The Vehicle Routing Problem with Pickups and Deliveries (VRPPD) constitutes a fundamental challenge in combinatorial optimization.
When extended to incorporate heterogeneous fleets, time-window restrictions, and vehicle capacity constraints, the problem’s complexity grows combinatorially, rendering exact methods computationally prohibitive at scale.

In such a setting, the central question becomes: how would you design an algorithm capable of producing high-quality solutions under these constraints? This was precisely the challenge posed by the 2024 Optimization Grand Challenge, hosted by LG CNS in collaboration with the Korean Institute of Industrial Engineers.

Our approach combined Adaptive Large Neighborhood Search (ALNS) with Set Partitioning Problem formulation. The hybridization of heuristic flexibility with exact optimization yielded a decisive outcome: 1st place out of 378 teams.

Further details and coverage:

This achievement was particularly meaningful as it was a joint effort with my longtime collaborator and friend, Wonjae.