A General-Equilibrium Approach to Distributed Transportation Planning

Michael P. WellmanAAAI 1992 · OpenReview

Abstract

Market price mechanisms from economics constitute a well-understood framework for coordinating decentralized decision processes with minimal communication. WALRAS is a general "market-oriented programming" environment for the construction and analysis of distributed planning systems, based on general-equilibrium theory. The environment provides basic constructs for defining computational market structures, and a procedure for deriving their corresponding competitive equilibria. In a particular realization of this approach for a simplified form of distributed transportation planning, we see that careful construction of the decision process according to economic principles can lead to effective decentralization, and that the behavior of the system can be meaningfully analyzed in economic terms.

Introduction

In distributed planning, multiple agents must coordinate to achieve a global goal, but central control is often impractical due to geographic separation or information silos. A core challenge is enabling independent, self-interested software modules to cooperate efficiently without excessive communication overhead. This paper addresses this through the lens of transportation planning, where shippers move cargo across a congested network. The tension lies between User Equilibrium (UE), where individuals act selfishly, and System Equilibrium (SE), the global optimum.

Market-Oriented Programming with WALRAS

The proposed solution is Market-Oriented Programming, implemented in the WALRAS environment. Instead of traditional optimization, the designer constructs an artificial economy:

  • Agents: Shippers act as consumers of transport services, while carriers act as producers transforming resources into capacity.
  • Bidding Process: Agents submit supply and demand curves (functions of price) rather than single bids. This decentralized relaxation method, known as tatonnement, iteratively adjusts prices until supply equals demand.
  • Price Discovery: Crucially, agents only interact with price signals for relevant goods, ensuring minimal communication overhead.

Experimental Setup

The approach was evaluated using a standard abstract transportation network benchmark. This network consists of four locations connected by directed links, each with quadratic cost functions to simulate congestion (diseconomies of scale).

The four-node abstract transportation network used as a benchmark for the WALRAS environment.

Results and Discussion

The WALRAS system successfully derived the System Equilibrium (SE), reaching 10% accuracy within 36 bidding cycles and 1% within 72 cycles. A key insight is the role of rents: by having carriers charge marginal costs (including congestion rents) and redistributing these to shippers, the decentralized market naturally converges to the global optimum.

Furthermore, the flexibility of the environment allows for simulating different economic regimes by altering agent logic. For instance, switching carriers from marginal-cost to average-cost pricing shifts the system from SE to UE.

Comparison of costs and link prices under marginal cost (SE) and average cost (UE) pricing regimes.

As shown in the table above, the marginal cost pricing regime results in a lower total system cost (1136 vs 1143) compared to the average cost regime, demonstrating the efficiency gains of the SE approach. The author also noted that asynchronous bidding helped stabilize the system by preventing the oscillations common in synchronized updates.

Conclusion

This research demonstrates that complex network routing can be solved effectively through decentralized market mechanisms. Prices serve as highly compressed signals, enabling global coordination with minimal information exchange. While the model currently assumes perfect competition and convex cost functions, it provides a robust foundation for future work addressing economies of scale, discrete movements, and temporal constraints in distributed planning.

Summary updated

Abstract

Market price mechanisms from economics constitute a well-understood framework for coordinating decentralized decision processes With minimal communication. WALRAS is a general "market oriented programming" environment for the construction and analysis of distributed planning systems, based on general-equilibrium theory. The environment provides basic constructs for defining computational market structures, and a procedure for deriving their corresponding competitive equilibria. In a particular realization of this approach for a simplified form of distributed transportation planning, we see that careful construction of the decision process according to economic principles can lead to effective decentralization, and that the behavior of the system can be meaningfully analyzed in economic terms.
BibTeX
@inproceedings{wellman1992general,
  author = {Michael P. Wellman},
  title = {{A General-Equilibrium Approach to Distributed Transportation Planning}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {1992},
  url = {https://openreview.net/forum?id=SybXcTguZS},
}
Figure