Integrated Location Routing and Inventory Optimization

39 concepts35 papers1 proposal

In traditional logistics, supply chain management is treated as a sequence of independent, isolated steps. First, an organization decides where to build its warehouses or distribution hubs (Facility Location). Next, it determines how much stock to hold at each hub to satisfy customer demand (Inventory Management). Finally, it maps out the daily paths for its delivery trucks (Vehicle Routing).

Because each of these problems is individually NP-hard, separating them into distinct operational silos makes mathematical sense. However, as single-author works by Huiyan Li and Yutong Gong argue, this fragmented approach creates a structural blind spot. Optimizing individual stages in isolation frequently results in global inefficiencies; a cost-saving measure in one area can create disproportionately high costs or bottlenecks in subsequent stages of the fulfillment process (Damla Kizilay et al.; Yiheng Zhu et al.).

The core insight of integrated supply chain optimization is that these distinct layers are deeply coupled. By breaking down the silos and solving these problems simultaneously, researchers have unlocked massive operational efficiencies. As reviewed by Gilbert Laporte et al. in A concise guide to existing and emerging vehicle routing problem variants, jointly optimizing strategic facility placement, inventory levels, and operational vehicle routes yields a 5% to 15% reduction in total costs compared to sequential planning.

The Coupling of Geography and Motion: Location-Routing

To understand why integrated optimization works, consider the relationship between facility location and vehicle routing.

If you choose a warehouse location based purely on static geographic proximity to customers, you will likely place it at the geometric center of the customer cluster. However, delivery vehicles do not travel back and forth from the warehouse to each individual customer; they visit multiple customers in a single continuous loop (a tour).

Sequential Optimization (Suboptimal):
[Warehouse at Center] ---> (Customer 1) ---> (Customer 2) ---> (Customer 3) ---> [Back to Center]
Result: Overlapping paths, higher total travel distance.

Integrated Optimization (Global Optimum):
[Warehouse at Perimeter] ---> (Customer 1) ---> (Customer 2) ---> (Customer 3) ---> [Back to Perimeter]
Result: Smooth, circular sweep; minimized total travel distance.

As Jingyang Zhao et al. demonstrate in Improved Approximation Algorithms for Capacitated Location Routing, the selection of depot locations fundamentally alters the optimal delivery paths. When infrastructure costs and operational travel distances are tightly coupled, an off-center depot that facilitates a smooth, circular vehicle tour can be significantly cheaper than a central depot that forces vehicles into overlapping, inefficient paths. Xueming Yan et al. note that because decisions in one domain constrain the other, finding a true global optimum requires evaluating infrastructure choices through the lens of actual delivery feasibility.

This geographic coupling is especially critical in modern and emergent logistics networks:

  • Electric Vehicle Fleets: In urban logistics, the placement of charging infrastructure dictates vehicle range and routing. U. Breunig et al. show that moving from a pre-defined list of charging locations to endogenous facility selection slashes routing detours. Similarly, a mathematical framework developed for Joint Optimization of Autonomous Electric Vehicle Fleet Operations and Charging Station Siting demonstrates that co-designing charging infrastructure and fleet routing reduces total costs and peak grid loads by up to 10%.
  • Emergency Response: In disaster relief, planners cannot afford sequential delays. Jia Luo et al. and Hanping Hou et al. show that integrating the placement of temporary supply depots with last-mile distribution routing ensures that facility location decisions are directly informed by the dynamic, real-time constraints of the physical environment.

The Coupling of Time and Quantity: Inventory-Routing

A similar interdependence exists between inventory management and vehicle routing. This is known as the Inventory Routing Problem (IRP).

In a siloed system, inventory managers minimize holding costs and stockout risks by ordering specific quantities at specific times. They then hand these delivery requirements over to the transportation team.

However, the timing of a delivery dictates which customers can be grouped into the same vehicle route. If Customer A needs a delivery on Monday and Customer B needs one on Tuesday, a vehicle must make two separate trips. If the inventory strategy is flexible enough to deliver to both on Monday—accepting a minor holding cost penalty at Customer B—the transportation team can combine them into a single route, drastically reducing travel costs.

