Hierarchical QAOA for the Vehicle Routing Problem via Clustered Decomposition and Local Feasibility Repair
We propose a hierarchical quantum approximate optimization framework for solving large-scale Vehicle Routing Problems (VRP) using Quantum Approximate Optimization Algorithm (QAOA). The method decomposes a VRP instance into balanced clusters of customer nodes. We formulate intra-cluster routing as Open loop Traveling Salesman Problems (OTSPs), and inter-cluster routing as a reduced VRP over the cluster representatives and depot. We then map the sub-problems to Ising Hamiltonians and solve with both standard and multi-angle QAOA variants at fixed depth p=3, and merge them to produce a routing path for the original VRP. Additionally, to improve solution feasibility and success probability, we introduce a polynomial-time post-processing protocol that samples candidate bit-strings from the QAOA output using a probability threshold and performs exhaustive local 1 and 2 bit-flip searches around these candidates. Benchmarking on 100 randomly generated 13-node, two-vehicle VRP instances, we show that the post-processed standard-QAOA implementation achieves high success rates and approximation ratios within 1.2-1.5 compared to classical optimizer (Gurobi) solutions, while requiring only 12 logical qubits per subproblem instead of 156 qubits for a direct edge-based encoding. These results provide a proof-of-concept demonstration that hierarchical decomposition, shallow QAOA, and local bit-flip repair can offer a scalable and resource-efficient pathway toward larger VRP instances on near-term quantum devices.