The Tour planning procedure links the demand, i.e. the generated trip requests, with the supply, i.e. the vehicle fleet and the pickup and dropoff points.
The dispatcher does not know the trip requests until the moment they arise and tries to serve them with the existing vehicles. The selection of the vehicles depends on the spatial proximity, the boundary conditions of the passengers already in the vehicles and the detours to be expected.
The pick-up and drop-off points of the trip requests are determined by the dispatcher. The candidates for these locations are defined by the set maximum number and the maximum PuT Walk access time from the start or end point to the pick-up or drop-off point. Increasing these parameters increases flexibility in dispatching. The selection of suitable drop-off and pick-up point pairs is controlled by a cost function. The coefficients of this cost function weight the requirements for efficient operation of the fleet ( = shorter additional travel time) and service comfort for passengers ( = shorter access and egress routes).
The dispatcher is based on the routing-enabled network of the transport model, i.e. on the opened links and turns of its PrT reference transport system.
Vehicles that do not receive new orders after submitting a trip request return to holding areas after an adjustable time. During the relocation and at the holding area, the vehicle is available to serve trip requests. The selection of the holding area influences the future availability of the vehicle due to its location.
The choice of the holding area is determined, among other things, by the distance to the respective location of the vehicle. However, it is a cost function that also takes into account the component of the potential of the holding area. The potential is determined by the occupancy of the holding area, the vehicles in its vicinity, and the attraction. This attraction can either be statically predetermined by the weight of the holding areas or is derived by proportionately accounting for future trip requests from the area surrounding the holding areas. The reason for considering future trip requests is that providers have a demand forecast that will direct vehicles to subsequent peak demand points. The quality of the demand forecast can be simulated by the proportion of future trip requests taken into account.
Tour planning results
The tour plan calculated is aggregated in various ways and adopted into the transport model.
The paths of the passengers are saved as PuT paths. Each path contains three path legs: one arrival path, one departure path and the path leg calculated that was covered using the responsive transport system. This path leg represents both the possible detour as well as the wait time for dropping off or picking up other passengers. From the total of these paths, skims are derived, analogous to conventional PuT.
For adjacent zones, the random distribution of trip requests at the node level can result in trip requests receiving the same pickup as well as dropoff point. These special cases are then represented by a public transport path composed only of walk links. The passengers with this trip request reach their destination this way. The tour planning of the fleet, however, remains unaffected by such pure walk links.
Due to the stochastic disaggregation of the trip requests, skims should only be used after multiple calculations of tour planning, taking into account various seed random numbers for trip request generation.
In addition, the trips of each vehicle are saved to a specified path set. They are used as a basis for estimating the operating expenses. PrT paths are created gradually by servicing pickup and dropoff points. The paths between these nodes are determined via a shortest path search. The resulting volume can be saved as a link attribute.
The individual pickup or dropoff points of a tour can be linked to a zone, just as for trip request generation via shortest path search. This allows you to create a PrT demand matrix that consists of individual legs of the tour. The subsequent assignment of such a matrix to an existing network creates a more realistic picture of the network volume generated through demand responsive transport.
