A Stochastic Approach for Modeling Lane-Change Trajectories - Digital Signal Processing for In-Vehicle Systems and Safety - page 264

Fig 19.7 Examples of generated (dotted line) and the selected (dashed line) trajectories for
maximum likelihood (ML) signal synthesis algorithm. The actual trajectory observed under the
given condition is also plotted (solid line)
Fig 19.8 Examples of generated ( dotted line ) and the selected ( dashed line )trajectories for sampling
method. The actual trajectory observed under the given condition is also plotted ( solid line )
change behavior can be modeled with HMM parameters. The trained hazard maps
M' for the two drivers shown in Fig. 19.6 also depict differences in sensitivity to
surrounding vehicles.
We generated possible lane-change vehicle trajectories over a 20-s period using
two methods: a maximum likelihood (ML) method [ 13 ] and a sampling method.
The possible generated trajectories and the selected optimal trajectory using the ML
method are shown in Fig. 19.7 , and those generated using the sampling method are
shown in Fig. 19.8 . The vehicles traveled about 600 m while changing lanes. The
trajectories of the two drivers are clearly different.