Báo Cáo An improvement of the driver trajectory model for path-planning

An improvement of the driver trajectory model for path-planning

The paper presents some new ideas to improve the accuracy and the generality of the driver
trajectory model used in the path-planning module of the NAICC (Navigation Aided Intelligent Cruise
Control) system. First, a new Sugeno fuzzy model is developed to calculate the characteristic points,
which are dependent on the driver and the road profile, needed to compute the coefficients of a polar
polynomial modeling the driver trajectory. The parameters of the membership functions used to
represent linguistic variables in Sugeno fuzzy model are estimated by using a Levenberg – Marquardt
algorithm so that the mean square error between the observed data and the fuzzy model output is
minimized. Second, the effect of the slope and curvature constraints on the shape of the polar
polynomial and the driver trajectory are discussed and some nice expressions for estimating these
constraints are proposed. Simulation on a wide range of bends with different radii and turn angles
shows on the one hand that the obtained fuzzy model is able to calculate the appropriate characteristic
points corresponding to the driver profile, and on the other hand that the estimation of the slope and
curvature conditions is good enough. An experimental validation with some available test bends shows
that the path, planned by using the new trajectory model, fits very well to the real driver trajectory.