Class ContractionHierarchyPrecomputation<V,E>

java.lang.Object
org.jgrapht.alg.shortestpath.ContractionHierarchyPrecomputation<V,E>
Type Parameters:
V - the graph vertex type
E - the graph edge type
public class ContractionHierarchyPrecomputation<V,E> extends Object
Parallel implementation of the contraction hierarchy route planning precomputation technique.

The original algorithm is described the article: Robert Geisberger, Peter Sanders, Dominik Schultes, and Daniel Delling. 2008. Contraction hierarchies: faster and simpler hierarchical routing in road networks. In Proceedings of the 7th international conference on Experimental algorithms (WEA'08), Catherine C. McGeoch (Ed.). Springer-Verlag, Berlin, Heidelberg, 319-333.

Parallel version of the algorithm is described in the article: Vetter, Christian. "Parallel Time-Dependent Contraction Hierarchies." (2009).

This algorithm speeds up shortest paths computation by contracting graph vertices. To contract a vertex means to remove it from the graph in such a way that shortest paths in the remaining overlay graph are preserved. This property is achieved by replacing paths of the form $\langle u, v, w\rangle$ by a shortcut edge $(u, w)$. Note that the shortcut $(u, w)$ is only required if $\langle u, v, w\rangle$ is the only shortest path from $u$ to $w$.

Contraction is performed as follows. First a priority is computed for each vertex in the graph. This implementation uses edge quotient, complexity quotient and hierarchical depth metrics for computing priority. A hierarchy is then generated by iteratively contracting independent sets of vertices. A vertex is independent iff it`s priority is less than the priority of every vertex in its 2-neighbourhood. A 2-neighbourhood of a vertex $v$ is defined as a set of vertices that are reachable from $v$ using at most 2 hops. After contraction each vertex gets its unique contraction level - its position in the computed hierarchy. Finally, after all vertices are contracted each edge is set to be either upward if its source has lower level that its target, or downward if vice versa.

Computing initial priorities, independent sets and shortcuts, updating neighbours priorities and marking upward edges is performed in parallel what gives this implementation performance speedup comparing to the sequential approach.

For parallelization, this implementation relies on the ThreadPoolExecutor which is supplied to this algorithm from outside.