Identification of the shear parameters for lunar regolith based on a GA-BP neural network

Abstract

Identifying the mechanical parameters of lunar soil through the rover’s wheel can provide the basic data for path planning, risk avoidance, and traction control. In this paper, the shear parameters of lunar soil are identified by a Back Propagation neural network optimized by Genetic Algorithm (GA-BP) based on a simplified wheel-soil model. For the GA-BP identified model, the input data are driving torque (T), vertical load (W) and slip ratio (s) of the wheel. The output data are internal friction angle (φ) and shear deformation modulus (K). A total of 315 sets of data are used to train GA-BP and BP algorithms. Data from single-wheel soil bin test are put into the trained algorithms to identify φ and K of lunar soil simulant. The test results demonstrate that GA-BP algorithm is accurate and effect to identify shear parameters of regolith online. The comparison between identified results and test results shows that the GA-BP algorithm is better than the BP algorithm. The cohesion is set to 1.5 kPa and then the drawbar pull is predicted according to the identified φ and K of GA-BP. The test results show that the prediction of DP is reasonable.