PaperlayerView exact revision on arXiv

Source-linked AI summary

Wireless Network Design for Control Systems: A Survey

Pangun Park, Sinem Coleri Ergen, Carlo Fischione, Chenyang Lu, Karl Henrik Johansson

arXiv:1708.07353v1eess.SYcs.ITcs.NI

TL;DR

WNCS require joint communication–control design because wireless delay, message dropout, sampling, and energy use interact with closed-loop performance. This survey synthesizes network parameters, standards, control analysis, and optimization methods, highlighting feasibility, tradeoffs, and complexity. It concludes that existing approaches often treat only subsets of the interdependent variables, while reviewed methods span interactive tuning and joint optimization.

  • Problem

    WNCS must coordinate wireless-network and control-system design because delay, message dropout, sampling period, and energy consumption interact and affect closed-loop performance.

  • Method

    The survey reviews wireless-network parameters across protocol layers, standards, control analysis, and interactive and joint network–control design approaches for WNCS.

  • Results

    The reviewed literature shows that practical WNCS design must consider feasibility and tradeoffs among interdependent communication and control variables, with approaches differing in performance and computational complexity.

  • Takeaways & Limitations

    WNCS design requires understanding how network choices shape delay, dropout, sampling, energy use, and control performance rather than tuning these aspects independently.

Abstract

from arXiv · show

Wireless networked control systems (WNCS) are composed of spatially distributed sensors, actuators, and con- trollers communicating through wireless networks instead of conventional point-to-point wired connections. Due to their main benefits in the reduction of deployment and maintenance costs, large flexibility and possible enhancement of safety, WNCS are becoming a fundamental infrastructure technology for critical control systems in automotive electrical systems, avionics control systems, building management systems, and industrial automation systems. The main challenge in WNCS is to jointly design the communication and control systems considering their tight interaction to improve the control performance and the network lifetime. In this survey, we make an exhaustive review of the literature on wireless network design and optimization for WNCS. First, we discuss what we call the critical interactive variables including sampling period, message delay, message dropout, and network energy consumption. The mutual effects of these communication and control variables motivate their joint tuning. We discuss the effect of controllable wireless network parameters at all layers of the communication protocols on the probability distribution of these interactive variables. We also review the current wireless network standardization for WNCS and their corresponding methodology for adapting the network parameters. Moreover, we discuss the analysis and design of control systems taking into account the effect of the interactive variables on the control system performance. Finally, we present the state-of-the-art wireless network design and optimization for WNCS, while highlighting the tradeoff between the achievable performance and complexity of various approaches. We conclude the survey by highlighting major research issues and identifying future research directions.

I. INTRODUCTION

WNCS replace point-to-point wiring with wireless links in safety-critical applications, bringing flexibility and installation benefits while creating tightly coupled communication–control design challenges. This survey reviews those challenges, interactive variables, standards, and network-design approaches.

  • Motivation: WNCS connect distributed sensors, controllers, and actuators over wireless networks for applications including automotive, avionics, building, and industrial automation.Their installation and maintenance flexibility, along with potential safety benefits, support use in safety-critical systems.
  • Design challenge: Wireless transmission introduces non-zero delay and message errors that can degrade control performance, destabilize systems, and cause economic or safety consequences.Control and network designs must therefore account for delay, reliability, and message-loss tolerance.
  • Design challenge: IEEE 802.15.4 control cost generally increases with sampling period, message delay, and message-loss probability, while the feasible region depends strongly on network performance.Shorter sampling periods can increase traffic, loss probability, and delay toward values above which the system becomes unstable.
  • Survey scope: The survey reviews critical interactive variables, wireless-network parameters across protocol layers, network standards, and control analysis and design for WNCS.It also covers recent design approaches, optimizations, algorithms, and protocols.

A. Intra-Vehicle Wireless Network

Wireless control architectures are being developed across automotive, avionics, building, and industrial settings to reduce wiring and installation burdens while supporting monitoring and automation. These applications combine wireless sensing and actuation with domain-specific control requirements.