Siloed Inventory & Routing:
Day 1: Depot -------> Customer A (Trip 1)
Day 2: Depot -------> Customer B (Trip 2)

Integrated Inventory & Routing:
Day 1: Depot -------> Customer A -------> Customer B (Single Joint Tour)
Day 2: No trips.

Hoong Chuin Lau et al. proved the power of this integration in Solving a Supply Chain Optimization Problem Collaboratively, achieving a 10.5% operational cost reduction over sequential methods by simultaneously balancing storage and transportation trade-offs. By consolidating data across multiple suppliers and clients into a single framework, Lukas Seeholzer et al. demonstrate that synchronizing inventory and routing decisions allows vehicle fleets to be shared across different product types, eliminating redundant trips.

The Unified Triad: Location, Inventory, and Routing

The frontier of this research direction lies in combining all three components into a unified optimization framework.

When facility location, inventory positioning, and vehicle routing are solved together, the mathematical model captures the full lifecycle of a supply chain network. Yang Jiao et al. provided the first constant-factor approximation algorithms for this integrated problem in Inventory Routing Problem with Facility Location, proving theoretically that minimizing the combined costs of infrastructure, transportation, and inventory storage simultaneously yields superior network designs.

Neglecting any piece of this triad can render a supply chain fragile. For instance, Yongzhen Li et al. point out that many robust facility location models fail in real-world post-disaster logistics because they neglect inventory positioning and delivery sequencing. Conversely, a single-author study by Saba Ghasemi Naraghi demonstrates that integrating facility allocation and inventory management under a unified robust model allows supply chains to balance cost-efficiency with waste reduction and community access, even under highly volatile demand.

Algorithmic Paradigms for Solving Coupled Systems

If integrated optimization is universally better, why wasn't it always done this way? The barrier is computational complexity. Combining multiple NP-hard problems creates an exponential explosion in the solution space. To make these models practical, researchers have developed several algorithmic strategies:

  1. Collaborative Decomposition: Instead of solving the massive joint problem directly, Hoong Chuin Lau et al. decompose it into sub-problems (e.g., an inventory sub-problem and a routing sub-problem) that iteratively pass information back and forth, converging rapidly on a near-global optimum.
  2. Advanced Mathematical Programming: Damla Kizilay et al. demonstrate that for complex, multi-stage logistics hubs, Constraint Programming (CP) can find optimal solutions for integrated operations orders of magnitude faster than traditional Mixed Integer Programming (MIP) by more naturally exploiting problem structure.
  3. Metaheuristics and Learning: For highly dynamic environments, researchers leverage evolutionary algorithms and machine learning. Xueming Yan et al. use a hierarchical fuzzy graph to share optimization "knowledge" across different echelons of a supply chain. In e-commerce fulfillment, Yiheng Zhu et al. introduce a Multi-Agent Reinforcement Learning (MARL) environment called OFCOURSE to let agents learn joint policies across packing and delivery stages simultaneously.

Ultimately, whether applied to warehouse automation (Lin Xie et al. on integrated order batching and robot routing) or critical infrastructure recovery (Pascal Van Hentenryck et al. on coordinating power grid restoration with repair crew routing), the core lesson remains the same: treating logistics as a single, deeply coupled system reveals efficiencies that are completely invisible from within an operational silo.

Key Foundational Work

Integrating Local Search and Network Flow to Solve the Inventory Routing Problem (2002) — contributed 2 concept elements to this direction.
Solving a Supply Chain Optimization Problem Collaboratively (2000) — contributed 1 concept elements to this direction.
Collaborative Model and Algorithms for Supporting Real-Time Distribution Logistics Systems (2000) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_integrated_supply_chain_optimization_8e8b85ad,
  title = {integrated supply chain optimization},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/integrated-supply-chain-optimization/rdr_8e8b85ad055246a889af6c0db7eb2a0e},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_integrated_supply_chain_optimization_8e8b85ad_v1,
  title = {integrated supply chain optimization},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/integrated-supply-chain-optimization/rdr_8e8b85ad055246a889af6c0db7eb2a0e/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake