---
title: Resource Substitution Problem Overview
url: https://www.emergentmind.com/topics/resource-substitution-problem-rsp
type: topic
---

# Resource Substitution Problem Overview

The Resource Substitution Problem (RSP) concerns determining optimal reassignments among a set of interchangeable and potentially composite resource types, so as to minimize persistent imbalances induced by asymmetrical flows in large-scale, directed networks. Originating in domains such as logistics, the RSP is to find a new assignment—subject to task requirements, resource compatibility, and operational fairness—that minimizes the number of resource substitutions needed to balance the overall network. Practical settings commonly include cost-sensitive substitution and decentralized scheduling, further motivating solutions that integrate formal optimization (operations research, OR) and data-driven guidance (machine learning, ML) [2511.18269].

## 1. Formal Problem Specification

The RSP is defined on a directed graph $\mathcal{G}=(\mathcal{N},\mathcal{A})$ where $\mathcal{N}$ is the set of nodes and $\mathcal{A}$ the set of arcs, each representing a required task assigned to one resource from a predefined set $\mathcal{R}$ (resources may represent equipment, vehicles, or composite arrangements). Each arc $a$ admits a subset $\mathcal{R}_a\subseteq\mathcal{R}$ of feasible resources, e.g., two short trailers may substitute for a long trailer if both endpoints permit. $\Phi_0:\mathcal{A}\rightarrow\mathcal{R}$ encodes the initial resource assignment.

For any assignment $\Phi$, the node-level imbalance of resource $r$ at node $n$ is defined as:
$$
I_{n,r}(\Phi) = \left|\,\,|\{a\in\mathcal{A}_n^+ : \Phi(a)=r\}| - |\{a\in\mathcal{A}_n^- : \Phi(a)=r\}|\,\,\right|,
$$
with $\mathcal{A}_n^+$ and $\mathcal{A}_n^-$ the sets of arcs into/out of $n$, respectively. The total network imbalance is:
$$
\mathcal{I}(\Phi) = \sum_{n\in\mathcal{N}}\sum_{r\in\mathcal{R}} I_{n,r}(\Phi).
$$

Binary variables $x_{a,r}$ indicate assignment $x_{a,r}=1$ iff resource $r$ is selected for arc $a$. Additional variables—$I_{n,r}$ (linearization of imbalance), $Z$ (max per-scheduler substitutions), and $D_{s_1s_2}$ (pairwise difference for Gini-based fairness)—are used in Stage 2 modeling.

## 2. Mixed-Integer Programming Formulations

The RSP solution process is structured in two mathematical programming stages:

### Stage 1: Minimize Network Imbalance

Formulation:

Minimize network imbalance:
$$
I^* = \min \sum_{n\in\mathcal{N}}\sum_{r\in\mathcal{R}} I_{n,r}
$$
subject to:
$$
\sum_{a\in\mathcal{A}_n^+} x_{a,r} - \sum_{a\in\mathcal{A}_n^-} x_{a,r} \le I_{n,r} \quad \forall n, r
$$
$$
\sum_{a\in\mathcal{A}_n^-} x_{a,r} - \sum_{a\in\mathcal{A}_n^+} x_{a,r} \le I_{n,r} \quad \forall n, r
$$
$$
\sum_{r\in\mathcal{R}_a} x_{a,r} = 1\quad \forall a;\qquad x_{a,r} \in \{0,1\}
$$

### Stage 2: Minimize Substitutions Under Fairness

Let $\Delta = \sum_{a}(1-x_{a,\Phi_0(a)})$ (total substitutions), $\mathcal{S}$ the set of schedulers, $\mathcal{A}_s\subseteq\mathcal{A}$ arcs owned by scheduler $s$, and $B_s = \sum_{a\in\mathcal{A}_s}(1-x_{a,\Phi_0(a)})$ scheduled substitutions.

- **Pure efficiency**: $\min\,\Delta$ s.t. $\sum_{n,r} I_{n,r}\le I^*$ (Stage 1 constraints)
- **Minimax fairness**: $\min\,Z$ s.t. $Z \ge B_s$ $\forall s$ and $\sum_{n,r} I_{n,r}\le I^*$
- **Weighted trade-off**: $\min (1-\alpha)\Delta + \alpha Z,$ $\alpha\in[0,1]$
- **Gini-based fairness**: $\min (1-\omega)\sum_s B_s + \omega\sum_{s_1<s_2} D_{s_1s_2}$ (with $D_{s_1s_2}\ge |B_{s_1}-B_{s_2}|$ for all $s_1<s_2$)

These formulations enable multiple fairness paradigms: minimizing peak burden (minimax), balancing overall changes (weighted), or reducing inequality (Gini).

## 3. Fairness Considerations in Resource Substitution

Multiple fairness metrics operationalize diverse organizational policies and scheduler equity:

- **Minimax fairness**: minimizes maximum substitutions incurred by any single scheduler, i.e., $\min\max_s B_s$.
- **Weighted objectives**: convex combinations, $(1-\alpha)\sum_s B_s + \alpha \max_s B_s$, interpolate between total-change and peak-burden criteria.
- **Gini-style fairness**: minimizes total pairwise difference $\sum_{s_1<s_2}|B_{s_1}-B_{s_2}|$, reflecting a Gini index–like inequality measure.