  • Automotive: In-vehicle wireless networks aim to reduce the cost, weight, and maintenance burden of extensive wiring harnesses.Vehicle wiring harnesses may contain up to 4,000 parts, weigh up to 40 kg, and extend up to 4 km.
  • Automotive: Automotive wireless applications span powertrain, chassis, and body functions, including tire-pressure monitoring and intelligent tires.These applications address vehicle energy use, handling, safety, occupant needs, and convenience.
  • Avionics: WAIC targets aircraft sensing and control while offering potential reductions in weight, maintenance costs, and wiring-related burdens.Its sensors can monitor aircraft structures and critical systems such as engines and landing gear.
  • Building automation: Wireless building automation supports occupant comfort and energy management through sensing and control of ventilation, heating, cooling, and related equipment.Wireless deployment can reduce installation costs in retrofit markets and new construction.
  • Industrial automation: Wireless sensor and actuator networks support industrial process control and factory automation, with standards including WirelessHART and ISA 100.11a.The survey notes an estimate of up to 90% cost savings versus wired field-device deployment in industrial automation.

III. WIRELESS NETWORKED CONTROL SYSTEMS

WNCS close feedback loops between physical plants and controllers through wireless sensors, actuators, and relay nodes. Their analysis uses several modeling approaches and evaluates stability and control performance under network imperfections.

  • WNCS architecture: A generalized WNCS contains multiple plants and controllers connected through wireless sensors, actuators, and relay nodes.Plant outputs are sampled, transmitted to controllers, processed into commands, and forwarded to actuators.
  • Control objectives: WNCS control objectives include stability, rapid and smooth setpoint responses, low steady-state error, limited control action, and robustness to delays and losses.Network imperfections can degrade performance or destabilize the closed loop.
  • Modeling approaches: NCS models use discrete-time, sampled-data, or continuous-time approaches depending on the controller and plant.The sampled-data approach can represent time-varying delays and sampling intervals without discretizing the continuous-time plant.
  • Stability: Stability analysis distinguishes input-output stability from internal stability, with asymptotic and exponential stability describing decay toward equilibrium.For nonlinear systems, input-output and internal stability are not necessarily equivalent.
  • Performance measures: Control cost commonly combines plant-state deviation and control effort, with Linear Quadratic formulations providing an explicit optimal policy through a Riccati equation.This cost formalizes regulation around a desired setpoint while limiting control actions.

4) Controller Design:

WNCS design must coordinate controller behavior, network timing, packet reliability, and energy use because these variables interact through the closed-loop wireless link. The survey relates controller and estimator choices to these communication effects.

  • Controller design: PID control uses proportional, integral, and derivative terms based on setpoint error and remains usable without precise plant-model knowledge.This model-light design helps explain its continued use in process control.
  • Controller design: LQR minimizes a quadratic cost subject to linear plant dynamics and produces a linear state-feedback controller.The algorithm automates finding the state-feedback controller.
  • Controller design: MPC optimizes control over a moving horizon, handles model uncertainty and hard constraints, and can accommodate missing measurements or control commands.Its predictive formulation distinguishes it from non-predictive PID and LQR controllers.
  • State estimation: State estimators predict plant states from partial measurements while compensating for measurement noise, delays, and packet losses.Kalman filtering is identified as a common estimation approach, and LQG combines estimation with LQR feedback.
  • Network interaction: Sensor traffic is often asymmetric, while battery-powered nodes make energy-efficient communication important for network lifetime.Sampling, delay, dropout, retransmissions, routing, and congestion jointly shape the interactive variables affecting control and communication performance.

A. Sampling Period

Sampling choices jointly shape control performance and wireless traffic. Higher sampling can improve control, but network-induced delay, loss, energy use, and access constraints can reverse that benefit.

