A Software Agent for Vehicle Driver Modeling

Abstract

The world is experiencing a paradigm shift towards intelligent agents in form of machine learning for modeling any given task or process. Human vehicle drivers are agents that operate under stochastic environments, full of other agents. Such environments are complex to perceive and model. This study explores how a utility-based agent could be used to model human vehicle drivers. The motivation behind the study was established on the assumption that a driver agent founded on GPS data, Mixture Models and probabilistic reasoning methodologies could effectively model human vehicle drivers. The data collected by GPS receivers was appropriately analysed to establish a driver behaviour dataset. The dataset was then divided into three sets: training, test and validation sets that were used to formulate the driver agent. The agent's successive actions were evaluated against sets of performance metrics to determine accuracy, precision and recall levels. The evaluation yielded over 50% successful performance rates at all levels. The significance of the study is four-fold: First, the function of the system could be extended to providing advisory services to drivers in real-time. Second, data gathered from the system could be used by road safety stakeholders to vet drivers and to diagnose causes of road accidents. Thirdly, the resulting knowledge-base could establish standards of rationality in driving and/or formulate rules for use in driverless vehicle control systems. Finally, the model could be used to build a dataset on driver behaviour for any given vehicle driver and type and nature of operational environment.