Modelling car drivers as believable autonomous agents for a traffic control training game

In the context of a training game for road traffic controllers, we discuss issues related to representing and implementing car drivers and their interaction with the driving environment to achieve behavior that fits the definition of believable agents. Drivers are implemented as autonomous agents with a set of individualized characteristics describing “driver personality” that affect their reaction time, tendencies for speeding, etc. Such autonomous agents interact with other cars, obstacles, traffic signals and traffic controllers - who can change traffic patterns - and serve to support simple and believable game simulation and allow instructors to create interesting and representative training units.