In order to assure safety in self-driving cars, the Autonomous Drive functionality needs to pass safety tests not only based on real scenarios collected from field driving tests, but also according to similar perturbed trajectories that might have not been collected in the data collection. To achieve this goal, we need to build a scenario database containing both real-world collected data and synthesized scenarios that are consistent with the real-world driving behaviour. This requires accurate and efficient annotation methods for extraction and analysis of driving scenarios trajectories. In this study, we propose an effective non-parametric trajectory clustering framework to annotate scenarios based on transitive relations of trajectories in an embedded space. We investigate the proposed framework's performance on real-world trajectory data sets and demonstrate its promising results, despite the complexity caused by having trajectories of varying lengths. Furthermore, we extend the framework to validate the augmentation of the real data sets with the synthetic trajectories generated by Generative Adversarial Networks (Recurrent AE-GAN) where we conclude the consistency of the generated and the real scenarios.
Vehicle Motion Trajectories Clustering via Embedding Transitive Relations
19.09.2021
4718165 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Clustering of Vehicle Trajectories
IEEE | 2010
|A Track Association Algorithm Based on Transitive Closure Fuzzy Clustering
Springer Verlag | 2021
|Review and Perspective for Distance-Based Clustering of Vehicle Trajectories
Online Contents | 2016
|