[Paper Review] Cloud Control AGV over Rayleigh Fading Channel -- The Faster The Better

Shreya Tayade, Peter Rost+2 more|arXiv (Cornell University)|May 10, 2019

Traffic control and management16 references4 citations

TL;DR

This paper investigates the stability of cloud-controlled AGVs over Rayleigh fading channels, showing that higher AGV velocities reduce instability risk by lowering channel correlation and consecutive packet loss probability. Counterintuitively, faster movement improves reliability despite stricter outage tolerance demands, revealing a key co-design opportunity for control and communication systems in industrial automation.

ABSTRACT

This paper analyzes the stability of the control system of an Autonomous Guided Vehicle (AGV) using a central controller. The control commands are transmitted to an AGV over a Rayleigh fading channel causing potential packet drops. This paper analyzes the mutual dependencies of control system and mobile communication system. Among the important parameters considered are the sampling time of the discrete control system, the maximum tolerable outages for the control system, the AGV velocity, the number of users, as well as mobile communication channel conditions. It is shown that increasing the velocity of an AGV leads to a lower risk of instability due to the higher time-variance of the mobile channel. While this still is a 'sandbox' example, it shows the potential for a manifold co-optimization of control systems operated over imperfect mobile communication channels.

Motivation & Objective

• To analyze the mutual dependencies between control system stability and mobile communication channel conditions in cloud-controlled AGVs.
• To determine the maximum tolerable consecutive channel outages for stable AGV operation under varying velocities and sampling times.
• To investigate how AGV velocity influences channel correlation and its impact on control system stability.
• To explore the potential for co-optimization of control and communication systems in industrial automation scenarios.

Proposed method

• Modeling the AGV control system with a discrete-time sampling period $ T_s $, where control commands are sent from a central cloud controller over a Rayleigh fading channel.
• Defining the maximum tolerable consecutive outages $ n_{\text{max}} $ based on control system stability requirements and channel outage probability.
• Using the Rayleigh fading channel model to compute the probability of consecutive packet errors $ P_e(n) $, which depends on Doppler spread and sampling time.
• Analyzing the impact of AGV velocity on Doppler shift and channel correlation, which directly affects the likelihood of consecutive outages.
• Evaluating the probability of system instability $ P_{us} $ as a function of sampling time $ T_s $, AGV velocity, and trace time $ T $, using numerical simulations.
• Applying the concept of channel correlation to derive the trade-off between higher velocity (lower correlation) and stricter outage tolerance.

Experimental results

Research questions

• RQ1 How does increasing AGV velocity affect the probability of consecutive packet losses in a Rayleigh fading channel?
• RQ2 What is the maximum number of consecutive channel outages $ n_{\text{max}} $ that a control system can tolerate before becoming unstable?
• RQ3 How does the sampling time $ T_s $ influence the stability of the AGV control system under varying channel conditions?
• RQ4 Does higher AGV velocity reduce the overall probability of system instability despite stricter outage requirements?
• RQ5 What is the optimal sampling time $ T_s $ that minimizes the probability of instability for a given AGV velocity and channel correlation?

Key findings

• Higher AGV velocities reduce the probability of consecutive packet losses due to lower channel correlation, which decreases the likelihood of sustained deep fades.
• Despite requiring stricter outage tolerance, the reduction in consecutive error probability at higher velocities dominates, leading to an overall lower instability probability.
• For a given velocity, increasing the sampling time $ T_s $ initially reduces instability probability by lowering channel correlation, but beyond an optimal point, the stricter outage requirements dominate and instability increases.
• At $ T = 333 $ s, the probability of instability $ P_{us} $ reaches a minimum at $ T_s = 5 $ s, demonstrating an optimal sampling interval for stability.
• For low velocities (e.g., $ T = 500 $ s), the optimal sampling time is 6 ms, while for $ T = 1000 $ s, it increases to 7 ms, indicating that optimal $ T_s $ depends on velocity and channel correlation.
• The results show that driving faster can improve system stability by reducing the impact of channel fading, highlighting a counterintuitive but valuable co-design insight for control and communication systems.