  • A. Sampling Period: Time-triggered sampling uses fixed intervals, while event-triggered sampling transmits when stability or specified control performance is about to be lost.Periodic sampling simplifies analysis and design; event-triggered traffic is asynchronous.
  • A. Sampling Period: A common sampling rule selects h so that ωh lies in [0.1, 0.6].Here, ω is the desired closed-loop natural frequency and h is the sampling period.
  • A. Sampling Period: Shorter sampling periods increase network traffic, message loss probability, and message delay, eventually degrading control performance.Thus, the wired-network intuition that faster sampling is always better does not hold over wireless networks.
  • A. Sampling Period: Time-triggered transmissions permit explicit scheduling that can reduce message loss and delay, whereas event-triggered access depends strongly on plant dynamics and control-loop count.Scheduled access suits few fast plants; random access can suit many slow plants, while high-load random access degrades reliability and delay.
  • A. Sampling Period: Self-triggered control can predict event times and schedule transmissions, but handling wireless message dropouts and disorders remains challenging.Most existing event- and self-triggered studies assume these impairments do not occur.
  • A. Sampling Period: Delay variation can destabilize closed-loop systems even when mean delay is small, especially when the delay distribution has a heavy tail.Buffering can reduce variation by trading it for additional delay.

2) Communication System Aspect:

Communication-system design determines delay, dropouts, energy consumption, and reliability in WNCS. These variables arise from transmission, access, queueing, channel, and control mechanisms that interact with plant operation.

  • 2) Communication System Aspect:: Transmission delay depends on packet size and rate, while increasing transmit power raises the node’s rate but also increases interference to neighboring transmitters.Power therefore creates a direct rate-versus-interference tradeoff.
  • 2) Communication System Aspect:: Medium access delay depends on the MAC protocol, network load, retransmissions, and the nodes’ encoding, decoding, and simultaneous-transmission capabilities.Contention-based access becomes slower as busy channels and failed transmissions increase.
  • 2) Communication System Aspect:: Queueing delay is governed by message generation, multihop forwarding load, and scheduling decisions that affect packet buildup and end-to-end delay.Multihop forwarding must be included in scheduling to limit source-to-destination delay.
  • 2) Communication System Aspect:: Dropouts arise from control-algorithm discards or wireless losses, with sensor–controller and controller–actuator dropouts forming the two main channel types.Controller–actuator losses are more critical because control commands directly affect the plant.
  • 2) Communication System Aspect:: Wireless losses result from blockage, multipath, Doppler shift, interference, and time-varying shadow fading caused by environmental obstructions.The shadow-fading distribution depends on obstruction number, size, and material.
  • 2) Communication System Aspect:: Reducing sampling period, delay, or dropout improves control performance but increases communication energy consumption through more packets, higher rates, retransmission-related capability, or transmit power.The survey therefore treats control performance and network lifetime as competing design objectives.

V. WIRELESS NETWORK

WNCS commonly use IEEE 802.15.4- and IEEE 802.11-based standards, with industrial extensions adding mechanisms for timing, reliability, energy efficiency, and quality of service. IEEE 802.15.4 organizes communication through configurable superframes and channel-access modes.

  • V. WIRELESS NETWORK: WirelessHART, ISA-100.11a, and IEEE 802.15.4e build on IEEE 802.15.4, while 6LoWPAN, RPL, and 6TiSCH provide compatible IPv6-related networking for low-power lossy networks.Industrial extensions add TDMA, frequency hopping, and multipath features for more reliable and timely transmission.
  • V. WIRELESS NETWORK: IEEE 802.11 targets high throughput and continuous connectivity, while IEEE 802.11e adds enhancements for QoS in soft real-time wireless industrial communications.The survey summarizes these standards in Table I.
  • V. WIRELESS NETWORK: IEEE 802.15.4 defines physical and MAC layers and supports star and peer-to-peer topologies managed by a PAN coordinator.Nodes may communicate directly with the coordinator or with neighboring nodes.
  • V. WIRELESS NETWORK: IEEE 802.15.4 offers beacon-enabled slotted CSMA/CA with optional GTS allocation and non-beacon unslotted CSMA/CA, organized into temporal superframes.The beacon-enabled coordinator synchronizes nodes and separates active communication from low-power inactive periods.
  • V. WIRELESS NETWORK: The superframe uses beacon order BO and superframe order SO, divides the active period into CAP and optional CFP, and reserves GTS bandwidth for time-critical frames.CAP uses slotted CSMA/CA for non-time-critical data and GTS requests.

