型的准确性，对结构化和非结构化不确定性敏感；（2）可变阻抗/导纳控制方法瞬态超调大，底层位置跟踪误差等非结构化不确定性对稳态力控精度影响大，难以实现交互力精确控制。

不同于现有的相对阶为零的静态交互模型，本文提出一种相对阶为一且同时建模了结构化和非结构化不确定性的机器人-环境动态交互模型，它能够更加全面、精确的捕捉交互动力学的复杂性，支持采用现有理论严格分析稳定性和收敛性。基于该模型，我们设计了一种自适应鲁棒交互控制器（ARIC）。该控制器通过引入自适应机制在线准确估计模型结构化不确定参数以实现精确的前馈补偿，同时通过集成鲁棒反馈控制律有效抑制包含底层位置跟踪误差在内等非结构化不确定性。最后给出了严格的稳定性分析和证明，并在七自由度机械臂上开展了不同场景下的接触力跟踪实验。其中，在3D人体腹部医学体模测试场景下，与其他多种典型力控算法相比，所提方法的稳态接触力跟踪MSE减少了20.0%-82.4%，且在率先到达期望接触力的同时，瞬态超调最小。

        J. Huang, M. Yuan*, Z. Huo, S. Zhang and X. Zhang. Adaptive Robust Interaction Force Control of a Robotic Manipulator in Uncertain Environments. IEEE Transactions on Industrial Electronics (T-IE), doi: 10.1109/TIE.2024.3525103.

Abstract

        This article presents a novel adaptive robust interaction force control approach that aims to achieve accurate and robust force modulation of a robot manipulator in uncertain environments. First, a novel interaction model between manipulator and environment is developed that captures both structured and unstructured uncertainties. According to the developed interaction model, a two-loop control scheme is developed, which consists of an adaptive robust interaction controller (ARIC) in the outer loop and a motion tracking controller in the inner loop. With the proposed ARIC, the unknown structured parameters of the developed interaction model are estimated online. These estimated parameters are then utilized to generate feedforward compensation actions in the mechanism of ARIC, while unstructured uncertainties are addressed through robust feedback control so that excellent interaction force tracking accuracy can be ensured. In addition, the theoretical stability analysis of the ARIC is presented and practical experiments are carried out on a robot manipulator. The experimental results confirm the superiority of the proposed approach over these typical algorithms in terms of interaction force tracking performance in different environments.