A study of rerouting beyond ad hoc decision making

Abstract

Rerouting is needed for a number of reasons such as overcoming a road incident, recovering from a navigation error, or avoiding a difficult turn. When obtaining a reroute, there is generally no choice but to take an inconvenient reroute - e.g., an overly long one. Although rerouting is a common operation, route guidance systems consider it as nothing more than finding an alternate route when needed. Such a view misses the opportunity to anticipate potential problems with reroutes on the way. This paper proposes a different perspective for rerouting where a path's potential reroutes are computed before navigation starts. This approach has the potential to identify desirable and adverse reroute properties and, thus, guide path planning. Such analysis requires formalising the concept of reroute. However, reroutes have previously been ambiguously or minimally defined. This research introduces formal definitions of reroutes and presents computational methods for obtaining sets of reroutes in a path. Subsequently, we assess the reroutes of shortest paths by running simulations in real-life street networks. The results show that at least 15% of shortest paths have one detour that adds more than 50% of travel time to the trip. Another result shows that in congestion, taking a reroute can save up to 9 minutes in trips that should take up to 15 minutes. The paper finishes by justifying and encouraging the use of rerouting as part of navigation query processing. Thus, a path planning method can more suitably handle road incidents, navigation errors, or dynamic navigation in general.