Hello pgRouting Mentors and Community!
My name is Kunal Chandola and i am currently pursuing Btech from IIT kharagpur. I’m excited to introduce myself and express my deep interest in participating in Google Summer of Code 2026 with pgRouting. Over the last few days, I have been analyzing the pgRouting codebase, researching its competitors (like OSRM and GraphHopper), and reviewing current issues (specifically Issue #3010 regarding performance).I wanted to pitch my project idea early to get your feedback and see if it aligns well with a 350-hour (Large).
Currently, running routing algorithms (like Dijkstra or A*) over massive, country-wide road networks in pgRouting is very slow. Because there is no preprocessing phase, the algorithms must traverse and evaluate too many nodes at query-time, creating a major performance bottleneck for large-scale analysis.
I propose the solution for Implementing Contraction Hierarchies (CH) natively in pgRouting using C++ and the Boost Graph Library. This involves two phases:
-
Preprocessing Phase (
pgr_createContraction): Evaluating and “contracting” less important nodes by adding “shortcut edges” between their neighbors. -
Query Phase (
pgr_contractedDijkstra): A lightning-fast, modified bidirectional Dijkstra search that utilizes these shortcuts and only traverses “upward” through the hierarchy, drastically pruning the search space and speeding up the query.
Looking forward to your feedback!