The current state of the art in scheduling ''point-to-point'' trains in a railway network utilizes the principles of centralized decision-making. A point-to-point train refers to either a freight or a passenger carrying express train wherein the originating (A) and destination (B) stations are known a priori but the route between A and B may be ascertained dynamically based on a complex cost function. In centralized scheduling such as the ATCS [1] system, a dispatcher, implemented through a single powerful mainframe computer, receives, at appropriate intervals of time, the destination and current location and speed of every train and the statuses i.e., whether occupied or empty, of all tracks in the railway system. The dispatcher analyzes the data and determines, based on a given cost function, the most efficient routes, for the subsequent time interval, for every train in the system. The; major difficulty of this approach is that the execution time and the memory requirements increase nonlinearly as the system grows in size with more trains, tracks, and stations. This paper introduces a new approach, ''DARYN'', wherein the overall decision process is analyzed and distributed onto every natural entity of the system-trains and stations. Tn DARYN, the decision process for every train is executed by an onboard processor that negotiates, dynamically and progressively, for temporary ownership of the tracks with the respective station controlling the tracks, through explicit processor to processor communication primitives. This processor then computes its own route utilizing the results of its negotiation, its knowledge of the track layout of the entire system, and its evaluation of the cost function. Every station's decision process is also executed by a dedicated processor that, in addition, maintains absolute control over a given set of tracks and participates in the negotiation with the trains. Since the computational responsibility is distributed over all of the logical entities of the system, DARYN offers the potential of superior performance over the traditional uniprocessor approach. In addition, as more trains and stations are added to the system, both the requirement for computational power and the number of computational engines increase. Assuming that a train's origin and destination is uncorrelated with that of any another train, the ratio of available computational power to the required computational power is expected to decrease only by a marginal fraction. Consequently, the time required by a train to travel from locations A to B at a given speed is expected, in general, to be marginally affected by an increase in the network size and the number of trains. This will be referred to as ''performance scalability'' in this paper. Presently, DARYN utilizes a simple cost function. However, if one chooses to increase the complexity of the cost function, DARYN's advantage over the traditional approach increases due to its enormous available computational power. Given that the current microprocessors such as MC68030, MC88000, Intel 486, and Intel 860 are powerful get relatively inexpensive, a network of concurrently executing processors may offer superior price-performance quotient over a single high performance computer. Unlike in the traditional approach where track statuses and train positions must be relayed over long distances to the central computer thereby increasing communication costs, in DARYN, communications between trains and stations are local and, therefore, fast and relatively inexpensive. This paper also reports on the development of a realistic model of a railway network based on the DARYN approach and its implementation on a loosely-coupled parallel processor system. Experimental results indicate DARYN's feasibility, significant superiority over the traditional approach, and that it exhibits, in general, the notion of ''performance scalability.'' The principles of the DARYN approach may find applications in other transportation systems including trucks and cargo-airline networks.
