English

Learning When to Act: Communication-Efficient Reinforcement Learning via Run-Time Assurance

Machine Learning 2026-05-14 v1 Robotics

Abstract

Safe reinforcement learning (RL) typically asks what\textit{what} an agent should do. We ask when\textit{when} it needs to act, and show that a single policy can jointly learn control inputs and communication-efficient timing decisions under a pointwise Lyapunov safety shield. We focus on stabilization around a known equilibrium, where CARE-based LQR backups, Lyapunov certificates, and classical Lyapunov-STC are well defined, enabling clean comparison against analytical baselines. A run-time assurance (RTA) layer overrides the policy via a one-step-ahead Lyapunov prediction and a precomputed LQR backup, providing a strictly stronger guarantee than constrained MDP methods that enforce safety only in expectation. On an inverted pendulum, cart--pole, and planar quadrotor, the learned policy achieves 1.91×1.91\times, 1.45×1.45\times, and 3.51×3.51\times higher mean inter-sample interval (MSI) than a Lyapunov-triggered baseline; a fixed LQR controller at the same average rate is unstable on all three plants, showing that adaptive timing, not a lower average rate, makes sparsity safe. A CARE-derived Lyapunov reward transfers across environments without redesign, with a single weight wcw_c controlling the stability--communication tradeoff; ablations confirm the RTA shield is essential, with its removal reducing MSI by 1.271.27--1.84×1.84\times and degrading state norms. A preference-conditioned extension recovers the full tradeoff frontier from one model at 211\tfrac{2}{11} of training compute, and SAC experiments show the results are algorithm-agnostic across discrete and continuous domains. A 12-state 3D quadrotor case study extends the framework to higher-dimensional systems where classical STC is intractable, and robustness to ±30%\pm30\% mass variation and disturbances shows graceful degradation, with the RTA absorbing what the learned policy cannot.

Keywords

Cite

@article{arxiv.2605.12561,
  title  = {Learning When to Act: Communication-Efficient Reinforcement Learning via Run-Time Assurance},
  author = {Adam Haroon and Erick J. Rodríguez-Seda and Cody Fleming and Tristan Schuler},
  journal= {arXiv preprint arXiv:2605.12561},
  year   = {2026}
}

Comments

27 pages, 6 figures

R2 v1 2026-07-22T07:08:27.253Z