Background: This research is driven by the general need for affordable, frequent and convenient Public Transport (PT) solutions. Over the last years the advances in the sector of autonomous systems have triggered studies on their effect on PT. AB lower the operational costs due to the removal of labor costs, which in developed countries account for more than half of the overall operational costs. These lower operational costs are expected to lead to higher service frequencies. The introduction of more diverse vehicle sizes is then possible and economical which will allow the operators to target the user demand better than with a fixed sized vehicle fleet. In this we analyse the use of AB systems in existing PT networks by:

- define an AV specific objective function

- integrate AV systems in a mesoscopic simulation framework

- extract KPIs for the economic deployment of AB systems 

This study aims at answering the following research questions:

- How can AB systems be used to improve passenger and operator costs on existing lines?

- What are the implications of the cost trade off in terms of the defined KPI?

- On which lines is the deployment of AB systems most interesting in terms of social welfare?

Methods: The implementation of the framework adopts a simulation based optimization approach. The multi-agent simulation software (BusMezzo) uses the networks routes and the decision variables as input values. Subsequently the simulation is executed, and the filtered results will be handed to the Genetic Algorithm optimization. The objective function minimizes the overall cost which is the sum of operator costs (capital costs & operation cost) and user cost (travel time, waiting time, ticket cost,..). This loop is executed until convergence. The decision variables for the optimization module are the vehicle capacity, the frequency per route and the vehicle type. For validation of the optimization approach described above a brute force analysis is done on the synthetic Network provided by Spiess & Florian. In the brute force analysis the entire solution space for the network is displayed and can be analyzed. This allows for deeper understanding of the underlying processes and validates the optimization results. With the knowledge of this approach the optimization parameters are configured.

Results: The proposed model was applied to a on-going pilot case study in the area of Kista in Stockholm. The proposed model is generally applicability for larger scale problems. Possible applications of the proposed methodology are:

- tool for identifying the most promising areas for introducing AB

- measurement for the economic impact of AB Systems on PT

- the design of the network for a mixed operation

Potential extensions of the model include the fleet composition and fleet size per line of special AB zones in high user demand areas.

The main results and conclusions are:

- Frequency and Capacity have comparable impact on total cost

- Introducing Autonomous Vehicle on high demand lines is beneficial

- There are "sweet spots" for operating for operating vehicle mixes

- some configurations are only profitably operable with autonomous vehicle