3) ISA-100.11a:

IEEE 802.15.4e provides MAC modes for industrial applications with delay and reliability constraints, including TSCH, DSME, and LLDN. These modes differ in scheduling, topology, synchronization, and implementation complexity.

  • IEEE 802.15.4e: IEEE 802.15.4e defines TSCH, DSME, and LLDN as its three major MAC modes for industrial applications.The standard was released to address delay and reliability constraints.
  • TSCH: TSCH combines time-slotted access, multichannel operation, and channel hopping to schedule collision-free links and support concurrent transmissions.Nodes synchronize through periodic slotframes and Enhanced Beacons.
  • TSCH: TSCH does not specify how to derive an appropriate link schedule.This leaves schedule derivation outside the standard.
  • DSME: DSME supports deterministic-delay multihop links through distributed beacon and GTS scheduling, but its complexity has prevented complete implementation.Current studies are limited to single-hop or cluster-tree networks rather than mesh topologies.
  • LLDN: LLDN targets very low-latency industrial automation in star topologies and can use multiple transceivers for simultaneous communication on different channels.Its short frames and group acknowledgments allow shorter time slots than TSCH.

5) 6LoWPAN:

This section describes routing and MAC mechanisms relevant to low-power wireless networks and industrial control. It contrasts flexible but unpredictable contention access with scheduled or QoS-aware alternatives.

  • Routing: RPL is an IPv6 routing protocol for low-power and lossy networks that uses destination-oriented directed acyclic graphs.It targets delay, reliability, and availability requirements in critical applications.
  • 6TiSCH: 6TiSCH integrates 6LoWPAN, RPL, and IEEE 802.15.4 TSCH while using 6top to manage schedules and monitor network performance.6top supports centralized and distributed scheduling.
  • 6TiSCH: Distributed 6top scheduling can reallocate poorly performing soft cells, providing an interference-avoidance mechanism.On-the-fly scheduling uses local information to update resources.
  • IEEE 802.11: 802.11 DCF uses random exponential backoff, producing unpredictable delays that can cause periodic control packets to miss deadlines under congestion.This behavior limits its suitability for real-time networked control.
  • IEEE 802.11: PCF and HCCA provide contention-free access, while EDCA differentiates traffic through access categories and shorter waiting parameters for higher priorities.PCF is optional and not widely implemented, whereas QoS-enabled standards do not explicitly address control deadlines.

B. Wireless Network Parameters

Wireless network parameters shape the distributions of delay, dropout, and energy consumption, so network design must account for their effects on control performance. The survey organizes these parameters across physical, MAC, and routing layers.

  • Network requirements: Existing QoS-enabled wireless standards do not explicitly consider WNCS deadline requirements, leading to unpredictable performance.Bandwidth must be allocated to high-priority sensing and actuation traffic with specific deadlines.
  • Physical layer: Transmit power and rate determine receiver SINR requirements, while higher rates require higher SINR thresholds.Receiver decoding capability and interference conditions also affect the applicable SINR criteria.
  • MAC layer: MAC protocols are classified as contention-based, schedule-based, or hybrid, with parameters controlling delay, loss probability, reliability, and energy consumption.Contention-based access exposes backoff settings, while schedule-based access assigns transmission resources.
  • Energy and scheduling: Low-duty-cycle mechanisms reduce energy use by cycling nodes between sleep and listening states, while scheduled protocols allow radios to sleep when inactive.Dynamic-priority scheduling has higher overhead but adapts priorities over time and performs better.
  • Routing: Routing must jointly support reliability, real-time forwarding, and energy efficiency in large-scale WNCS.Graph routing provides multiple paths, while RPL objective functions select routes using QoS-related metrics.
  • Control-system interaction: Control analysis defines network requirements and supports controller design, while joint treatment of sampling, delay, and dropout remains difficult because of complexity and tradeoffs.Analytical requirements can be conservative, motivating tighter bounds.

