Prediction of human driver behaviors based on an improved HMM approach

Research and development of predicting driving behaviors play an important role in the development of Advanced Driver Assistance Systems (ADAS) for assisting drivers. In this contribution, an approach is developed based on Hidden Markov Model (HMM) for predicting human driving behaviors. Three different driving maneuvers including left/right lane change and lane keeping are modeled as hidden states for the HMM. Based on observations (training), the HMM approach is able to calculate the most possible driving behaviors using observed sequences. Furthermore, the observed sequences are also used for training of HMM in the modeling process. To improve the prediction performance of the model, a prefilter is proposed to quantize the collected signals into observed sequences with specific features. In this contribution the definition of a suitable prefilter will be discussed and finally optimized. The approach focuses on the definition of optimal prefilters. Here optimality is defined as the optimal segments describing a quantized prefilter mapping the vehicle's environment to quantized states. In combination with related HMM-based results in terms of accuracy, detection, and false alarm rates an optimal parameter set of the prefilter can be determined. Using experimental data from real human driving behaviors (taken from driving simulator) it can be concluded that the optimal definition of the prefilter can increase the detection rate and accuracy, and in the meanwhile decrease the false alarm rate. The effectiveness of driving behaviors prediction has been successfully proved by comparison with other methods in this contribution.