Traffic Simulation of Automated Shuttles in Linköping University Campus
2Automated shuttles (ASs) are designed to provide a cleaner transportation and improve access to areas such as where travelers have to walk long distances to/from the bus stops. The introduction of ASs in the road network might affect the safety of other road users specially pedestrians and cyclists as well as the performance of conventional cars due to the lower speed of ASs. Several demonstration trials are being conducted to study how ASs operate in a real traffic condition, but they are limited to few vehicles and evaluations of traffic effects at higher penetration rates are not possible.
Traffic simulation is a tool that can be used to study effects on traffic performances at different penetration rates. However, ASs have not been modeled, calibrated, and validated in microscopic traffic simulation tools. Therefore, the objective of this paper is to model, calibrate and validate automated shuttle’s behavior using the simulation tool SUMO and data collected from the demonstration trial on the area of campus Valla Linköping University, Sweden. The ASs operate on a 2.1 km fixed route.
The collected data is analyzed to study the behavior of the AS along the shuttle route. A free driving behavior of the AS when it is unconstrained by other road users is obtained. The analysis result shows that the AS has different maximum operation speeds at different locations and defining one value for the maximum speed when setting up the simulation in SUMO is not enough. Therefore, virtual speed limit is derived by mimicking the maximum operation speed of the AS from the data. In the simulation network, edge speed limit is defined based on the virtual speed. Additionally, acceleration and deceleration parameter values are extracted from the data and used as input to the simulation. Krauss and IDM car-following models have been investigated by tuning the acceleration and deceleration parameters for calibration of the free driving situation. The preliminary results indicate that the IDM car-following model shows a slightly better agreement with actual observation.