In this work different approaches to simulation of vehicle longitudinal movement in car-following regime have been tested in order to develop an Adaptive Cruise Control (ACC) system with human-like driving capabilities. The main idea is to develop an ACC system that can be continuously trained by drivers to accommodate their actual driving preferences as these change among drivers and over time. In particular, this work is a preliminary study on a learning machine capable of learning and memorizing the driving attitude of a given driver and reproducing it automatically. The simulation of driving behaviour and the identification of a driving desired trajectory have been the subject of much research work, which led to the formulation of several longitudinal movement models, known as “car-following models”. The Gipps car-following model has been used in this work to test and compare the performances of different approaches, based on simple regression models (linear and polynomial) and complex regression models (artificial neural networks). The comparisons among different models will be based on accurate experimental data consisting in long trajectories of vehicle platoons gathered on urban roads. Keywords:
Linear, polynomial and neural network approach to longitudinal driving behaviour simulation in ACC context
Simonelli F;
2008-01-01
Abstract
In this work different approaches to simulation of vehicle longitudinal movement in car-following regime have been tested in order to develop an Adaptive Cruise Control (ACC) system with human-like driving capabilities. The main idea is to develop an ACC system that can be continuously trained by drivers to accommodate their actual driving preferences as these change among drivers and over time. In particular, this work is a preliminary study on a learning machine capable of learning and memorizing the driving attitude of a given driver and reproducing it automatically. The simulation of driving behaviour and the identification of a driving desired trajectory have been the subject of much research work, which led to the formulation of several longitudinal movement models, known as “car-following models”. The Gipps car-following model has been used in this work to test and compare the performances of different approaches, based on simple regression models (linear and polynomial) and complex regression models (artificial neural networks). The comparisons among different models will be based on accurate experimental data consisting in long trajectories of vehicle platoons gathered on urban roads. Keywords:I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
