Abstract
This paper proposes an adaptive lane departure warning algorithm based on driving habits. Since drivers differ in their driving habits and skill levels, false alarm may be triggered and causes disturbance to drivers if the same warning algorithm is used. First, the models of driving habits are established based on GM-HMM (Gaussian Mixture - Hidden Markov Model) utilizing experiments data from driving simulator. Then, an adaptive virtual adjacent lane boundary is designed according to driving habits recognized in real-time; meanwhile, the time to lane crossing (TLC) is calculated based on the adaptive virtual boundary and is compared to a certain threshold which can be set by drivers according to their driving skill levels and response time. If the TLC is shorter than the threshold and the winker is off, the system will warn the driver. The experiments show that the adaptive lane departure warning method can significantly reduce false alarms and won’t miss necessary alarm.
Keywords: lane departure warning, driving habits, TLC, GM-HMM