A. Time-Triggered Sampling

Time-triggered WNCS analysis distinguishes hard and soft sampling periods and models delay and dropout through stochastic, switched, or Markov approaches. These methods derive stability-related network requirements but face modeling and computational limitations.

  • Time-triggered sampling: Hard sampling discards messages not transmitted within one sampling period, whereas soft sampling continues transmission beyond that period.The classification is based on the relationship between sampling period and message delay.
  • Message dropout: Message dropouts are modeled with either unbounded or bounded numbers of consecutive losses.The bounded model imposes a limit on consecutive dropouts, although strict limits may be unreasonable for wireless links.
  • System modeling: Markov jump and switched-system models characterize control under message dropout and delay, but Markov models require statistically independent or simply modeled network transitions.These approaches can provide less conservative requirements than switched systems under suitable assumptions.
  • Model limitations: Some stability criteria assume fixed sampling and delay, eventual successful transmission, or Bernoulli losses, limiting direct applicability to wireless networks.Wireless link reliability is correlated over time and space rather than generally Bernoulli.
  • Joint design: Numerically quantifying control cost over sampling, dropout, and delay can identify feasible network requirements but has high computational complexity.This complexity is a major drawback for joint wireless-network and controller design.
  • Event-triggered sampling: Event-triggered control generates traffic when the plant state crosses a threshold, reducing traffic load but requiring continuous sensing and creating randomly generated network traffic.Its network design must therefore accommodate traffic that is not periodic.

C. Comparison Between Time- and Event-Triggered Sampling

The survey compares time-triggered and event-triggered sampling in networked control, emphasizing that their relative performance depends on the network protocol and topology. Event-triggered approaches can reduce network utilization, but collisions and scaling effects may reverse that advantage.

  • Comparison and tradeoffs: Event-triggered control often reduces network utilization, but random-access implementations can suffer when many control loops share the network.The survey identifies this as a central comparison issue across sampling and channel-access mechanisms.
  • Comparison and tradeoffs: In large-scale NCSs using pure ALOHA, collision-induced packet losses drastically reduce event-triggered control performance.The cited comparison finds time-triggered control superior in this setting.
  • Comparison and tradeoffs: Slotted ALOHA improves event-triggered control cost relative to pure ALOHA, but time-triggered control still performs better.The comparison concerns the tradeoff between delay and loss for event-triggered control.
  • Comparison and tradeoffs: The survey concludes that no general performance ranking exists between time-triggered and event-triggered sampling because outcomes depend on network protocol and topology.This dependence motivates joint consideration of control and wireless-network design.
  • Scheduling implications: SSF significantly decreases event-triggered components’ maximum delay compared with EDF and, with retransmissions, reduces average missed deadlines per unit time.SSF minimizes the maximum total active length of subframes, while retransmissions provide time diversity against lost packets.
  • Wireless-network design: IEEE 802.11n parameter selection can ensure deterministic real-time behavior, with suitable MIMO configurations improving reliability at the expense of throughput.The result is supported by theoretical analysis and experiments.

2) Network Resource Schedule:

Network-resource scheduling assigns transmissions to satisfy delay and reliability constraints under varying sampling periods, deadlines, and link conditions. The surveyed methods span dedicated or shared schedules, fixed or dynamic priorities, retransmission planning, and multipath routing.

