RailCab effects individual traffic on rails by means of small, fully automated vehicles that are employed on demand for an uninterrupted journey. Except for local traffic, operation will not follow any fixed schedule. Booking systems will directly compute journeys from customer requests. To comply with any request, a vehicle can be provided that is suitable for the respective task. From the customer’s point of view this means that he can book a vehicle for his desired departure time and will be taken automatically from his starting point to his destination.

In order to meet customers’ requests as quickly as possible, we pursue a decentralized policy for the RailCabs. Its basis is a multi-agent system in which one intelligent agent is assigned to every relevant object in the transport network. These agents negotiate the terms of rivalling transport requests and submit to the customers several alternative proposals to choose from. Linking the multi-agent system with the MFERT modelling procedure can ensure that limited resources like rails, switches, or stations will not be overcharged.

When a quick fix for customers’ requests has been found the second stage comprises optimisation of the initial plan. In a strongly parallelised routine, the local plans of the agents are consolidated and optimised, thus reducing costs by as much as 30 %. Initial planning and optimisation in the multi-agent system are continuously effected in turns so that incoming transport orders can be processed without delay.