Modeling multiple vehicle interaction constraints for behavior prediction of vehicles on highways. (March 2022)

• Record Type: Journal Article
• Authors: Tripicchio, Paolo; D'Avella, Salvatore
• Is Part Of: Computers & electrical engineering. Volume 98(2022)
• Publication Date: 2022-03
• Subjects: Autonomous driving -- Recurrent deep neural networks -- Support vector machine -- Behavior prediction; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.107700
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: In the context of autonomous driving and road situation awareness, this manuscript introduces a Bayesian network that enables the prediction of participant vehicles (PVs) circulating on a highway. The network architecture combines Long-Short-Term-Memory recurrent Deep networks and Support Vector Machines as computational nodes in a graph. The network inputs are real data acquired by radar and synthetic data generated mimicking the real ones. The interaction among multiple vehicles is handled explicitly by introducing the Allowance Factors that model the constraints of the possible interactions of the PVs and the ego car in the trajectory forecast estimation. Results obtained on the conducted tests show the ability of the system to predict the motion of the vehicles up to 6 s in advance with a 92% mean average accuracy. Additionally, if implemented into automatic trajectory planners, the introduced modeling approach enables the prediction and avoidance of possible vehicles' collisions. Graphical abstract: Highlights: Fusion of DL prediction and ML classification in a Bayesian graph. Definition of reference trajectories for PVs behavior classification and prediction. Integration of multi-vehicle constraints and decision factors to model feasibility.