Robust unsupervised learning of temporal dynamic vehicle-to-vehicle interactions. (September 2022)

• Record Type: Journal Article
• Authors: Guha, Aritra; Lei, Rayleigh; Zhu, Jiacheng; Nguyen, XuanLong; Zhao, Ding
• Is Part Of: Transportation research. Volume 142(2022)
• Publication Date: 2022-09
• Subjects: Unsupervised learning -- Temporal -- Robust; Transportation -- Periodicals; Transportation -- Technological innovations -- Periodicals; 388.011
• DOI: 10.1016/j.trc.2022.103768
• Languages: English
• ISSNs: 0968-090X

Abstract

Abstract: Robust unsupervised learning of temporal dynamic interactions is an important problem in robotic learning in general and automated unsupervised learning in particular. Temporal dynamic interactions can be described by (multiple) geometric trajectories in a suitable space over which unsupervised learning techniques may be applied to extract useful features from raw and high-dimensional data measurements. Taking a geometric approach to robust unsupervised learning for temporal dynamic interactions, it is necessary to develop suitable metrics and a systematic methodology for comparison and for assessing the stability of an unsupervised learning method concerning its tuning parameters. Such metrics must account for (geometric) constraints in the physical world as well as uncertainties associated with the learned patterns. In this paper, we introduce a model-free metric based on the Procrustes distance for robust learning of dynamic vehicle interactions, and an optimal transport-based distance metric for comparing between distributions of interaction patterns. These distance metrics can serve as an objective for assessing the stability of an interaction learning algorithm. They are also used for comparing the outcomes produced by different algorithms. Moreover, they may also be adopted as an objective function to obtain clusters and representative interaction primitives. These concepts and techniques will be introduced, along with mathematical properties, while their usefulness will be demonstrated in unsupervised learning of vehicle-to-vehicle interactions extracted from the Safety Pilot database, one of the world's largest databases for connected vehicles. Results reveal that the proposed methods outperform existing techniques under several well-founded notions of clustering efficiencies. Highlights: Robust representation learning for temporal dynamic interactions. Distributional metric for assessing the stability of an unsupervised learning method relative to tuning parameters. Simulations of traffic interaction scenarios. Interpretable inference for clustering V2V encounters.