Applying Entropy to Understand Drivers’ Uncertainty during Car-following

As one of the main processes in most microscopic simulation models and modern traffic flow theory, carfollowing has drawn huge academic attention from the engineering and physiological domains. However, given the inherently uncertain and unpredictable nature of human behaviour, car-following models have always faced challenges in capturing drivers’ behaviour accurately and objectively. Therefore, to better capture drivers’ uncertainty in car-following, this paper contrasts four different entropy algorithms (Shannon Entropy, Steering Wheel Entropy, Approximate Entropy and Sample Entropy) as a novel measure, based on time headway data during car following. Results showed that not all the entropy measures tested are suitable for the context of carfollowing, especially when it comes to measuring uncertainty in time headway data. Approximate and Sample entropy algorithms in a moving time window seem to be the most appropriate, as they consider drivers’ prior time headway data as a factor in the perceived uncertainty. This paper contributes to the fields of microsimulation and human factors, as it demonstrates how entropy can be a precise and replicable measure of changes in behaviour, as well as anomalies in patterns of time headway data in car-following situations.