Modeling Techniques for Driver's Lane Changing Behaviour

Advances in sensor technology and multi-core Graphical Processing Unit (GPU) is going to drive the future of self-driving vehicle. To mark a commercial success, it will require another 10-15years for such vehicles to come on the road for different safety reasons. Until we see the driverless vehicles, various research groups are active to improving current state-of-the-art of Advance Driver Assistance System (ADAS). The use of ADAS is not new and had benefitted drivers through cruise control, collision avoidance, parking assistance, lane departure warning, blind spot detection. In past few years, there are several research attempts are made to monitor the driver's behaviour including braking, steering wheel manoeuvre, drowsiness, on the road surface and environment conditions, vehicle health etc. All these research studies are influenced by a goal of minimizing accidents on the road. In this paper, we study lane-changing behavior using Hidden Markov Model (HMM) and Dempster-Shafer-Theory (DST) for steering angle and brake. The effectiveness of the proposed system is presented by conducting simulation experiments. We found the accuracy of the system is remarkably increased with combined approach of HMM & DST.