Weighted trade-off and Gini approaches support generating Pareto fronts to facilitate scheduler review and selection of context-specific trade-offs between efficiency and equity.

## 4. ML-Enhanced Candidate Pruning and Decision Support

To address scale and complexity, ML augments the OR formulations in two principal ways:

### Learning Arc-Resource Preference Scores

A deep neural network (DNN) is trained on historical labeled (arc, resource) assignments $y_{a,r}$ to estimate $p_{a,r}=P(\Phi(a)=r\,|\,\text{features})$. Feature vectors $X_{a,r}$ include node IDs, volume, scheduler identities, temporal indicators, distance, equipment size, and other operational variables. The DNN outputs a softmax vector $Y_a=(p_{a,1},\dots,p_{a,|\mathcal{R}|})$ trained via cross-entropy.

### Dynamic Top-$\kappa$ Resource Selection

For each arc $a$, betweenness centrality $B(a)$ partitions arcs into Low, Medium, or High centrality. Thresholds $\tau_1<\tau_2$ determine $\kappa_a\in\{\kappa_\text{low},\kappa_\text{med},\kappa_\text{high}\}$. For each arc, only the top $\kappa_a$ resources (ranked by $p_{a,r}$) are retained; other assignment variables are fixed to zero. This procedure sharply reduces problem dimensionality:
```
for each arc a∈𝒜:
    Compute B(a)
    if B(a)<τ₁ then κ_a←κ_low
    else if B(a)<τ₂ then κ_a←κ_med
    else κ_a←κ_high
    Let R̂_a← top‐κ_a resources by p_{a,r}
    Impose x_{a,r}=0 ∀r∉R̂_a
end
```
This dynamic pruning executes prior to Stage 1 optimization, resulting in an arc-resource candidate set tailored for both operational centrality and empirical preference.

## 5. Integrated OR–ML Workflow and Computational Acceleration

The solution architecture integrates ML-driven candidate reduction with two-stage mixed-integer programming, as follows:

1. Offline DNN training on historical $(a, r)$ data.
2. For new instances: predict $p_{a,r}$, compute $B(a)$, and dynamically select reduced resource candidate sets $\hat{\mathcal{R}}_a$ for each arc.
3. Solve Stage 1 MIP on the pruned assignment model for the minimum imbalance $I^*$.
4. Solve Stage 2 for diverse fairness objectives, yielding a portfolio of solutions using the same reduced model.
5. Output the Pareto frontier of efficiency–fairness trade-offs for scheduler selection.

Additional acceleration strategies include warm-starting via iterated local search around the initial assignment $\Phi_0$, Dantzig–Wolfe decomposition or column generation for $x_{a,r}$, and targeted MIP solver tuning (e.g., cut aggressiveness, presolve configuration).

## 6. Empirical Performance in Large-Scale Logistics Networks

Applied to a package delivery network (2–40 schedulers, up to 1,588 nodes, up to $8.3\times10^6$ arcs, 13–23 equipment types), the hybrid OR–ML framework yields:

- **Model-size reduction**: 80–93% via dynamic top-$\kappa$ pruning.
- **Execution-time reduction**: 90–99% (MIP times cut from hundreds to 1.0–26.5 s in Stage 1).
- **Imbalance reduction**: Final imbalance $I^*$ reduces 16–22%.
- **Optimality preservation**: Reduced models retain global optima for $\kappa\ge 3$ or with dynamic pruning.
- **Fairness trade-offs**: As fairness weight $\alpha$ (in weighted trade-offs) increases $0 \rightarrow 1$, total substitutions increase up to 89% in the largest instances, but maximum per-scheduler burden drops by 17–18%.
- **Portfolio insights**: Weighted objectives generate a Pareto front for schedulers, and partially implemented fair solutions (applying only top-$N$ changes) improve over truncations of pure-efficiency plans. Increased collaboration in arc ownership further reduces both imbalance and fairness cost.

For a representative case (50 schedulers, 8,000 nodes, 24 million arcs): static $\kappa=1$ yields a 9.7% optimality gap in 133 s; $\kappa=5$ achieves optimality in 2,376 s; dynamic $\kappa_a$ averaging 2.2 attains the optimum in 132 s with 86% arc reduction. These results demonstrate preservation of optimality and substantial computational gains using the ML-augmented, candidate-pruned approach.

## 7. Summary of Benefits and Implications

The hybrid OR–ML approach for the Resource Substitution Problem addressed in [2511.18269] demonstrates significant scalability and fairness advances:

- Up to 93% reduction in model size.
- Up to 99% reduction in solution times.
- 16–22% reduction in network imbalance.
- Up to 18% reduction in maximum per-scheduler burden for fairer solutions.

A plausible implication is the feasibility of deploying near-optimal, fairness-aware substitution plans in decentralized, real-world logistics contexts at scales unattainable by classical methods. Further, portfolio-based Pareto-efficient trade-offs facilitate more informed, preference-aligned choices by human schedulers within large organizations.

Source: https://www.emergentmind.com/topics/resource-substitution-problem-rsp