Optimal pricing control in distribution networks with time-varying supply and demand

Mathias Bürger, Claudio De Persis, Frank Allgöwer

Abstract

This paper studies the problem of optimal flow control in dynamic inventory systems. A dynamic optimal distribution problem, including time-varying supply and demand, capacity constraints on the transportation lines, and convex flow cost functions of Legendre-type, is formalized and solved. The time-varying optimal flow is characterized in terms of the time-varying dual variables of a corresponding network optimization problem. A dynamic feedback controller is proposed that regulates the flows asymptotically to the optimal flows and achieves in addition a balancing of all storage levels.

Introduction

In traditional network optimization, solutions typically address the most efficient way to route goods or resources given a static set of requirements. However, real-world infrastructures—such as global supply chains, power grids, and data networks—are inherently dynamic. Supply and demand fluctuate constantly, requiring a control system that can adapt in real-time.

The challenge in these systems is twofold. First, there is the problem of Optimal Routing: adjusting flows instantaneously to minimize transportation costs as conditions shift. Second, there is Inventory Balancing: ensuring that storage nodes across the network maintain equitable levels to prevent localized overflows or stockouts. This paper proposes a continuous-time control law that addresses these challenges while respecting the physical capacity limits of the transportation lines.

Methodology

Network Modeling and Legendre-type Costs

The authors model the distribution network as a directed graph where nodes represent inventory storage and edges represent transportation lines. To manage the physical constraints of these lines, the framework utilizes Legendre-type cost functions. Unlike simple quadratic costs, these convex functions are designed such that as the flow approaches a physical capacity limit, the marginal cost approaches infinity. This mathematical property naturally prevents the controller from violating hard capacity constraints without the need for complex switching logic.

Dual Variables and the Dirichlet Problem

A central contribution of the paper is the characterization of optimal flow through time-varying dual variables (denoted as $\zeta$). In economic terms, these variables represent "shadow prices" for the goods at each node. To enable real-time control, the authors derive an explicit differential equation for these dual variables by solving a generalized version of the Dirichlet problem on a graph. This allows the system to determine how prices should evolve dynamically to maintain network optimality.

Dynamic Feedback Control

The proposed controller is structured as a continuous-time Newton algorithm. It integrates two primary components:

  1. Feed-forward control: This component anticipates changes in external supply and demand signals.
  2. Feedback control: This component corrects for inventory imbalances, driving the system toward a state of "consensus" where storage levels are equalized across the network.

Key Insights

Prices as Control Signals

One of the most significant insights of the research is the functional use of dual variables. By treating shadow prices as the state of the controller, the system naturally finds the most efficient flow. The "price" at a node dictates the flow direction and magnitude, effectively decentralizing the logic of the optimization.

Consensus and Balancing

The research links the physical problem of inventory balancing to the mathematical problem of "output agreement" in multi-agent systems. By integrating balancing directly into the optimization objective, the controller ensures that the network does not just move goods efficiently, but also maintains a robust distribution of inventory.

Soft Constraints for Hard Limits

The use of Legendre-type functions allows the system to treat hard capacity limits as "soft" constraints within the mathematical optimization. This ensures the control law remains continuous and stable, even when the network is operating near its physical limits.

Simulation and Results

As the paper is primarily theoretical, the authors validated the controller using a numerical simulation of a three-node inventory system. The supply and demand were modeled as harmonic signals (sine waves) generated by an external "exo-system."

  • Asymptotic Optimality: The authors provided rigorous mathematical proof that the controller causes network flows to converge to optimal values over time, regardless of the initial state.
  • Robustness to Shocks: During simulations, the system was subjected to instantaneous phase shifts in supply and demand. The controller successfully readjusted flows and re-balanced inventories without violating capacity bounds.
  • Operational Efficiency: The simulation demonstrated that the controller intelligently prioritizes low-cost routes, only utilizing more expensive transportation lines as the cheaper options approach their capacity limits.

Limitations and Future Directions

While the proposed controller is robust, it currently assumes that external supply and demand (and their derivatives) are known or measurable. Future research is needed to develop designs that can operate with limited or estimated information. Additionally, the current derivation is centralized; for massive networks like the global internet, a distributed implementation—where each node communicates only with its immediate neighbors—would be a necessary evolution. Finally, the model currently assumes a balanced global supply and demand; handling net surpluses or deficits dynamically remains an open challenge for the field.

Summary updated

Abstract

This paper studies the problem of optimal flow control in dynamic inventory systems. A dynamic optimal distribution problem, including time-varying supply and demand, capacity constraints on the transportation lines, and convex flow cost functions of Legendre-type, is formalized and solved. The time-varying optimal flow is characterized in terms of the time-varying dual variables of a corresponding network optimization problem. A dynamic feedback controller is proposed that regulates the flows asymptotically to the optimal flows and achieves in addition a balancing of all storage levels.
BibTeX
@article{burger2015optimal,
  author = {Mathias Bürger and Claudio De Persis and Frank Allgöwer},
  title = {{Optimal pricing control in distribution networks with time-varying supply and demand}},
  journal = {IEEE Trans. Control. Netw. Syst.},
  year = {2015},
  url = {https://openreview.net/forum?id=Q1JpDp0lfZK},
}
Figure