Trajectory Tracking Control of Wall-Climbing Robots for Nuclear Power Plant Maintenance with Dynamic Slip-Skid Compensation

Wall-climbing robots are increasingly deployed in high-risk industrial environments such as nuclear power plants, where high-precision trajectory tracking is critical for tasks including surface inspection, coating, and corrosion removal. However, severe slip and yaw deviations caused by gravity, surface irregularities, and adhesion uncertainty significantly degrade tracking accuracy and control stability when conventional kinematic or model-based controllers are applied. To address this challenge, this paper proposes a dynamic slip–yaw compensation trajectory tracking framework for tracked wall-climbing robots. First, a coupled slip–yaw dynamic model is established to characterize the nonlinear deviation between ideal and actual robot motion under varying pose and adhesion conditions. To overcome the computational complexity of the analytical model, a data-driven slip and yaw prediction model based on Support Vector Regression (SVR) is developed, with its hyperparameters optimized using the Butterfly Optimization Algorithm (BOA) to achieve high prediction accuracy and real-time performance. On this basis, a Model Predictive Control (MPC) strategy incorporating dynamic slip–yaw compensation is designed, in which the predicted motion deviations are explicitly embedded into the prediction model and optimization constraints. Experimental results on both simulated and physical wall-climbing robot platforms demonstrate that the proposed method significantly improves trajectory tracking accuracy and control smoothness under complex trajectories and curved surfaces. Compared with conventional controllers, the proposed approach achieves a 23.7% reduction in RMS tracking error and a 34.2% reduction in peak error in sinusoidal trajectory tracking, while effectively suppressing control input jitter. These results indicate that the proposed framework provides a robust and practical solution for high-precision autonomous operation of wall-climbing robots in complex industrial environments.