A new real-time algorithm for off-road terrain estimation using laser data

Abstract

Gaussian mixture algorithm (GMA) is an effective approach for off-road terrain estimation, but still suffers from some difficulties in practical applications, such as complex calculation and object abstraction. In this paper, GMA is modified to improve its real-time performance and to provide it with a potential ability of obstacle detection. First, a selection window is designed based on the dominant-ellipse-principle to limit the probability distribution area of each measurement point, therefore avoiding the calculation on the cells outside the dominant ellipse. Second, a clustering approach is proposed in order to distinguish objects efficiently and decrease the operation area of one laser scan. Third, a virtual point vector is introduced to further reduce the computational load of the mean square error matrix. The modified GMA is experimented on a tracked mobile robot, and its improved performance is shown in comparison to the original GMA.real-time, off-road, terrain estimation, laser data

A Kalman-filtering-based Approach for Improving Terrain Mapping in off-road Autonomous Vehicles

“…Terrain map estimation is an important component in outdoor mobile robot navigation, and several authors have addressed it ( [2][3][4][5][6][7][8][9][10][11][12][13]). Some of them have concentrated their effort on the fusion problem [12] while others have addressed the representation of the terrain data within the map [5] or the use of elevation functions over a 2D domain to represent the map [3].…”

“…Some of them have concentrated their effort on the fusion problem [12] while others have addressed the representation of the terrain data within the map [5] or the use of elevation functions over a 2D domain to represent the map [3]. Other authors have tackled the filtering of the laser range finder data to improve map quality [8], the assessment of the traversability of the terrain [9], the accurate registration of the laser range finder data [10][11][12][13] or the loop-closing problem when building a global map [5]. Some of these works use small rovers in the experiments or cars moving at low speeds (<10 [m/s]) while in some cases the map is built offline [5].…”

“…Different types of representations have been used to map terrains (e.g., 2D occupancy grid, 3D occupancy voxel), being the elevation or 2.5D height map one of the preferred options in ground robot mapping applications (for example [10][11][12]) due to their compactness, lower computation and insensitivity to height discretization errors [12]. Elevation maps "store in each cell of a discrete grid the height of the surface at the corresponding place in the environment" [5].…”

“…Elevation maps "store in each cell of a discrete grid the height of the surface at the corresponding place in the environment" [5]. In some mapping systems the height's variance is also stored in each cell [10,11].…”

“…In [11], three enhancements were proposed for improving the real-time performance of the system proposed in [10]: a selection window was used to limit the probability distribution area of each measurement, a clustering algorithm was used to simplify the object detection tasks and a virtual point vector was introduced to reduce the computational cost of the algorithm. Despite these enhancements, no fundamental changes were made to the terrain estimation framework proposed in [10].…”