BinR-LRP: A divide and conquer heuristic for large scale LRP with integrated microscopic agent-based transport simulation

Abstract

The holistic optimisation of transportation systems is one of the key challenges in transportation science, because it requires the simultaneous consideration of the numerous interactions between the strategic planning level (e.g., the Facility Location Problem [FLP]) and the tactical and operational planning levels (e.g., Vehicle Fleet and Vehicle Routing Problem [VRP]). Traditional methods for solving the Location Routing Problem (LRP) often focus on the fixed constraints and ignore the variable vehicle characteristics, dynamic operations, different modes or underlying infrastructure. This paper proposes an integrated approach for modular and intuitive metaheuristic for LRP. The route planning phase is incorporated by means of agent-based transport simulation, which provides additional flexibility with respect to the vehicle fleet, demand characteristics, or the use of external problem constraints. Therefore, this approach can be easily applied to practical problems and used to optimise transport networks in a flexible and modular manner. Moreover, the algorithm developed here can independently converge to the near-optimal number and location of logistics sites. We also demonstrate the effectiveness and the performance of our approach by performing several simulation experiments in the context of a sensitivity analysis and comparing the results with well-known benchmark solutions. The results indicate that the Binary-Partition LRP heuristic (BinR-LRP) is able to identify better solutions than the benchmark heuristics in most cases. This emphasises its suitability as a scalable and robust optimisation framework, even for oversized LRP instances.