[Paper Review] Towards Mission-Critical Control at the Edge and Over 5G

Per Skarin, William Tärneberg+2 more|arXiv (Cornell University)|Mar 6, 2018

Energy Efficient Wireless Sensor Networks4 citations

TL;DR

This paper presents a fully operational edge cloud test-bed integrating 5G, IoT, and cloud-native control systems to enable mission-critical, low-latency control applications. It demonstrates that model predictive controllers can be dynamically relocated across edge nodes with minimal jitter and latency, outperforming both on-plant and distant cloud deployments in stability and responsiveness under high noise and set-point changes.

ABSTRACT

With the emergence of industrial IoT and cloud computing, and the advent of 5G and edge clouds, there are ambitious expectations on elasticity, economies of scale, and fast time to market for demanding use cases in the next generation of ICT networks. Responsiveness and reliability of wireless communication links and services in the cloud are set to improve significantly as the concept of edge clouds is becoming more prevalent. To enable industrial uptake we must provide cloud capacity in the networks but also a sufficient level of simplicity and self-sustainability in the software platforms. In this paper, we present a research test-bed built to study mission-critical control over the distributed edge cloud. We evaluate system properties using a conventional control application in the form of a Model Predictive Controller. Our cloud platform provides the means to continuously operate our mission-critical application while seamlessly relocating computations across geographically dispersed compute nodes. Through our use of 5G wireless radio, we allow for mobility and reliably provide compute resources with low latency, at the edge. The primary contribution of this paper is a state-of-the art, fully operational test-bed showing the potential for merged IoT, 5G, and cloud. We also provide an evaluation of the system while operating a mission-critical application and provide an outlook on a novel research direction.

Motivation & Objective

• To evaluate the feasibility of deploying time-sensitive, mission-critical control applications in a distributed edge cloud environment.
• To address the limitations of centralized data centers and on-plant monolithic systems in handling latency, jitter, and system resilience.
• To enable dynamic runtime reconfiguration of control applications across geographically dispersed edge nodes without sacrificing stability.
• To assess the performance of cloud-native control systems under real-world network conditions using 5G and edge infrastructure.
• To establish a reproducible, observable test-bed for studying the interplay between control systems and distributed cloud platforms.

Proposed method

• Built a multi-tiered test-bed integrating 5G wireless access, distributed edge cloud nodes, and a physical control plant (inverted beam with ball).
• Deployed a Model Predictive Controller (MPC) as a cloud-native application, executed on edge nodes, a Raspberry Pi at the plant, and an AWS instance.
• Utilized 5G's Ultra-Reliable Low-Latency Communication (URLLC) to ensure sub-10ms end-to-end latency and low jitter for control loops.
• Implemented dynamic reconfiguration by migrating the MPC controller across edge nodes during runtime to evaluate resilience and performance.
• Measured execution times, network latency, and controller stability across different deployment points: plant, edge, and cloud.
• Used time-series data to analyze controller behavior during high-noise and set-point change scenarios, comparing feasibility and solution convergence.

Experimental results

Research questions

• RQ1 Can mission-critical control applications be reliably and responsively deployed over a 5G-enabled edge cloud infrastructure?
• RQ2 How does the performance of a Model Predictive Controller (MPC) vary when deployed at the edge, on the plant, or in a distant cloud?
• RQ3 To what extent can dynamic runtime reconfiguration of control applications be achieved without compromising system stability or introducing excessive jitter?
• RQ4 What is the impact of network latency and jitter on MPC feasibility and solution convergence in real-time control scenarios?
• RQ5 How does computational load and noise affect controller performance across different deployment locations?

Key findings

• The edge node successfully maintained control of the ball on the beam under high noise and set-point changes, while the plant-based controller failed due to computational overload.
• On the AWS cloud instance, despite high computational capacity, network latency and jitter caused repeated failures in finding feasible MPC solutions, leading to the ball falling off the beam.
• The edge node required only 80 ms to complete an MPC optimization iteration, compared to 40 ms on AWS, but its lower latency and jitter enabled stable control where AWS failed.
• At the plant, when the MPC failed to find a feasible solution, execution time exceeded the 50 ms sampling period, indicating severe computational bottlenecks.
• Dynamic reconfiguration of the controller across edge nodes was successfully achieved at runtime without loss of stability or significant performance degradation.
• The system demonstrated that edge cloud deployment enables reliable operation of latency-sensitive, mission-critical control applications, even under challenging conditions where traditional deployments fail.