ESO-Based Safety-Critical Control for Robotic Systems With Unmeasured Velocity and Input Delay
Sihua Zhang, Di‐Hua Zhai, Juncheng Lin, Yuhan Xiong, Yuanqing Xia, Minfeng Wei
Abstract
For practical robots, obtaining precise dynamic models and states is a challenge, which presents difficulty in achieving safety-critical control. When faced with an uncertain dynamic model of the robotic system and the absence of measurements for joint velocity, this article proposes a method by combining extended state observer (ESO) and control barrier function (CBF) for safety-critical control. Firstly, an ESO is used to estimate the model and states in real time. Then, according to the estimation error, the ESO-based CBF (ESO-CBF) is proposed, and a quadratic programming subject to ESO-CBF is constructed to calculate the control input for robotic systems. In addition, input delay is also considered for robotic systems with uncertain models. In cases involving input delay, a predictive ESO is designed to estimate the model, and the corresponding estimation error boundary is derived. Based on the estimation error, ESO-CBF is constructed to ensure the safety constraint. Finally, the effectiveness of the proposed method is verified by the obstacle avoidance task of Franka Emika Panda manipulator.