TitleGenetic Algorithm in Trajectory Optimization for Car Races |
|||
AuthorsAssoc. Prof. Dr. Dana Vrajitoru, Indiana University South Bend, United States of America |
|||
AbstractIn this paper, we present several methods of constructing trajectories for autonomous cars in a car race setting using the TORCS car race simulation system with the goal of improving race completion time. The first method builds the trajectory procedurally using known efficient curves. It starts by mapping the race track using available sensors provided by TORCS. The track is then reconstructed and segmented based on the curvature direction. Last, efficient curve profiles are applied to each segment. The second method applies smoothing and gradient descent optimization algorithms to improve such trajectories built by the first method. Finally, we use a genetic algorithm to find an optimal trajectory by minimizing the overall length of the curve. A coarser segmentation is used in this case and a number of key points are selected for the trajectory definition. We compare the results of these different methods on two chosen race tracks and show that the genetic algorithm is capable of building a trajectory of a lower total length than other methods, but the results in an actual race are less predictable. |
|||
Keywordsgenetic algorithms, autonomous cars, trajectory optimization |
|||
CitationVrajitoru, D. (2019). Genetic Algorithm in Trajectory Optimization for Car Races. In M. Shelley & V. Akerson (Eds.), Proceedings of IConEST 2019--International Conference on Engineering, Science and Technology (pp. 1-10). Monument, CO, USA: ISTES Organization. Retrieved 21 November 2024 from 2019.iconest.net/proceedings/30/. |
|||
LinksDownload Fulltext |
Dear Participant, We will use the Adobe Connect Meeting platform for the virtual presentations. You will receive an email titled “Adobe Connect - Your Account Information” from Adobe for the Account URL link and password to participate in the vi...
14.09.2019
