Optimal Cross-Layer Design of Sampling Rate Adaptation and Network Scheduling for Wireless Networked Control Systems

Abstract

Wireless Networked Control Systems (NCS) are increasingly deployed to monitor and control Cyber-Physical Systems (CPS). To achieve and maintain a desirable level of performance, NCS face significant challenges posed by the scarce wireless resource and network dynamics. In this paper, we consider NCS consisting of multiple physical plant and digital controller pairs communicating over a multi-hop wireless network. The control objective is that the plants follow the reference trajectories provided by the controllers. This paper presents a novel optimization formulation for minimizing the tracking error due to (1) discretization and (2) packet delay and loss. The optimization problem maximizes a utility function that characterizes the relationship between the sampling rate and the capability of disturbance rejection of the control system. The constraints come from the wireless network capacity and the delay requirement of the control system. The solution leads to a joint design of sampling rate adaptation and network scheduling, which can be naturally deployed over existing networking systems which have a layered architecture. Based on a passivity-based control framework, we show that the proposed cross-layer design can achieve both stability and performance optimality. Simulation studies conducted in an integrated simulation environment consisting of Matlab/Simulink and ns-2 demonstrate that our algorithm is able to provide agile and stable sampling rate adaptation and achieve optimal NCS performance.

A closed-loop context aware data acquisition and resource allocation framework for dynamic data driven applications systems (DDDAS) on the cloud

“…A duality theory which decomposes the network utility maximization problem into a rate control problem and a scheduling problem is presented in another work [15]. Minimization of tracking error of control systems by transforming the problem into a utility maximization problem, where the utility function illustrates the relationship between the capacity of disturbance rejection and sampling rate, has also been investigated [16]. Another group of work specifically focuses on energy harvesting sensor networks that rely on renewable energy sources [17,18,19,20].…”

Wireless Network Design for Control Systems: A Survey

“…On the other hand, [173] derives experimental based models by using curve fitting techniques and validation through extensive experiments. An adaptive algorithm was also proposed to adjust the coefficients of these models by introducing a learning phase without any explicit information about data traffic, network topology, and Contention-based Access [212], [213], [175], [174], [214], [129], [84], [107], [215], [173], [125] [166], [202], [194], [195], [216], [196], [217] [204], [211], [206], [207], [208], [209], [210] Schedule-based Access [126], [127], [176], [177], [178] [4], [190], [203], [202], [192], [193] [81], [218], [205], [62], [83] Physical Layer Extension [6], [87], [183], [219], [220], [221] --Network Resource Schedule Scheduling Algorithm …”

Energy and Delay Constrained Maximum Adaptive Schedule for Wireless Networked Control Systems

“…If a particular t * value is infeasible, then all t < t * are infeasible and if a particular t * value is feasible, then all t > t * are feasible due to Theorem 4. Based on this finding, algorithm iteratively shrinks the range for the optimal t * in Lines (12)(13)(14)(15)(16)(17)(18)(19)(20) until the selected t * falls into the desired neighborhood of the optimal value, which is specified by the relative error bound value .…”

“…Scheduling algorithms based on these standards however are designed mostly to provide low deterministic end-to-end delay and controlled jitter for realtime traffic across a very large mesh network distributed over a large area without considering the periodic nature of the transmissions [13], [14]. On the other hand, the algorithms designed for the scheduling of periodic controller tasks running on a processor, such as Earliest Deadline First (EDF), Least Laxity First (LLF) and Deadline Monotonic (DM) scheduling algorithms [15], have been adopted for the scheduling of the direct transmission of the periodic data packets of the sensor nodes to their corresponding controllers for the case where no concurrent transmissions are allowed [16], [17], [18]. However, since they assign time slots to the tasks as soon as they are available, none of these algorithms provide any adaptivity to the packet losses.…”