Niching Genetic Programming for Co-Joint Optimization: A Case Study on Replenishment and Transshipment Policies in Dynamic Multi-Site Inventory Management

This paper proposes a novel niching genetic programming (GP) framework for co-joint optimization, which is designed to maintain exploration and solution quality in dynamic environments. The framework comprises a multi-tree representation, with separate trees encoding for the respective optimization tasks, and incorporates a niching mechanism to maintain population diversity and avoid premature convergence. This approach facilitates a comprehensive exploration of the solution space, enabling the discovery of effective and synergistic policies. The key contributions of this work include: (1) a multi-tree GP representation for jointly optimizing interdependent tasks, (2) a niching strategy to enhance diversity and robustness, (3) a unified framework for co-joint optimization tailored to the dynamic problems, and (4) benchmark scenarios to evaluate algorithmic performance. In particular, we present the proposed niching GP on replenishment and transshipment policies co-optimization in dynamic multi-site inventory management as the case study. Jointly optimizing replenishment and transshipment policies is a critical yet challenging problem in dynamic multi-site inventory management. Traditional methods including classical GP struggle to adapt to rapidly changing environments or fail to consider the complexities of co-joint optimization, where the interdependence between policies expands the solution space significantly. Experimental results on synthetic and real-world datasets demonstrate the proposed framework’s superiority over state-of-the-art methods, providing a scalable and algorithmically robust solution to the challenges of co-joint optimization in dynamic systems.