Algorithms for obstacle detection in an off-road environment using lidar

Abstract

The functions of autonomous or semi-autonomous driving of vehicles involve precise and detailed monitoring of the environment in order to detect free and safe areas and possible obstacles. For this, it is possible to use different technologies, among which LiDAR or rotating laser provides good performance. For this technology, numerous algorithms have been developed in the past that allow the detection of obstacles with robustness in highways and urban environments. However, applications that try to detect obstacles in off-road environments are not so common. In addition, the use of low-cost sensors, forces to adopt specific solutions that mitigate the limitations that may involve low resolutions, for example. In this work, the algorithms for the detection of ditches in off-road scenarios are presented in order to warn the driver or automatically act on the vehicle using a VLP-16 3D laser scanner. These algorithms have been implemented on a military tactical vehicle with remote guided and autonomous movement capabilities and have been tested, showing their satisfactory response in road environments with other vehicles or pedestrians and in off-road environments with ditches, positive obstacles and other elements that impede circulation. In this way, the vehicle is able to navigate without a driver and detect areas that it considers to be potentially dangerous, stopping automatically and notifying an operator that the situation should be remotely supervised. The novel contribution of the paper lies in the development of a perception configuration for off-road environments, taking into account the obstacles that can be encountered, as well as the programming, implementation and testing in real conditions of specific algorithms. This configuration and algorithms must contemplate different situations, so the classical algorithms for a structured environment are not directly extrapolated. In addition, the solution has been tested on an autonomous vehicle developed by the research team.