k-Shortest path algorithms

Abstract

This paper focuses on algorithms to solve the k-shortest path problem. Three codes are described and compared on random generated and real-world networks. One million paths were ranked in less than 3 seconds (3 microseconds per path), with at most 1 second of preprocessing, on random generated networks with 10 000 nodes. For real-world instances with more than one million nodes, the preprocessing time rises up to 2,7 hours and the CPU time to rank one million paths is less than 30 seconds (30 microseconds per path).