Lightweight accident detection model for autonomous fleets based on GPS data
Αλέξανδρος Παπαδόπουλος, Athanasios Sersemis, Georgios Spanos, Antonios Lalas, Christos Liaskos, Konstantinos Votis, Dimitrios Tzovaras
Abstract
Autonomous Vehicles (AVs) will be the future of automotive including both the Public and the Private Transportation. One of the major concerns of the corresponding research community is the safety of the AVs. Considering this, a lightweight accident detection model for autonomous fleets is presented, utilizing only GPS data. The proposed accident detection model combines well-known statistical and machine learning techniques such as data normalization, PCA transformation, and DBSCAN clustering. In order to validate the proposed methodology simulated data were utilized exploiting well-established techniques, such as Dead-Reckoning, accident speed profiles, and pre-crash acceleration models. The preliminary results highlighted that the proposed methodology managed to achieve its accurate accident detection purpose presenting accuracy higher than 98%.