关键词: 双轮腿机器人；子空间辨识, 关节摩擦力；积分滑模控制

Abstract: The two-wheeled robot switches between foot and wheel modes to adapt to different terrains and environments, and the frequent switching of motion modes makes the joint friction have an inevitable nonlinear and uncertain effect on the system. To address the problem of model errors caused by neglecting or simplifying joint friction during the modeling process, this paper proposes a nonlinear integral sliding mode control strategy for a two wheeled legged robot based on subspace identification. Firstly, the influence of joint friction on the motion performance of the robot was analyzed based on the Coulomb-Stribeck friction theory. Then, a state space model for a two-wheeled legged robot considering joint friction was established using a subspace identification algorithm. To address the interference issue of colored noise generated by joint friction, a sparse projection method was employed to eliminate the bias of the noise term on the subspace identification algorithm. Subsequently, an improved integral sliding mode control law was designed to achieve attitude control of the two-wheeled legged robot, and the stability of the control system was proved using Lyapunov theory. Finally, the simulation experiment results show that compared with the control system of the two wheeled legged robot that ignores joint friction, the model errors are respectively reduced by 28.94%, 41.97%, 35.43%, and 40.61%. The overshoot of pitch angle and yaw angle has been respectively reduced by 24.26% and 21.1%, and the system has good stability, disturbance rejection, and robustness.

Key words: Two-Wheeled-legged robot, Subspace identification, Joint friction force, Integral sliding mode control