Nonparametric Terrain Estimation for Planetary Rovers Based on Noncontact Rut Measurement

Abstract

The estimation of terrain mechanical properties is an important mission of planetary exploration rovers. Since only limited resources and human supervision are available, efficient and accurate methods for terrain estimation are required. In this paper, a nonparametric estimation method for the prediction of the terrain parameters is presented. Four characteristic parameters based on telemetry information of planetary rovers were adopted in the method, including wheel load W, torque T, rut depth (hereinafter referred to as apparent sinkage za), and slip ratio s, and 72 identification parameters composed by the characteristic parameters were put forward for terrain estimation. The terrain consisted of simulant regolith classified nonparametrically and artificially defined as three states (loose, natural, and compact states) based on different hardnesses, and four terrain estimation criteria (slack, ideal, partially strict, and strict) were established. Wheel–soil interaction tests were conducted for the terrain estimation, and apparent sinkage based on rut noncontact measurement was analyzed. The optimal identification parameters and estimation criteria were derived through the analysis of evaluation indexes (correct rate, accurate rate, and conservation rate). Experimental validation results indicated that the proposed method could accurately and efficiently estimate the mechanical state of deformable terrain for planetary rovers.