Bezier curve based lattice planner and lateral controller for highway scenarios
Sammanfattning: The autonomous driving technology is transforming the road-based transport industry into a new era that will change the way we think about transport solutions. With advancements in computational power and sensor precision, self-driving vehicles are predicted to be a reality in the near future. This thesis proposes a spatiotemporal motion planner for highway scenarios. The planner is designed to consider static and dynamic obstacles to find a suboptimal and collision-free trajectory. The driveable region is sampled using a state-lattice to reduce oversampling and generates paths from different orders of Bezier curves. The resulting planner successfully performs complex maneuvers such as overtaking and merging into heavy trafficked roads. A lateral PIDcontroller is designed to compare the different orders of Bezier curves curvature continuity. The experiments show that the lateral controller performance is directly related to paths with low curvature. In the highway scenario studies, it is observed that Bezier curves of lower orders and discontinuous curvature performs better than higher-order curvature continuous Bezier curves.
HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)