Research On Terrain Parameter Estimation And Non-geometry Obstacle Identification For Lunar Rover
- Paper number
IAC-07-A3.I.A.17
- Author
Prof. Ju Hehua, Chinese Society of Astronautics, China
- Coauthor
Mr. LIU Bing, Beijing University of Technology, China
- Year
2007
- Abstract
This paper presents the research on terrain parameter estimation and non-geometry obstacle identification, which are used for lunar rover navigation and control. Firstly, because of low cohesion of lunar terrain, the key soil parameters to be identified are internal friction angle and pressure-sinkage coefficient. The proposed technique is based on mechanics sub-system which is constructed by six-axis force/torque sensor and other on-board sensors. The mechanics model for rigid wheel of six-wheel drive lunar rover on soft terrain is derived, which the wheel turns in steady state. And simplified distribution function of shear stress and the relevant simplified resolving equations for estimation terrain parameters are obtained. Based on sensed quantities by force sensor and other on-board sensor, internal friction angle and pressure-sinkage coefficient could be estimated. The algorithm could fast and efficiently identify key terrain parameters. Secondly, a novel method of non-geometry identification of long range terrain is discussed in this paper. Using vision sub-system of lunar rover, the texture feature of terrain underneath and near the rover front wheel can be obtained, then using terrain texture feature and the corresponding mechanics parameters to update the feature database which represents the traversed terrain, and lastly, based on the information in database and the texture feature of terrain in front of the rover, the mechanics parameters of terrain in front of the rover can be estimated, which in turn can be used to identify the non-geometry obstacle in rover’s scene. In order to realize the idea that discussed above, the validity of texture feature is crucial, so a terrain classification method based on distribution feature of high frequency energy is presented. The method implement wavelet multi-resolution analysis for the multi-level decomposition of a terrain image to compute high frequency energy of every decomposed level, and the expectation of the high frequency distribution curves is taken as the texture feature. Then, after selective filter and vote classification, the classification result is obtained based on maximum likelihood approach. Finally, simulation and experimental results show that the proposed technique is feasible.
- Abstract document