Detection of road navigability for ATLASCAR2 using LIDAR and inclinometer data

• Autor(a) principal: Marques, Tiago Simões
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/29678
• topic: Point density; Point cloud; Curb detection; ATLASCAR2; LIDAR; Laser projection; ROS

This work is developed as part of the AtlasCar2 project, and intends to create mechanisms for the detection of area traversable by the vehicle, delimited by the side-walks or road barriers and other small obstacles, such as road delineators, or small holes on the road surface, that together condition the vehicle’s navigation. As such this work aims to obtain from the lasers a point cloud that delimits these obstacles, so that it can be used in future navigation algorithms. The approach was to first, create a reconstruction of the environment ahead of the vehicle, using for this purpose the SICK LD-MRS400001 LIDAR sensor. To achieve this reconstruction, systems were developed to establish: the global position of the vehicle, using the GPS system of the AtlasCar2; and its orientation relative to the road plane, obtained through the inclinometer module also present in the vehicle. A filtering method was developed that allows to extract, from the created road reconstruction cloud, the information regarding the obstacles to the navigation, present in the road. This filtering method is performed through a new proposed approach based on the study of the accumulated density of the laser beams, dependent on the topology of the terrain they hit. A simulator was also built to study these accumulation behaviours in a controlled environment, in order to evaluate the performance of the accumulation in various situations, including several configurations of the sensor and the road obstacles. Lastly, some results testing the individual performance of the position and orientation systems, are presented, as well as the results of their integration in the road reconstruction system. Results of the performed simulations detailing the accumulation behaviour in various situations, are also presented, along with the results of the filtering algorithm: first using static filtering parameters, and then using dynamic ones, that change based on vehicle velocity, aimed at solving some of the limitations of the static method. The methodology for delimiting the road area accessible by the vehicle, despite some limitations, shows capable of accurately identifying, in real time, the road boundaries with relatively low computational cost, generating a point cloud ready to be used in navigation algorithms.