Abnormal driving behavior detection based on an improved ant colony algorithm

As one of the most serious hazards in the world, more than 80% of traffic accidents are caused by driver misconduct. The detection of abnormal behavior of drivers is important to improve safety in public transportation. The anomaly measurement is not only determined by objective rules such as laws, but also distinguished due to the biological characteristics. The same driving behavior may present completely opposite judgment results for different categories of drivers. In this paper, we propose a novel detection method that measures the preference path length of drivers for various driving operations via pheromones, and identifies abnormal driving behavior by calculating the cumulative conversion probability of operation switching. An improved ant colony algorithm based on fixed point simplicial theory is proposed to improve the convergence efficiency by optimizing the initial population state. Experimental results show that the proposed method can effectively detect abnormal driving behavior and significantly reduce false alarms.