关键词: 参数识别, 估值误差, 气垫车, 土壤参数, 越野机器人

Abstract: Correct on-line identification of soil parameters is the basis of performance optimization and control for robots navigating on soft terrain, in which, the problems of multiple-solution and accuracy should be solved. By taking advantage of the additional degree of control freedom of air-cushion-typed off-road robots for vertical force to eliminate the multi-solution problem, a new identification algorithm, g-function algorithm, is proposed to estimate the three soil parameters related to the tractive effort in real time. The implementation of the g-function algorithm requires three sampling points so that it is essential to establish their selection rules to decrease the estimation error sourced from state noises and measurement noises. This decrease is embodied in three aspects, with respect to which, selection rules are respectively established by mathematical derivation firstly and then are balanced after a trade-off consideration. Based on an engineering example, experiments on estimation accuracy are conducted in different levels of state noises and measurement noises. The results show: ① The g-function algorithm succeeds to identify the three targeted soil parameters in all noise levels despite existing more or less errors; ② The ideal combination of sampling points that is obtained by using the selection rules presents an obvious advantage over random combinations in all noise levels; ③ System nonlinearity makes estimation accuracy sensitive to both state noises and measurement noises. The above results, therefore, support the necessity and feasibility of the g-function algorithm and its sampling points’ selection rules.

Key words: Air-cushion vehicle, Estimation error, Off-road robot, Parameter identification, Soil parameter