Performance and reliability improvement of cyber-physical systems subject to degraded communication networks through robust optimization
Introduction
Cyber physical system (CPS) is a sophisticated system integrating communication network and physical control processes (Cao, Cheng, Chen, & Sun, 2013); in this system, the communication network serves as an interaction among different physical devices. A major difference between a traditional control system and a CPS is the use of communication network, which provides the CPS with great benefits in terms of low installation fee and easy system maintenance (Al-Dabbagh & Lu, 2010). Given the wide applications in diverse fields, including vehicular systems and transportation (Ahmadi, Abdelzaher, Jiawei, Nam, & Ganti, 2011), medical and health-care systems (Kocabas et al., 2016, Sonntag, 2016, Wrobel et al., 2015), smart homes and building (Ahmad et al., 2016, Li et al., 2011), and power grid (Farzin et al., 2016, Pan and Das, 2016, Susuki et al., 2012), CPS has generated extensive interest in recent years (Gupta and Chow, 2010, Khaitan and McCalley, 2015).
However, communication networks are not always reliable and subject to communication degradations, such as packet dropouts and transmission delays (Ghostine, Thiriet, & Aubry, 2011). As for packet dropouts, for example, a packet loss in the controller to the actuator channel prevents the actuator from receiving the latest control command. The control process consequently becomes unstable, and the CPS is unreliable (Ghostine et al., 2011). The CPS design encounters several challenges because of uncertain network degradations and interactions between two different layers, i.e., cyber level and physical layer. Therefore, numerous works have studied the CPS from different aspects to overcome these problems. Considerable literature has explored the CPS from the viewpoint of control engineering, existing control policies, such as networked predictive control (Renquan et al., 2016, Wang et al., 2016), robust control (Latrach et al., 2014, Wang et al., 2007, Wen et al., 2016, Xiao et al., 2009), and fuzzy control (Mahmoud and Almutairi, 2016, Xinchun et al., 2009), are improved and implemented by taking network delay and packet loss into consideration, in order to alleviate the impact of the network degradations on reducing system performance and reliability. Many studies have also been devoted to improving communication protocols based on the realistic application requirements from the aspect of communication network (Zhipeng, HanYong, Yuanyuan, Chen, & Jiaxuan, 2016).
In view of the degraded communication network, in most of the existing literature cited above, it has been assumed that the packet loss or delay only happening on the channel from sensor to controller (Wang et al., 2016, Xiao et al., 2009), although some work include two aspects, i.e. channel from the sensor to the controller and from the controller to the actuator, only packet loss (Renquan et al., 2016, Wang et al., 2007) or transmission delay (Latrach et al., 2014, Wang et al., 2007, Wen et al., 2016) phenomenon is considered. However, when the controller and the actuator communicate over real-time network, the transmission delay and packet dropout are inevitable in practical CPS. In the aforementioned works, the application of control strategies is only valid in specified systems, mainly in the linear system, and many assumptions are made. By contrast, the proportional-integral-derivative (PID) controller is used in most realistic control systems. The root reason is that this controller is simple and applicable in most cases, such as in the linear or nonlinear system, in addition, it can provide good performance by selecting appropriate controller parameters. Therefore, in this study, without loss of generalities, a PID controller is selected to control the physical devices via the communication network. The integrator and filter method are selected in the real hardware implementation of the discrete-time PID controller. However, these methods would introduce disturbances on the controller parameters (Pan & Das, 2016). In most existing literature, the design of PID controller for CPS subject to degraded communication network and parameters perturbation has not been adequately studied. We thus need to develop a robust PID controller against these disturbances.
The explicit reliability function of the CPS cannot be obtained because the feedback mechanism conceals the mapping between the deterioration level of communication network and the system performance. Therefore, we use a Monte Carlo simulation (MCS)-based method to evaluate the CPS reliability. The basic idea is that the CPS performance is evaluated multiple times under the same network configurations. The capability of the controller to ensure whether the system satisfies the performance criteria can be assessed in the presence of random network degradations. For a nonlinear and nonconvex optimization problem, the canonical particle swarm optimization (CPSO) can escape from being trapped in local optima (Chung-Cheng, 2015, He et al., 2013). Thus, the CPSO is applied here to search for the robust optimal parameters of the controller. Given that the objective function is stochastic, it needs to be evaluated multiple times to achieve the expectation. An archive strategy is applied here to reduce the computation expense (Kruisselbrink et al., 2010, Pan and Das, 2016).
The main contributions of this study can be summarized as follows:
- (1)
In the model of cyber physical system, both transmission delay and packet dropout over the sensor to controller channel and the controller to actuator channel are considered.
- (2)
In view of network degradations, the model of transmission delay is more stochastic rather than the one which is uniformly distributed on an interval. We use the Bernoulli distribution to describe the packet dropout and study the effect of packet dropout on the behaviors of controller and actuator.
- (3)
We develop a robust PID controller against controller parameter perturbations, introduced by real hardware implementation of the discrete-time PID controller.
- (4)
We conduct a simulation-based validation test on the CPS model to show that the proposed CPSO-based robust optimization method is effective in improving the system reliability.
The rest of this paper is organized as follows. In Section 2, the control structure of CPS is described, and the main features of packet dropout and time delay along with their effects on the control process are discussed. In Section 3, we use TrueTime toolbox to simulate a CPS, and a demonstration about the detailed design for every module is provided. In Section 4, in the presence of design variable fluctuations, we propose CPSO-based robust optimization for the performance and reliability improvement of a CPS subjected to network degradations. In Section 5, a case study carried about an industrial heat exchanger is used to illustrate the effectiveness of the proposed method. Finally, the conclusions and future work are given in Section 6.
