Path Inference Based on Voronoi Graph

Abstract

Traditional path planning approach is generally focused on finding a single one shortest path given the temporal or spatial constraints. However, the strategy of single-one optimal path is unable to adapt to various situations in real applications. For example, in the scenarios of vehicle patrolling and search &rescue, the comprehensiveness of coverage is much more important. Rather than finding a single one optimal shortest path as the traditional path planning does, our path inference approach aims at inferring all the possible paths which satisfy the specified temporal and spatial constraints. Path inference is able to provide more valuable insights for internet routing, migration pattern study, urban construction and adversary profiling. In this work, we propose a novel path inference algorithm based on Voronoi graph. We model the 2D geospatial topology with Voronoi cells according to the spatial constraint, represent the adjacency relationships of Voronoi cells with an adjacency matrix and infer all the possible paths in the derived Voronoi graph satisfying the user-specified temporal and spatial constraints. Experimental results indicate that our path inference algorithm based on Voronoi graph is able to perform more delicately and efficiently than the traditional path planning approach in term of both space partition and path inference.