Abstract:The paper describes the case study of the mobile robot Andabata navigating on natural terrain at low speeds with fuzzy elevation maps (FEMs). To this end, leveled three-dimensional (3D) point clouds of the surroundings are obtained by synchronizing ranges obtained from a 360 • field of view 3D laser scanner with odometric and inertial measurements of the vehicle. Then, filtered point clouds are employed to produce FEMs and their corresponding fuzzy reliability masks (FRMs). Finally, each local FEM and its FRM are processed to choose the best motion direction to reach distant goal points through traversable areas. All these tasks have been implemented using ROS (Robot Operating System) nodes distributed among the cores of the onboard processor. Experimental results of Andabata during non-stop navigation on an urban park are presented.

Reactivity is a key component for autonomous vehicles navigating on natural terrains in order to safely avoid unknown obstacles. To this end, it is necessary to continuously assess traversability by processing on-board sensor data. This paper describes the case study of mobile robot Andabata that classifies traversable points from 3D laser scans acquired in motion of its vicinity to build 2D local traversability maps. Realistic robotic simulations with Gazebo were employed to appropriately adjust reactive behaviors. As a result, successful navigation tests with Andabata using the robot operating system (ROS) were performed on natural environments at low speeds.

Autonomous navigation of ground vehicles on natural environments requires looking for traversable terrain continuously. This paper develops traversability classifiers for the three-dimensional (3D) point clouds acquired by the mobile robot Andabata on non-slippery solid ground. To this end, different supervised learning techniques from the Python library Scikit-learn are employed. Training and validation are performed with synthetic 3D laser scans that were labelled point by point automatically with the robotic simulator Gazebo. Good prediction results are obtained for most of the developed classifiers, which have also been tested successfully on real 3D laser scans acquired by Andabata in motion.

Progress in applying supervised learning for natural scene classification is impeded by the lack of appropriate datasets for training. This paper describes the automatic generation of synthetic three-dimensional (3D) scans of natural environments with each point labelled individually with its element class. The developed software employs the robotic simulator Gazebo to obtain range and intensity measurements from a 3D laser rangefinder aboard a ground mobile robot. Precisely, the returned intensity values are used to annotate every 3D point within its corresponding class 100% error free. Several examples are provided to show the utility of the proposed approach.

No abstract