Developing Original Meta-Heuristic Algorithms for Large-Scale Multi-Echelon Inventory Optimization in Nigerian Distribution Networks

Multi-echelon inventory optimization in large-scale distribution networks is a computationally intractable problem for which exact methods fail to deliver solutions within operationally relevant timeframes, motivating continued development of effective meta-heuristic approaches tailored to real-world problem characteristics. This dissertation develops original meta-heuristic algorithms for multi-echelon inventory optimization in Nigerian distribution networks, addressing the combined challenges of demand uncertainty, multiple product families, stochastic lead times, and capacitated storage facilities. A comprehensive literature review identifies gaps in meta-heuristic performance on multi-echelon problems with the specific structural features present in Nigerian distribution contexts, including sparse transport networks, seasonal demand clustering, and informal distribution tier interactions. Three novel algorithms are developed and theoretically analyzed: a hybrid genetic algorithm with adaptive population dynamics, a variable neighbourhood search with domain-specific neighbourhood structures, and an ant colony optimization variant exploiting echelon-specific pheromone representations. Algorithm convergence properties are formally characterized, and computational complexity bounds are derived. Extensive numerical experiments on benchmark instances and empirical problem instances derived from a nationwide distribution company operating across all geopolitical zones benchmark the proposed algorithms against state-of-the-art alternatives. Results demonstrate that the hybrid genetic algorithm consistently achieves solutions within 2.1 percent of a lower bound on total inventory cost while solving instances 40 times larger than those tractable by commercial solvers. The dissertation contributes novel algorithmic methodologies to the operations research literature on inventory optimization with direct applicability to Nigerian and sub-Saharan African distribution contexts. Keywords: meta-heuristics, multi-echelon inventory, distribution optimization, genetic algorithm, Nigeria