Set-membership based real-time terrain modeling of mobile robots with a laser scanner

This paper presents a new set-membership based real-time terrain estimation algorithm to generate the elevation map of rough terrain by a mobile robot with a scanning laser in outdoor environments. Unlike conventional probabilistic algorithms, all underlying uncertainties are modeled as unknown-but-bounded noises without other assumptions, and the set-membership framework is then proposed to construct the set-valued elevation map with elevation variation in each cell bounded by intervals. The pose of the mobile robot is also estimated simultaneously by the set-membership filter and fused to the elevation map to modify the accuracy of the terrain model. Simulations and experimental results demonstrate the effectiveness and robustness of the proposed method.