Piecewise linear continuous-curvature path planning for autonomous vehicles

• Autor(a) principal: Silva, Júnior Anderson Rodrigues da
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: http://www.teses.usp.br/teses/disponiveis/18/18153/tde-02032018-095552/
• topic: Autonomous vehicles; Clothoids; Clotóides; Mapeamento; Mapping; Path planning; Planejamento de caminho; Veículos autônomos

Autonomous vehicles have increasingly become an attractive field due its promising capabilities of improvements regarding safety, comfort, traffic flow etc. A required attribute for those vehicles is the ability of autonomously compute its path towards a destination point. The path must be planned considering the constructive aspects of the vehicle in order to guarantee the maneuver feasibility. This work consists on computing a feasible path for autonomous vehicles with non-holonomic constraints. Piecewise linear continuouscurvature paths constituted of clothoids, circular arcs, and straight lines are used for this purpose, providing passenger\'s comfort. The road network is modeled from GPS (Global Positioning System) vehicle trajectories by defining lanes, roundabouts and intersections. GPS points are used later to parameterize lanes using clothoids and to extract roundabout centers and radii. This approach provides a sparse road network model since GPS points are replaced by parameterized curves. The information about connections between roads coming from the model is used by a global path planner, which computes a minimal length route from the vehicle current position to the destination point. After that, path planners compute intersection and roundabout paths depending on the nature of connections between roads. Also, lanes changes are performed to obey traffic rules. These three path planners that connects adjacent roads use clothoids, circular arc, and straight lines as interpolating curves whose curvature is constrained to that the vehicle can perform: the intersection path planner uses only a minimal amount of steering to perform the maneuver, increasing the comfort level; the roundabout path planner takes the roundabout center and radius as well as parameters that defines the entrance and exit maneuvers to compute the path; the lane change path planner connects lanes belonging to the same road with a prescribed longitudinal traveled distance depending on whether this maneuver is required. In the end, an global continuous-curvature path is generated. As the result of this work, a real urban scenario is modeled and the proposed approaches are validated.