Abstract:

In order to improve the calibration accuracy of the traditional car-following model based on fuzzy reasoning, this paper proposes a new model based on the TSK model. In this model, an improved genetic algorithm-based learning algorithm is adopted for the identification of TSK model, and the situation with insufficient sub-group data point is specially considered. Moreover, some real car-following data are collected by GPS for the establishment and validation of the TSK model. It is found that the experimental results accord well with the theoretical ones, and that the accuracy increases by one order of magnitude, meaning that the application of TSK model to the car-following model is feasible.

Key words: car following, TSK model, genetic algorithm