Efficient Batch Processing of Shortest Path Queries in Road Networks

Finding the shortest path from one place to another is an essential operation for various location-based services (LBS), and most of the computations run on the server side. However, as the business grows, the service providers are facing an increasing swarm of path requests submitted during a short time period. The most straightforward solution is deploying more servers, while the operation cost increases at the same time. Therefore, in this work, we aim to improve the efficiency algorithmically by answering a large set of shortest path queries in a batch and reusing sharable computations. Specifically, we first propose the petal A*-1N algorithm to process 1-N shortest path queries by batch without repeated computation. Then we introduce several decomposition methods to cluster the start/target set and answer the whole query set with zigzag scheduling methods to further reduce the total running time. Extensive evaluations on both synthetic and real-world data verify the superiority of our algorithm compared with state-of-the-art methods.