Tensor-Train Decomposition and Quantum En- coding of Multi-Port Voyage Matrices for Maritime Logistics

Kr : The maritime inventory routing problem (MIRP) relies on a voyage‑cost tensor that scales as ports × cargoes × time windows. For a global fleet of 50+ vessels this tensor reaches over 10⁹ entries, exceeding standard RAM and rendering real‑time optimisation impossible. we present a first‑person account of building, compressing, and validating a Tensor‑Train (TT) representation of a 5‑year real‑world voyage matrix. Using a control plane and 8× A100 instance for linear algebra, we decomposed a 1,200‑port × 80‑cargo × 365‑day tensor into a chain of rank‑8 TT‑cores, reducing storage from 1.2 TB to 4 MB—a compression factor of 300,000. The reconstruction error was held below 10⁻⁴ and the compressed tensor degraded an optimal MIRP solution by merely 0.2% while speeding up the solver 50‑fold. we exposed the results via a FastAPI endpoint, enabling other researchers to submit their own AIS data and receive TT‑cores ready for classical optimisation or quantum amplitude encoding on Quantum. The paper details the hardware/software stack, the experimental procedure, in‑depth visualisations, and a practical protocol for integrating Quantum Platform with a reusable, API‑driven research service.