Controller and architecture co-design of wireless cyber-physical systems

Abstract

Recently, more and more factories begin to employ wireless mesh networks for process control. These wireless Cyber-Physical Systems (WCPSs) need to meet performance requirements. However, the control algorithm and system are usually designed with an isolated design flow, which leads to higher integration, testing and debugging costs and poor resource utilization. To bridge the gap, we propose a controller and architecture co-design framework based on architecture analysis and design language (AADL). Firstly, we show how to build control and architecture models of WCPSs. Then methods that translates control and architecture models to performance models are proposed. With these models, we can analyze the non-function properties of the system before implementation. Finally, Matlab/Simulink and OSATE are integrated and several plugins are developed to support the realization of this framework. The synthesis of above contributions is a largely automated co-design and analysis process of WCPSs. Comparing with other co-design frameworks, our framework provides better capability, usability, and extensibility. A case study shows that our co-design framework achieves better control performance (reducing the pulp level variations to 42.3% of the former controller).

Introduction

In the last decade, Wireless communication has been practiced in the industrial market for many years. Today more and more plants begin to use wireless mesh networks for industrial automation and control system applications [1], [2]. These systems can be considered as wireless Cyber-Physical Systems (WCPSs). The migration of wired industrial infrastructure to wireless technologies can improve flexibility, scalability, and efficiency [3]. Typical industrial wireless mesh network technologies suit for WCPSs include WirelessHART [4], ISA100 [5], [6], and WIA-PA [7], Blink [8].

WCPSs are characterized by the tight interactions between computational components, communication networks and physical dynamics. The design of WCPSs is a challenging process that usually involves the participation of control theorists and embedded system engineers. Control theorists are responsible for control algorithm design and performance analysis of controller. Embedded system engineers are responsible for system architecture design, task partition, computations and communication scheduling, timing and performance analysis etc. The control algorithm and system usually designed separately by different person. This isolated design flow leads to higher integration, testing and debugging costs and poor resource utilization.

To overcome the gap, we propose a controller and architecture co-design framework which supports the analysis and verification of system nonfunctional properties before the implementation. As shown in Fig. 1, we setup a control algorithm and structure for wireless control and use architecture analysis and design language (AADL) to model the system architecture. These models are transformed to performance models with our proposed model transformation algorithms. Parameters and constraints are also extracted. Based on these performance models, parameters and constraints, we can analyze and verify the system performance of WCPSs at architecture level. This process is carried out before system implementation, so it can overcome the drawbacks of the isolated design flow.

Related Works: (1) Wireless Control Design: Several promising control strategies, including Kalman Filters [9], [10], passivity-based control [11], event-triggered control [12], model predictive control [13], [11], and command buffer [13], [11] have been proposed to deal with the dynamics and randomness in wireless control systems. However, these works only focus on centralized wireless network. [9] discussed the close-loop control design strategy for WirelessHART networks with Kalman filters and model predictive control.

(2) Co-design Approaches of CPSs: There have been a lot of attempts to close the gap between controller design and system design and implementation in NCSs (Network Controlled Systems). [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24]. These results, however, cannot be applied to WCPSs directly. In NCSs, the processing units, sensors, controllers, and actuators communicate with each other via a shared bus, e.g., CAN or FlexRay, for which the delay of packets from sensor to controller, and from controller to actuator can be treated as uniform and easily estimated probability distributions. But for WCPS, as multi-hop wireless network is used to connect these nodes, a lot of network and physical factors including network protocol, network topology, data flow configuration, message scheduling, and channel condition strengthen the interaction between computational components, communication networks and physical dynamics. However, previous works do not fully consider the impact of these network and physical factors.

Recently, there are few works concentrating on the co-design principles of industrial WCPSs. These co-design approaches of WCPSs can be classified into two categories: interactive approaches and joint approaches [25]. Interactive design approaches tun the wireless network parameters to satisfy given constraints on the critical interactive system variables [26], [27], [28], [29]. Joint design approaches optimize the wireless network and control system parameters jointly considering their interaction through the critical system variables [30], [31], [32], [33]. In this paper, we only focus on the joint co-design approaches that support multiple hop wireless networks. Paper [30] addresses the sampling period optimization of the overall control performance with existing delay bounds for WirelessHART network, and propose four methods to solve the problem. Paper [31] jointly optimizes the sampling period, packet forwarding policy and control law for a multi-hop WirelessHART network. The objective is to minimize the closed-loop control cost subject to the energy and delay constraints of the nodes. Paper [32] proposes a framework for modeling and analyzing of WCPSs based on Mathematica. The framework jointly considers control, network topology, routing, scheduling, and communication error to the robustness of WCPSs. Paper [33] studies utility maximization problem subject to wireless network capacity and delay requirement of control system. The joint optimization method is based on embedded-loop approach. In the inner loop solves the sampling period as a relaxed problem with fixed delay bound. The outer loop then determines optimal delay bounds based on the output of the inner loop.

Contributions: The contributions of this paper include: (1) we propose a controller and architecture co-design framework of WCPSs based on AADL. With the virtual integration of AADL, the framework is able to analyze and verify the system performance before implementation, so it allows us to overcome the gap of controller design and system implementation; (2) comparing with other joint co-design frameworks that support multiple wireless network scenario, our framework provides better design ability, usability, and extensibility; (3) we show the method of deriving tool chains for co-design and analysis of WCPSs in a formal manner by using Matlab/Simulink and OSATE.

The organization of this paper is as following: Section 2 describes the problems encountered in designing of WCPSs; Section 3 shows how to model system architecture with AADL; Section 4 gives methods to translate AADL architecture model to performance models; Section 5 gives a detailed description of this co-design framework and shows how to analyze system properties; Section 6 provides an example of co-designing WCPS; Section 7 concludes this work.