Dissimilar Alternative Path Search Algorithm Using a Candidate Path Set

Abstract

There is a growing need for navigation services providing multiple dissimilar alternative paths that reflect a variety of user preferences and a dynamic/stochastic variety of travel time and cost. Providing multiple dissimilar paths, in addition to the shortest path based on a certain attribute (travel cost, time or length), may contribute to extending the market of navigation services and attract new customers since navigation users may prefer other attributes, such as driving convenience and fewer right/left turns, to just saving some minutes of driving time. The traditional method for searching multiple paths is the K-shortest path search method, which provides paths in the order of cost (Yen, 1971; Shier, 1976; 1979; Martins, 1984; Azevedo et al., 1993). However, the multiple paths found by this method tend to be very similar to each other and cannot truly be called alternative paths. A reasonable alternative path should not only have acceptable attribute value(s) but should also be dissimilar to previously identified paths in terms of the links used (Park et al., 2002). To avoid unreasonable searches for the multiple paths, some methods were suggested in the previous literature. Barra et al. (1993) suggested a link penalty method. In this method, links of a path which was searched in the previous step have a penalty and the next path is searched using a shortest path search algorithm with consideration of the penalties of links. The link penalty method has no way of evaluating the quality of the set of the paths it generates in terms of the spatial differences and the costs of the paths while it can be easily used due to its simplicity (Akgun et al., 2000). Furthermore, this method cannot consider the cost constraint of the alternative paths while generalized cost is one of the most important to select alternative paths in navigation services. The link penalty method is, therefore, not suitable for the path search algorithm for navigation services. Meng et al. (2005) and Marti et al. (2009) formulated the multiple dissimilar path problems as a multi-objective problem and suggested several solution algorithms. These problems have multiple objectives varying with the purpose of researches such as minimizing average or total costs of the paths, minimizing total population exposed and maximizing the average of the dissimilarity between the paths. However, it takes significant time to solve the multi-objective problem while these problems can produce the paths that meet the given