Obstacle detection and avoidance for a mobile robot

Abstract

Off-road robots are complex vehicles used in a variety of applications and are capable of operating over rough terrain, and its application has been growing more and more nowadays. With the growing importance of the study and application of autonomous vehicles in rough areas for the users’ safety reasons, researches concerning this kind of vehicle have increased. Works were already done to propose algorithms for non-linear predictive controllers, for the improvement of the stability of the vehicle, for path following and also for a nonlinear observer to estimate the contact cornering stiffness in real-time, but there is also the question of how to proceed if the robot encounters an obstacle that could obstruct its path while it is tracking a path. Thus, this work aims to propose an algorithm that allows a four-wheel, fast off-road, double-steering mobile robot to detect obstacles from the terrain in real-time using the dynamic mapping of the environment, and that also allows the robot to avoid obstacles by following a local path created using a composite Bézier curve, optimized based on the maximum steering that the robot can perform. For the experiments, a sensor for position and perception were used, including the Lidar (Light Detection And Ranging) Velodyne HDL-32E. The treatment of the point cloud provided by it was treated using mainly the PCL library. For reasons of internship duration, the tests performed were done mostly in a virtual environment considering different types of trajectory to be followed by the SPIDO, with obstacles positioned along the way. The final results obtained were satisfactory concerning the expected, thus concluding the validity of the proposed algorithm.