"Cost switching Lange-Change Adaptive Cruise Control (LC-ACC)"
Cost switching Lange-Change Adaptive Cruise Control (LC-ACC)
Sprache des Titels:
This thesis presents a novel approach to investigate traffic information for the development of better ACC systems. The goal is to incorporate this information into the control algorithm of an ACC to decrease conservativity and increase performance in certain traffic conditions. At first, two generic performance indicators, namely fuel economy and comfort, are defined and used for traffic analysis. Data from naturalistic driving behavior are analyzed in order to find traffic situations which are more likely to offer optimization potential for one of the performance indicators. In a second step, regression analysis and machine learning are applied to describe these conditions mathematically and to find interesting clusters of traffic situations.
A two layer control structure for the high fidelity vehicle simulator IPG CarMaker is used to analyze the impact of different priorities of control objectives. The acceleration trajectories are computed by model predictive control (MPC) and translated into throttle, brake and steering inputs by lower level controllers. For collision avoidance, a stochastic traffic prediction model is introduced and trained with real world data. Simulation of the ego car in an interesting traffic condition showed that indeed, it is more sensible to concentrate on improving comfort instead of fuel efficiency. A considerable advantage in comfort was possible while the increase in fuel consumption was comparatively low. However, several difficulties for validating such ACC systems have been identified and substantiated.