Performance comparison of lane-changing models for merging scenarios in traffic simulation for driving simulators

Virtual intelligent traffic is essential for an immersive experience in a driving simulator, and crucial for obtaining reliable results in complex scenarios. However, current research has not focused on actors with humanlike behaviour in driver behaviour models. The current paper presents results from a series of game theory based lane-changing models specifically designed for motorway merging situations. Different models were implemented in SmartActors, a newly developed microscopic traffic simulation platform and were tested in two different scenarios, one of which was a simulated road layout of the I-80 motorway, for which real traffic data is available from the NGSIM project. The results from the base merge scenario showed that trade-offs occurred between traffic and safety performance of the models as models that let to higher traffic throughput also let to higher number of crashes and unsafe merge behaviour. Results from the I-80 scenario showed that the road layout may affect the performance of the models, when compared to a base scenario with a shorter acceleration lane. As a conclusion, game theory models can be used for microsimulation applications, although in some cases additional restrictions to their original specification may need to be implemented to ensure safe performance.