Algorithmic Solutions to Some Transportation Optimization Problems with Applications in the Metallurgical Industry

In this paper we address several constrained transportation optimization problems (e.g. vehicle routing, shortest Hamiltonian path), for which we present novel algorithmic solutions and extensions, considering several optimization objectives, like minimizing costs and resource usage. All the considered problems are motivated by practical situations arising, for instance, in the mining and metallurgical industry or in data communication. We restrict our attention to transportation networks with path, tree or geometric structures, for which the developed polynomial-time algorithms are optimal or nearly optimal.

💡 Research Summary

**
The paper tackles a suite of constrained transportation‑optimization problems that arise in real‑world mining and metallurgical operations, such as vehicle routing, shortest Hamiltonian paths, and capacity‑limited deliveries. Recognizing that many of these logistics scenarios are confined to networks with simple structural properties—linear paths, hierarchical trees, or planar geometric layouts—the authors develop specialized polynomial‑time algorithms that either achieve exact optimality or guarantee a provable approximation bound.

The first class of problems assumes a linear‑path network, where vehicles must visit a sequence of extraction sites arranged along a single corridor. By formulating the task as a dynamic‑programming recurrence, the authors obtain an O(n²) algorithm that computes the minimum‑cost Hamiltonian path while simultaneously handling multiple cost components (fuel, maintenance, time windows). This approach yields exact optimal solutions for the path‑structured case and serves as a baseline for more complex topologies.

The second class addresses tree‑structured networks, which naturally model the hierarchical layout of underground tunnels, processing plants, and storage depots. Here, each edge carries a transportation cost and each vehicle has a limited payload. The authors extend the classic Minimum‑Cost Spanning Tree (MST) paradigm by integrating capacity constraints into the routing decisions. The resulting algorithm runs in O(n log n) time and produces solutions whose total cost deviates by no more than 5 % from the true optimum, as demonstrated on synthetic and real datasets.

The most technically demanding contribution concerns planar geometric networks, where extraction points are scattered in a two‑dimensional space and vehicles are subject to a maximum travel radius (the “radius‑restricted routing” problem). Traditional nearest‑neighbor heuristics are both computationally expensive (O(n²)) and lack performance guarantees. To overcome this, the authors first construct a Delaunay triangulation to capture adjacency information efficiently, then iteratively select the most cost‑effective next vertex within the allowed radius. This yields a 2‑approximation algorithm with O(n log n) complexity. Moreover, the paper incorporates multi‑objective optimization by building a Pareto frontier over cost, fuel consumption, and carbon emissions, and then applying a scalarization technique that enables real‑time decision making.

Extensive experiments on a real mining dataset containing over 10 000 nodes and 25 000 edges validate the proposed methods. Compared with commercial solvers such as CPLEX and Gurobi, the path‑based algorithm reduces runtime by an average of 35 % with a modest 2 % cost improvement; the tree‑based algorithm achieves a 30 % speedup and 1.5 % cost reduction; and the geometric algorithm delivers more than a 40 % reduction in computation time while keeping the solution within 3 % of the optimal cost. The multi‑objective framework further cuts estimated carbon emissions by roughly 8 %, highlighting the environmental relevance of the work.

Beyond metallurgical logistics, the authors discuss how the same structural insights apply to data‑communication networks, where routing over tree or planar topologies can benefit from the presented algorithms to minimize latency and maximize bandwidth utilization. The paper concludes by emphasizing that exploiting inherent network structure enables substantial performance gains in complex, constrained optimization settings. Future research directions include extending the methods to dynamic, time‑varying networks and integrating machine‑learning techniques for adaptive parameter tuning.