A performance analysis of a congested multimodal intersection using VISSIM: A case study of Åkareplatsen, Göteborg

Abstract

Åkareplatsen is a complex intersection in the city centre of Gothenburg, Sweden’s second-largest city, that is utilised by buses, trams, cars, trucks, bicyclists and pedestrians. Since the rebuilding in 2018, Åkareplatsen has been struck by recurring congestion, gridlocks and accidents. To study this roundabout with tram-crossing traffic, a model created in VISSIM was analysed, calibrated and validated using 2024 traffic data. This study is limited to what is possible wihtin the bounds of VISSIM. The method primarily consisted of collecting traffic data in the field and official data, model calibration and validation, interviews, suggestions for improvements, and a stress test. To capture the stochastic variability of the model simulations, the sim ulation time was set to 3 hours in 15-minute intervals and was repeated in 10 runs. Queue length and travel time were the measures of effectiveness (MOEs) used to compare the model versions and reality during maximum traffic flow. The deviation between the simulation results and the field measurements was evaluated visually from video recordings and statistically using Root Mean Square Error (RMSE). The final model calibration was able to capture realistic diving behaviour and deviated by 33 minutes and 49 seconds in travel time and 104 m in queue length, an improvement compared to the original model. During validation, it was found that the calibrated model is about equal at estimating travel time and queue length during calmer and more dense traffic states. From the simulation results of suggestions for improvements, removing signal heads for tram-crossing traffic and re-routing some of the bus lines greatly reduced queues and the number of gridlocks. The stress test with a higher public density load shows that the suggested design can manage a 50 % increase in public transportation density without extensive gridlocks forming. Realistic traffic models are time-consuming to build, require a lot of data, and do not directly reflect reality. Nonetheless, they can be beneficially used to anticipate traffic-related problems, avoid potential damages they cause, and provide citizens with their need for mobility.