VERA improves safe multi-agent control with verifiable action feasibility
VERA: Verifiable Feasibility Representations with Counterfactual Credit for Constrained Multi-Agent Control
Machine Learning
Summary
Controlling many agents like drones or robots is tricky because they must choose actions that are not just good but also possible under changing conditions, like limited shared resources or strict deadlines. The authors created VERA, a system where each agent checks if its actions are doable and then gets credit for how well its choices help the whole group. This approach lets agents learn better decisions without needing special knowledge when they act. Tests show VERA works well in complex network control tasks, keeping actions safe and successful.
What this means in practice
- •For network control engineers: Improve scheduling and coordination for nodes in integrated space-air-ground communication networks with guaranteed action feasibility and minimized constraint violations.
- •For robotics system designers: Design multi-robot controllers that better respect shared resource limits and timing constraints by separating feasibility checks from credit assignment in learning.
Authors
Bo Yin, Dongbo Li, Hongkai Chen, Jie Liu, Guoliang Xing
Abstract
Constrained multi-agent control requires more than predicting rewarding actions: an action can cease to be executable as contact windows, shared capacity, and deadlines change. We introduce VERA, a centralized-training, decentralized-execution framework that separates feasibility estimation from credit assignment. Each actor predicts a five-dimensional verifiable feasibility representation (VFR). After an action is proposed, exact action-conditioned margins available only during training supervise that representation, while a counterfactual group-relative advantage (CGRA) ranks candidate representation-action pairs. Execution uses one actor pass and no privileged state. In a dynamic space-air-ground integrated network (SAGIN), VERA obtains 55.33% +/- 3.60% success with 0.45% +/- 0.81% coverage violation, within 1.33 percentage points of a privileged-mask reference. With rewards matched over ten paired seeds, VERA improves success over the strongest baseline by 8.74 percentage points (p=0.023) and reduces violation by 52.19 percentage points (p=5.7e-8). A ten-seed 4-by-2 factorial attributes a 14.16-16.48 percentage-point gain to CGRA across handcrafted, learned, random, and latent representations; evaluation on seven unseen topologies preserves a 24.33-30.02 percentage-point advantage over multi-agent proximal policy optimization. From 10 to 40 users, success remains 50.1-53.8%, and VFR adds only 0.026 ms to a central processing unit (CPU) actor step. Cross-domain tests further identify the governing condition: counterfactual credit succeeds when candidate scores respect shared constraints and fails under incompatible reward geometries. These results establish action-conditioned feasibility as an auditable training interface and counterfactual credit as a geometry-dependent optimization mechanism.