David Leffler

Simulation and Modelling of Autonomous Road Transport

Supervisors: Erik Jenelius, Wilco Burghout, Oded Ctas

Background and objective:

”There is great interest in the automation of the road traffic system. The vehicle industry, road administrators and other authorities have high expectations on road transport automation and hope that it will contribute to a safer traffic system that is also more efficient and environmentally friendly. Today’s car-based transport system is widely viewed as unsustainable. In company with the ongoing increase in availability and usage of shared mobility options (e.g., bikesharing, carsharing, ridesharing and ridesourcing), shared autonomous vehicles have also shown potential in reducing car ownership. Furthermore, in contrast to more traditional schedule and line-based public transit, fleets of shared autonomous vehicles may allow for a public transit service that is more flexible and demand-responsive. The development of autonomous vehicles is progressing at a high pace, but there is still limited knowledge of what effects these have, and will have, on the traffic system.

Traffic simulation is a powerful and common tool used to both study the effects on the traffic system resulting from changes in infrastructure and traffic control, as well as in decision support systems for traffic planning or advanced driver support systems. However, to enable analysis of the traffic system effects of automated vehicles, todays traffic simulation models need to be enhanced and extended with more demand-responsive capabilities. In the era of Intelligent Transportation Systems, traffic planners have access to, and utilize, greater quantities and categories of real-time data. To analyze the operation of future traffic systems, simulation models must also take into account varying degrees of available real-time data and how they will be used.

The scope of this research is to develop models to assess the effects of partially- and fully-autonomous fleets of vehicles on traffic networks using agent-based simulation. By effects we refer to both cost/benefits (e.g., travel time, travel cost) for various stakeholders (e.g., operators, passengers), as well as network effects (e.g., environmental, congestion). To this aim, Mezzo (an open source, mesoscopic, agent-based traffic simulator developed at KTH) will be extended to model demand-responsive public transit solutions and strategies.

In relation to the ongoing project ADAPT-IT (Analysis and Development of Attractive Public Transport through Information Technology) the first steps of the extension of Mezzo will be to implement more demand-responsive control strategies (such as short-turning and vehicle insertion).

Potential lines of research in no specific order (SAV = Shared Autonomous Vehicles):
1. Using real-time information for SAV service, real-time control of SAV service
1a. Holding/Waiting strategies for SAV (predict future demand to decide when to hold)
2. Operator game, competing SAV services or with public transit
3. Objective diving heuristic on top of algorithm that solves static DARP
4. Degree of dynamism (known demand before assignment) effect on SAV performance
5. Compare SAV system vs SV system”


University: KTH Royal Institute of Technology