Dijkstra ’ s Shortest Path Algorithm

The underlying principle of the algorithm may be described as follows: the algorithm starts with the source; it visits the vertices in order of increasing cost, and maintains a set V of visited vertices (denoted by UsedVx in the article) whose cost from the source has been computed, and a tentative cost D(u) to each unvisited vertex u. In the article, the set of all unvisited vertices is denoted by UnusedVx. D(u) is the cost of the shortest path from the source to u in the subgraph induced by V ∪{u}. We denote the set of all unvisited vertices whose Dvalues are not infinite (i.e. in the subgraph each of which has a path from the source to itself) by OuterVx. Dijkstra’s algorithm repeatedly searches OuterVx for the vertex with minimum tentative cost (this procedure is called findmin in the article), adds it to the set V and modifies D-values by a procedure, called Relax. Suppose the unvisited vertex with minimum tentative cost is x, the procedure Relax replaces D(u) with min{D(u),D(u)+ cost(x,u)} where u is a vertex in UnusedVx, and cost(x,u) is the cost of edge (x,u). In the Mizar library, there are several computer models, e.g. SCMFSA and SCMPDS etc. However, it is extremely difficult to use these models to formalize the algorithm. Instead, we adopt functions in the Mizar library, which seem to be pseudo-codes, and are similar to those in the functional programming language, e.g. Lisp. To date, there is no rigorous justification with respect to the correctness of Dijkstra’s algorithm. The article presents first the rigorous justification.