Section snippets
Preliminary model
Block diagram is adopted here to build a feasible architecture for the CPSs, as shown in Fig. 1. The architecture consists of interdependent cyber and physical layers. In the physical layer, considering that devices are usually distributed in different areas, data exchanges among them are via a communication network in a form of data packets. The plant is a continuous-time system, and the controller is a discrete controller; thus, a digital-to-analog converter is necessary.
The sensor measures
Test platform based on TrueTime simulator
TrueTime, a MATLAB Simulink tool, is readily applied in modeling the CPS. Numerous works have used the test platform based on the TrueTime simulator to validate their frameworks (Balasubramaniyan et al., 2016, Grenier and Navet, 2008).
The CPS consists of five basic modules (see Fig. 2), i.e., sensor module, actuator module, controlled module, network module, and interference module. TrueTime Send and TrueTime Receive are selected to simulate message transmission through the network (Cervin,
Reliability assessment
In the presence of network degradation, the system output will be different even for the same control strategy. The MCS is widely used to assess the reliability of complex systems given the analytical reliability function is unavailable. The MCS technique is introduced here to generate a number of samples for real CPS, with scenarios of network degradation, and then estimate the system reliability.
When conducting the reliability assessment, the control quality should is usually defined by
System description
This section illustrates the effectiveness of the proposed robust optimization method for the CPS. An industrial heat exchanger given in Fig. 4, commonly used in the chemical industry as in Vidal and Banos (2010), is employed in this case study.
In the industrial heat exchanger system, the temperature control of the inlet fluid receives most concerns. The inlet fluid temperature should converge to a predefined value in the given time, so as to ensure a satisfied reaction environment for the
Conclusion
This work aims to improve the reliability and the performance of CPSs in the presence of network degradations. The proposed CPSO-based robust optimization method for finding a robust discrete-time PID controller is more general and thus has wider applications in nonlinear and stochastic control systems than other complicated control strategies. The hardware implementation of the discrete-time PID controller introduces controller parameter perturbation, and the system performance deteriorates if
References (41)
- et al.
Smart cyber society: Integration of capillary devices with high usability based on cyber-physical system
Future Generation Computer Systems-the International Journal of Escience
(2016) - et al.
Reliability modeling of networked control systems using dynamic flowgraph methodology
Reliability Engineering & System Safety
(2010) - et al.
Design and verification of Cyber-Physical Systems using TrueTime, evolutionary optimization and UPPAAL
Microprocessors and Microsystems
(2016) - et al.
Robust optimization – A comprehensive survey
Computer Methods in Applied Mechanics and Engineering
(2007) - et al.
Comparative studies on the high-variability embedded robust parameter design from the perspective of estimators
Computers & Industrial Engineering
(2013) - et al.
Multi-objective reliability-redundancy allocation problem using particle swarm optimization
Computers & Industrial Engineering
(2013) - et al.
Variable delays and message losses: Influence on the reliability of a control loop
Reliability Engineering & System Safety
(2011) - et al.
Multilevel redundancy allocation using two dimensional arrays encoding and hybrid genetic algorithm
Computers & Industrial Engineering
(2013) - et al.
Feedback fuzzy control for quantized networked systems with random delays
Applied Mathematics and Computation
(2016) - et al.
Handling packet dropouts and random delays for unstable delayed processes in NCS by optimal tuning of (PID mu)-D-lambda controllers with evolutionary algorithms
ISA Transactions
(2011)
Reset compensation for temperature control: Experimental application on heat exchangers
Chemical Engineering Journal
An online optimization approach for control and communication codesign in networked cyber-physical systems
IEEE Transactions on Industrial Informatics
“TRUETIME 2.0 beta—Reference manual
Robust data envelopment analysis approaches for evaluating algorithmic performance
Computers & Industrial Engineering
Particle swarm optimization
Enhancing power system resilience through hierarchical outage management in multi-microgrids
IEEE Transactions on Smart Grid
Fine-tuning MAC-level protocols for optimized real-time QoS
IEEE Transactions on Industrial Informatics
Networked control system: Overview and research trends
IEEE Transactions on Industrial Electronics
Cited by (15)
Risk propagation and mitigation of design change for complex product development (CPD) projects based on multilayer network theory
2020, Computers and Industrial EngineeringCitation Excerpt :Decision makers tend to avoid some extremely complex or expensive components when choosing the change path, but it may lead to the lag of change propagation and eventually lead to “superchange” with greater losses (Giffin et al., 2009). Reference to other research areas such as supply chain networks (Golpîra, Najafi, Zandieh, & Sadi-Nezhad, 2017; Dehghan, Nikabadi, Amiri, & Jabbarzadeh, 2018; Darestani & Hemmati, 2019), infrastructure networks (Fang, Mo, Wang, & Xie, 2017), one effective way to mitigate the design change risk propagation is improving the robustness and resilience of the CPD projects by allocating design/control resources. However, less attention is paid on the aspect of mitigation resource allocation especially in the feedback and control process when design change risks occur.
A resilient network recovery framework against cascading failures with deep graph learning
2024, Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and ReliabilityModeling two-stage failure mechanism of cascading in cyber-physical power systems
2023, Physica ScriptaDeclaration of Interests
2023, SSRNReliability of spare routing via intersectional minimal paths within budget and time constraints by simulation
2022, Annals of Operations ResearchMethodology support and algorithms for dynamic synthesis of technologies and programs for proactive management of information processes in the industrial internet under a crisis
2022, International Journal of Risk Assessment and Management