Predictive Control and Communication Co-Design via Two-Way Gaussian Process Regression and AoI-Aware Scheduling

Abstract

This article studies the joint problem of uplink-downlink scheduling and power allocation for controlling a large number of control systems that upload their states to remote controllers and download control actions over wireless links. To overcome the lack of wireless resources, we propose a machine learning-based solution, where only one control system is controlled, while the rest of the control systems are actuated by locally predicting the missing state and/or action information using the previous uplink and/or downlink receptions via a Gaussian process regression (GPR). This GPR prediction credibility is determined using the age-of-information (AoI) of the latest reception. Moreover, the successful reception is affected by the transmission power, mandating a co-design of the communication and control operations. To this end, we formulate a network-wide minimization problem of the average AoI and transmission power under communication reliability and control stability constraints. To solve the problem, we propose a dynamic control algorithm using the Lyapunov drift-plus-penalty optimization framework. Numerical results corroborate that the proposed algorithm can stably control 2x more number of actuators compared to an event-triggered scheduling baseline with Kalman filtering and frequency division multiple access, which is 18x larger than a round-robin scheduling baseline.

Task-Oriented Prediction and Communication Co-Design for Haptic Communications

“…In the 5G New Radio system, the expression of the decoding error probability, ϵ d m , is given by (12). Given the queuing system in Section III, the relationship between the queuing delay bound and the queuing delay violation probability, f ϵ q m (•), is derived in (16).…”

“…Gaussian mixture regression model was used to predict haptic feedback to surgeon in a needle insertion process, where the prediction accuracy was measured by Root Mean Square Error (RMSE) between predicted force and ground truth. In [12], authors proposed a predictive actuation framework by predicting missing state at receiver using previously received state. A model-based prediction algorithm was used to minimize average age of information (AoI) and transmit power.…”

“…A model-based prediction algorithm was used to minimize average age of information (AoI) and transmit power. However, in both [11] and [12], human perception capabilities were not taken into account.…”

Predictive Closed-Loop Remote Control Over Wireless Two-Way Split Koopman Autoencoder

“…In the literature of remote control over wireless links, one central question is how to cope with communication imperfections on the control system operations. These works are often termed communication and control co-design (CoCoCo) that deal with the trade-off between wireless resource consumption and control stability [24], [25]. In this direction, dynamic sensor scheduling approaches were proposed to cope with timevarying control and channel states, thereby improving control stability and communication efficiency [26], [27].…”

“…In order to meet stringent latency and reliability requirements in the context of URLLC and time-sensitive control systems, controldynamics and CSI aware resource allocation and scheduling solutions were developed in [30], [31], which can partly relax URLLC requirements without compromising the control performance. In our prior work [25], a predictive and controlaware scheduler was proposed, in which the future states and actions are locally predicted using a Gaussian process regression approach (GPR) at the controller and actuator, respectively. Nevertheless, these works only consider linear control systems, questioning their feasibility when the system dynamics are non-linear.…”