Vibration suppression of welding robot based on chaos-regression tree dynamic model

A torque feedforward vibration suppression tactic, premised on the chaotic-regression tree dynamic model, is proposed to enhance the precision of motion in welding robots during low-velocity movement and to curtail robot oscillation. Chaotic theory is employed to scrutinize the nonlinear characteristics inherent in joint torque amid the low-velocity operation of welding robots. A holistic approach is adopted toward the non-rigid body dynamics component of the joint torque across all joints. By merging ordered fitting with unordered regression tree techniques, the robot kinematic model is derived. The parameters for phase space reconstruction are identified through autocorrelation and pseudo-nearest neighbor methods, improving nonlinear dynamic prediction accuracy via the phase space reconstruction process. To boost the tracking precision of specific motion segments within the trajectory planning, a torque feedforward compensation control algorithm paired with trajectory planning is proposed. Experiments were conducted in low-velocity welding on a 6R welding robot platform, revealing that the improved torque compensation strategy proposed reduces the average position error by 25.2% in comparison with traditional torque compensation tactics. This enhancement optimizes the robot’s motion accuracy and concurrently suppresses end effector vibrations.