Outdoor experiment: Achieving full autonomy in a mobile robot requires combining robust environment perception with on-board sensors, efficient environment mapping, and real-time motion planning. All these tasks become more challenging when we consider a natural, outdoor environment and a robot that has many degrees of freedom (d.o.f.). In this paper we address the issues of motion planning in a legged robot walking over a rough terrain, using only its on-board sensors to gather the necessary environment model. The proposed solution takes the limited perceptual capabilities of the robot into account. A~multi-sensor system is considered for environment perception. The key idea of the motion planner is to use the dual representation concept of the map: (i) a higher-level planner applies the A* algorithm for coarse path planning on a low-resolution elevation grid, and (ii) a lower-level planner applies the guided-RRT (Rapidly-exploring Random Tree) algorithm to find a sequence of feasible motions on a more precise but smaller map. This paper contributes a new method that can learn the terrain traversability cost function to the benefit of the A* algorithm. A probabilistic regression technique is applied for the traversability assessment with the typical RRT-based motion planner used to explore the space of traversability values. The efficiency of our motion planning approach is demonstrated in simulations that provide ground truth data unavailable in field tests. However, the simulation-verified approach is then thoroughly tested under real-world conditions in experiments with two six-legged walking robots having different perception systems.

 

References:

  1. D. Belter, P. Łabęcki, P. Skrzypczyński, Adaptive Motion Planning for Autonomous Rough Terrain Traversal with a Walking Robot, Journal of Field Robotics, vol. 33(3), pp. 337-370, 2016 (pdf)