The electro-hydraulic controlled bionic robots suffer from path contour deviation, which is affected byinertia, system lag, and control system accuracy. The focus of this paper is on the study of a data-driven systemfor the compensation of errors in the robot’s internal arithmetic model and its own motion. A deviation predictionmodel of the robot motion process is constructed using a machine learning approach, while the spatial error, generated in the transfer process between the input trajectory and the actual output, is de-parameterized and regardedas an attribute value of the robot’s motion process. Furthermore, a data-based adaptive compensation method isproposed. The simulation model and a hardware-in-the-loop simulation platform of the proposed control strategyare constructed, in order to verify the proposed control approach. Simulation and experimental results show that theproposed compensation strategy can significantly reduce system deviation.