Predictive over-the-air sensing and controlling under limited communication and computational resources

Abstract

Recently, the wireless networked control systems (WNCS) have become a key infrastructure technology for critical control systems due to their low cost of deployment and maintenance, flexibility in installations, and potential for making the working environment safe. To improve the scalability of WNCS under limited wireless resources, the novel research works focus on utilizing predictive sensing and controlling techniques available in machine learning (ML) in addition to the advanced communication strategies. In this work, a novel communication and computation resource aware control system scheduling algorithm for wireless and computational resource allocation for a scalable WNCS with over-the-air sensing is proposed to solve the joint problem of achieving control stability and utilizing wireless and computational resources. Therein, a two-stage Gaussian Process Regression solution is used to predict the missing state information due to the lack of resources for sensing a low-complexity GPR with low prediction accuracy and low resource consumption; and a high-complexity GPR with high prediction accuracy with an expense of high computational resource consumption. The joint problem of scheduling wireless and computational resources under strict limitations of resource availability is formulated as a stochastic optimization problem, which is solved using the Lyapunov optimization framework. As the solution, a novel predictive stability and resource-aware scheduling algorithm is proposed in the end. Finally, the performance of the proposed solution is evaluated with respect to several standard scheduling techniques.