Vehicular traffic flow is a complex interaction between drivers, vehicles and the road environment, involving for example, individual driver decisions and actions or the relationship between the driver-vehicle systems and the environment. TRL’s car simulator offers an ideal application for the use of intelligent driver agents to understand such complex interactions. This paper introduces modelling concepts and techniques for improving behavioural intelligence and realism in driving simulation scenarios. A Neural Driver Agent (NDA) is one type of modelling technique which is able to show behavioural intelligence. Behavioural intelligence can be defined within the broader context of Artificial Intelligence (AI) such that, at the minimum level, the system can exhibit human-like properties, for example, planning, perception, learning, knowledge, reasoning and decision making. The work described in this paper makes a number of contributions: (1) increased ‘intelligence’ of agent interactions in driving scenarios; (2) improvement in simulator trials by creating autonomous agents capable of responding realistically both to the behavioural responses of the participant and to any pre-programmed autonomous vehicle behaviour (e.g. a vehicle programmed to disobey a red traffic light); (3) improved participants’ immersion in simulator scenarios, increasing the likelihood that they will drive in a realistic and representative manner with the consequence that greater confidence can be placed in resulting analyses. This paper describes research which extended previous work (development of a Synthetic Driving SIMulation, SD-SIM framework) conducted at Loughborough University. NDAs were developed to learn and successfully replicate human lane changing behaviour based on data collected from the TRL car simulator.

Want to know more about this project?