Intelligent algorithm based on support vector data description for automotive collision avoidance system

Abstract

This paper presents a theoretical expansion of a new intelligent algorithm called extended support vector data description (E-SVDD) for the analysis and control of dynamic groups to realize macroscopic and microscopic behavior prediction in an automotive collision avoidance system. The time to collision concept was extracted as a key parameter via system modeling and used with the E-SVDD algorithm to set up the relevant generalized theoretical system. A new method, along with its practical application, to predict the behavior of micro- and macro-systems in real time and improve the control logic for collision avoidance was realized. A numerical simulation based on actual driving data was performed to compare the proposed collision avoidance logic and the conventional one. The results confirmed the improved performance and effectiveness of the proposed control logic.