Integrating the Intelligent Driver Model With the Action Point Paradigm to Enhance the Performance of Autonomous Driving

Abstract

Autonomous driving is designed to enhance the overall performance of vehicular traffic. The primary objective of this study is to develop an improved car-following model for adaptive cruise control (ACC) by integrating conventional automated logic with human factors. Specifically, a modeling framework is proposed with a generalization of a classical action point paradigm describing drivers' psychological reactions and expectations. The action points are determined from collected data, from which the unconscious and conscious reaction regimes can be identified. Based on this identification, the human-like driving mode is carried out, and its corresponding acceleration model is developed to reproduce empirical car-following spiral within the unconscious regime, whereas the conventional autonomous mode driven by the Intelligent Driver Model (IDM) is used to calculate changes in acceleration for the conscious regime. To integrate the two modes, a switching rule considering the random occurrence of action points is also determined. The results of numerical experiments reveal that simulated macroscopic traffic is compatible with the three-phase theory. Moreover, traffic mobility in terms of throughput and speed is significantly improved compared to that achieved by the classical IDM. Asymptotic stability is achieved in four typical fluctuation scenarios, as the amplitudes of fluctuations converge while travelling along the vehicular platoon. However, the classical IDM fails to maintain such stability. The proposed model enables ACC to drive in accordance with drivers' psychological traits, and consequently has the potential to increase the acceptance of autonomous driving. It also has the ability to help ACC enhance traffic efficiency and safety.