Iterative Negotiation and Oversight: A Case Study in Decentralized Air Traffic Management

TL;DR

This paper tackles the challenge of achieving desirable system-level objectives in decentralized, noncooperative multi-agent coordination by integrating an iterative negotiation framework with a taxation-like oversight mechanism. Building on TACo, the trading auction for consensus, the authors add a central oversight loop that adaptively adjusts a coordination factor and imposes a tax-like intervention to steer negotiations toward efficiency and fairness, while guaranteeing finite-time termination. They prove that increasing the coordination weight yields diminishing cost spreads and that the framework converges in finite rounds, with explicit bounds linking the tax parameter κ to convergence rate and system-optimality gap. A case study on decentralized CTOP demonstrates that the approach can achieve near-centralized performance and provide controllable trade-offs between convergence speed and system efficiency. Numerical experiments corroborate the theoretical bounds and show that higher κ improves fairness and system cost, while still converging reliably, suggesting broad applicability to safety-critical decentralized coordination problems.

Abstract

Achieving consensus among noncooperative agents remains challenging in decentralized multi-agent systems, where agents often have conflicting preferences. Existing coordination methods enable agents to reach consensus without a centralized coordinator, but do not provide formal guarantees on system-level objectives such as efficiency or fairness. To address this limitation, we propose an iterative negotiation and oversight framework that augments a decentralized negotiation mechanism with taxation-like oversight. The framework builds upon the trading auction for consensus, enabling noncooperative agents with conflicting preferences to negotiate through asset trading while preserving valuation privacy. We introduce an oversight mechanism, which implements a taxation-like intervention that guides decentralized negotiation toward system-efficient and equitable outcomes while also regulating how fast the framework converges. We establish theoretical guarantees of finite-time termination and derive bounds linking system efficiency and convergence rate to the level of central intervention. A case study based on the collaborative trajectory options program, a rerouting initiative in U.S. air traffic management, demonstrates that the framework can reliably achieve consensus among noncooperative airspace sector managers, and reveals how the level of intervention regulates the relationship between system efficiency and convergence speed. Taken together, the theoretical and experimental results indicate that the proposed framework provides a general mechanism for decentralized coordination in noncooperative multi-agent systems while safeguarding system-level objectives.

Figures (6)

• Figure 1: Overview of the iterative negotiation and oversight framework. Each iteration consists of three stages: (a) Candidate generation: each agent proposes candidate options based on the broadcast coordination factor $\mathbf{w}$, (b) Negotiation with TACo: agents negotiate and trade secondary assets under the taxation parameter $\kappa$, and (c) Shortfall computation: the oversight mechanism computes shortfalls based on asset shortages and (d) broadcasts the updated coordination factor for the next round.
• Figure 2: Comparison between the conventional and decentralized collaborative trajectory options program (CTOP). (a) Each airline submits a set of trajectory option sets (TOS) that differ in route, altitude, or arrival time. (b) In the proposed decentralized CTOP, each Air Route Traffic Control Center (ARTCC) operates as an autonomous agent that generates candidate trajectory bundles (CTBs) and engages in iterative negotiation and oversight to coordinate with neighboring centers. (c) The final trajectories are determined through this decentralized mechanism, achieving coordinated outcomes without relying on a central controller. (d) In the conventional CTOP, a centralized air traffic control command center assigns trajectories based on predicted sector demand and capacity constraints.
• Figure 3: ARTCC boundary and waypoint map over the continental United States. The map includes the boundaries of all 20 ARTCCs (black lines) and 4,998 waypoints (blue dots). The three ARTCCs selected for the case study—Chicago (ZAU), Indianapolis (ZID), and Atlanta (ZTL)—are highlighted in red.
• Figure 4: Relative system cost and fairness improvement across $\kappa$ values. Each point represents the final value normalized by the performance after first-iteration, grouped by the agents’ average asset value $\bar{b} = \sum_{i\in\mathbf{N}}{b_i}/n$. The colors indicate the asset value groups: blue for low $b$, orange for medium $b$, and yellow for high $b$. Larger $\kappa$ values lead to slower but more balanced convergence, reducing both system cost and Gini index. Boxed numbers indicate the proportion of cases that improved ($<1$), remained unchanged ($=1$), or degraded ($>1$).
• Figure 5: Performance comparison between the proposed decentralized oversight framework and baseline coordination mechanisms. As $\kappa$ increases, the framework achieves efficiency and fairness comparable to the centralized CTOP, outperforming the FCFS and Voting baselines.

Theorems & Definitions (31)

• Lemma 1: Iterative choice generation yields convergent choice sets
• proof
• Lemma 2: Individual transfer value bounds
• proof
• Corollary 1: Payment bound for payers
• proof
• Theorem 1: Finite termination under unit-count reserves
• proof
• Theorem 2: Sufficient condition for termination in terms of $\kappa$
• proof