RYNSORD: a novel decentralized algorithm for railway networks with "soft reservation"

Abstract

This paper presents a novel decentralized algorithm with soft reservation for efficient scheduling and congestion mitigation in railway networks (RYNSORD), wherein every train utilizes lookahead to dynamically replan its route. Thus, a train, currently at position N and headed for the ultimate destination Y, first requests and reserves N tracks, i.e., the lookahead, ahead of its current position for use at appropriate future times. The train moves through the N tracks and, upon completion, it again requests and reserves N subsequent tracks ahead of itself. The process continues until the train reaches its destination. The reservation process is characterized as "soft", i.e., less abrupt and more flexible in negotiation between the trains and stations, in contrast to the traditional rigid "hard" reservations, RYNSORD is modeled and implemented on a network of SUN sparc workstations, configured as a loosely coupled parallel processor. Experimental results, obtained from simulating realistic example train networks and utilizing stochastic input train traffic data, reveal that: (1) soft reservation is superior to hard reservations in achieving both fast train travel times and higher quality of routing; and (2) while trains under low lookahead experience higher average hop count and high probability of doublebacks, relative to large lookahead values, they also have lower travel times.