Validating of complex agent-based models of social activities and travel behaviour

Recent travel forecasting models are based upon the fact that travel is derived from the activities in which people participate. The underlying architecture of these models are becoming more behavior- and individual-driven, and as a result, researchers are now looking towards agent-based modelling techniques for model development. This move provides new challenges with respect to validation. We describe a validation process, developed using techniques from transport modelling, agent-based modelling, and social simulation, for a complex model of joint social activity generation and scheduling. The tests focus on the activities predicted by the model, in particular when and where, and measure the extent of their similarity with theory, expert opinion and data collected in Eindhoven. We provide a recommendation of how agent-based travel behaviour models can be validated and the resources required.