Resilient and Efficient Allocation for Large-Scale Autonomous Fleets via Decentralized Coordination

TL;DR

This work tackles scalable allocation of scarce resources for large autonomous fleets operating under stochastic consumption. It introduces DESIRA, which combines side-information conditioned risk shaping with a consensus-ADMM coordination scheme to satisfy chance constraints while sharing capacity over a sparse graph. Core contributions include a tractable risk reformulation $r_i(\phi_i, \varepsilon)=\mu_i(\phi_i)+\Phi^{-1}(1-\varepsilon)\sigma_i(\phi_i)$, an ADMM-based decomposition with efficient projections, online conformal calibration, and cross-domain validation on urban and orbital platforms. Results show 30–55% reductions in failure rates at near-centralized costs and near-linear scalability to thousands of agents, underscoring practical resilience for real-world deployments.

Abstract

Operating large autonomous fleets demands fast, resilient allocation of scarce resources (such as energy and fuel, charger access and maintenance slots, time windows, and communication bandwidth) under uncertainty. We propose a side-information-aware approach for resource allocation at scale that combines distributional predictions with decentralized coordination. Local side information shapes per-agent risk models for consumption, which are coupled through chance constraints on failures. A lightweight consensus-ADMM routine coordinates agents over a sparse communication graph, enabling near-centralized performance while avoiding single points of failure. We validate the framework on real urban road networks with autonomous vehicles and on a representative satellite constellation, comparing against greedy, no-side-information, and oracle central baselines. Our method reduces failure rates by 30-55% at matched cost and scales to thousands of agents with near-linear runtime, while preserving feasibility with high probability.

Figures (5)

• Figure 1: Urban road network (Austin) with stations (blue) and demand nodes (orange). The network topology is extracted from OpenStreetMap data using OSMnx.
• Figure 2: Performance comparison across allocation methods. Lower values indicate better performance for both cost and failure metrics. DESIRA demonstrates reduced failure rates while maintaining competitive cost efficiency.
• Figure 3: Relationship between failure probability and communication radius. The diminishing returns at larger radii demonstrate that limited connectivity suffices for effective coordination.
• Figure 4: Computational runtime as a function of fleet size. The near-linear scaling of DESIRA demonstrates superior performance compared to centralized optimization for large-scale deployments.
• Figure 5: Temporal evolution of satellite constellation health status. Each point represents a satellite in its orbital plane, with color gradient from green (healthy) to red (degraded) indicating operational status across four time instances.