Hardware-in-the-loop Oriented Modeling of the Road Condition for Unmanned Roller Motion Simulation Using Gaussian Process Regression

The rolling earth-rock dam's undulating surface has a significant impact on the path-following control performance of the roller vehicle. In this paper, a road condition model in terms of slope for the hardware-in-the-loop (HIL) platform application, is proposed, for the unmanned roller. To address the randomness and nonlinearity of the dam's surface slope under repetitive compactions, the Gaussian process regression (GPR) algorithm is adopted. The model inputs take account of the times of compactions and local location, and the output variable is the dam's local surface slope. Experimental results show that the coefficient of determination between the estimated slope and the measured value is up to 0.828 on average. By combing the proposed model with the existing kinematic model, the accuracy of the combined model is improved, in terms of the maximum deviation between the estimated lateral coordinate curve and actual curve reduced by 16.99%.