Reinforcement Learning-Driven Edge Management for Reliable Multi-view 3D Reconstruction
Motahare Mounesan, Sourya Saha, Houchao Gan, Md. Nurul Absur, Saptarshi Debroy
TL;DR
The paper tackles reliable, latency-constrained real-time multi-view 3D reconstruction in edge environments by introducing two cooperative Q-learning agents for camera and server selection, learning online under disruptions. By formalizing reliability as $\mathbb{R}_t = 1$ when $Q_t \ge \Theta$ and $L_t \le \Phi$, and employing adaptive Q-learning to cope with non-stationarity, the approach achieves disruption-aware decisions that balance reconstruction quality and end-to-end latency. Evaluations on a lab plus FABRIC-based testbed show up to about $62\%$ camera-selection reliability and substantial server-selection gains with adaptive policies, outperforming several baselines. The results demonstrate that online, decoupled RL can deliver robust, near real-time 3D reconstructions in unpredictable edge environments with practical implications for mission-critical smart-city applications.
Abstract
Real-time multi-view 3D reconstruction is a mission-critical application for key edge-native use cases, such as fire rescue, where timely and accurate 3D scene modeling enables situational awareness and informed decision-making. However, the dynamic and unpredictable nature of edge resource availability introduces disruptions, such as degraded image quality, unstable network links, and fluctuating server loads, which challenge the reliability of the reconstruction pipeline. In this work, we present a reinforcement learning (RL)-based edge resource management framework for reliable 3D reconstruction to ensure high quality reconstruction within a reasonable amount of time, despite the system operating under a resource-constrained and disruption-prone environment. In particular, the framework adopts two cooperative Q-learning agents, one for camera selection and one for server selection, both of which operate entirely online, learning policies through interactions with the edge environment. To support learning under realistic constraints and evaluate system performance, we implement a distributed testbed comprising lab-hosted end devices and FABRIC infrastructure-hosted edge servers to emulate smart city edge infrastructure under realistic disruption scenarios. Results show that the proposed framework improves application reliability by effectively balancing end-to-end latency and reconstruction quality in dynamic environments.