In contrast to conventional public transport, demand responsive transport (DRT) generally does not require a timetable, a fixed sequence of stops or predefined stops. Demand responsive transport includes ride sharing, student transport, taxi on call services and carpool services.
Only in combination with the conventional public transport, ride sharing systems lead to a sustainable offer in urban transport (Ride sharing).
In order to be able to evaluate the feeder function of these offers, an assignment procedure must combine the modes of public transport and ride sharing. Within an assignment, the transfer processes are depicted and a temporal and spatial consistency of the connections is ensured. The result of the timetable-based assignment with consideration of ride sharing systems are intermodal connections including their volumes.
While the timetable-based assignment in conventional public transport follows macroscopic principles, ride sharing systems can only be modeled sufficiently by microscopic simulations. In order to be able to achieve stable results for such a system, a large number of realizations, i.e. in this case served trip requests and characteristic values must be summarized. In addition, the the key values of ride sharing systems are volume/capacity dependent.
Carrying out several microscopic tour planning procedures in each iteration as part of a capacity-dependent assignment would unreasonably prolong the runtime of the procedure. Therefore, the average of the realizations is not obtained from several simulations, but from spatial and temporal aggregation. For this purpose, nodes are combined in larger areas.
The aggregates of the characteristic values between these areas and within the individual areas are used for the route search and choice within the assignment. The choice of the area division plays a central role. Large areas coarsen the result, but lead to stable skims. As a rule, small areas mean few realizations (trip requests) per relation. This makes the influence of stochastic fluctuations too great to achieve convergence and thus relevance.
The determined characteristic values serve the evaluation of potential public transport path legs, which are taken on the network of the assigned PrT transport system. When such a DRT path leg is chosen, a trip request for tour planning arises for the next iteration loop, which in turn will generate new values. The individual tour plans are thus fed by trip requests that end or begin at transfer stops. However, unimodal paths are also possible.
The tour plans from the individual iterations of the procedure can be saved as files. Due to the aggregation of the results from tour planning in the procedure, no tour plan is consistent with the resulting paths of the assignment. However, an analysis of the tour plans can provide useful results for an operational consideration of the ride sharing system.