  • Scheduling approaches: Dedicated scheduling reserves slots for specific packets, whereas shared scheduling lets packets share scheduled slots to improve reliability.Both approaches are evaluated for common-deadline transmission problems.
  • Scheduling approaches: Scheduling sensor-to-access-point transmissions for a common deadline is NP-hard, motivating algorithms that provide packet-delivery-time upper bounds.The formulation accounts for many-to-one transmission characteristics.
  • Priority policies: Fixed-priority and dynamic-priority methods address networks with varying sampling periods and deadlines.WirelessHART analyses cover fixed-priority scheduling, EDF, and dynamic-priority policies.
  • Priority policies: Dynamic-priority scheduling of periodic deadline-constrained WirelessHART flows is NP-hard; EDF outperforms fixed-priority scheduling in real-time performance.A branch-and-bound optimum and a faster conflict-aware least-laxity heuristic are described, but the heuristic offers no timely-delivery guarantee.
  • Robustness enhancement: Retransmission mechanisms use known schedules and link-quality statistics to allocate shared or separate slots while respecting delay bounds.Network-layer methods limit retransmissions according to deadlines and already guaranteed delays.
  • Routing: Multipath routing methods are classified as disjoint-path, graph, controlled-flooding, or energy/QoS-aware routing.These categories target improved wireless-network reliability and energy efficiency.

3) Network Routing:

WNCS routing methods trade reliability against energy consumption and network lifetime. Joint design also tunes sampling, access, routing, and traffic parameters because communication conditions affect control cost.

  • Routing reliability: Disjoint and graph routing use multiple paths to improve resilience and worst-case reliability against node or link failures.Node-disjoint paths share no relay nodes, whereas link-disjoint paths share no links; graph routing standards use multiple paths without explicitly defining path-construction mechanisms.
  • Energy-aware routing: Graph-routing optimization maximizes network lifetime for a fixed connectivity graph and node battery capacities while preserving high routing reliability.The optimization is NP-hard; integer programming and greedy heuristic approaches provide suboptimal solutions with significant lifetime improvement.
  • Controlled flooding: Controlled flooding improves tolerance to topology changes through redundant paths and flooding, while distributed node-list generation reduces gateway workload.REALFLOW targets industrial applications with stringent reliability requirements, and its flooding schedule can incorporate received signal strength.
  • Energy/QoS-aware routing: Energy/QoS-aware routing jointly considers application requirements and energy consumption to maximize network lifetime under industrial reliability and delay constraints.Breath jointly optimizes randomized routing, MAC, and duty cycling, minimizing energy subject to packet reliability and delay requirements; EARQ tracks energy, reliability, and delay metrics.
  • Joint design: Joint design optimizes control and network parameters, including sampling, transmission, MAC, duty cycling, routing, and traffic generation, but its complexity motivates layered abstractions.Time-triggered approaches are classified into contention-based access, schedule-based access, and routing or traffic-generation control.
  • Sampling-period tradeoffs: Realistic wireless conditions make shorter sampling periods nonoptimal because higher traffic loads increase packet loss and delay, while longer periods can stabilize throughput and control cost.Under an ideal network, control performance improves monotonically with sampling period; for longer periods, sampling becomes more influential than loss and delay.

2) Event-Triggered Sampling:

Event-triggered and self-triggered methods adapt transmissions to plant errors or predicted events, reducing unnecessary communication while addressing shared-network control constraints. The literature models access contention, losses, stability, energy, and global resource allocation jointly.

  • Event-triggered control: Event-triggered control transmits when plant-output level crossings occur, with packet losses modeled probabilistically to analyze communication–control tradeoffs.Extensions use multidimensional Markov chains and combine retransmission models with closed-loop performance analysis.
  • Random access: Over multichannel slotted ALOHA, control loops separately decide whether to transmit and which channel to select, adapting error thresholds to network resources.Each loop makes a local threshold-based transmission decision and selects an available channel uniformly when transmitting.
  • Shared-channel stability: Control-aware random-access policies derive sufficient conditions intended to preserve the stability of other control loops sharing the wireless channel.The analysis assumes packet losses caused by interference between simultaneous transmissions.
  • Contention management: CSMA-based approaches model interactions between event-triggering and contention resolution, while error-dependent scheduling combines deterministic blocking with probabilistic access.Lower-error requests can be blocked below predefined thresholds before remaining requests receive probabilistic medium access.
  • Distributed adaptation: Distributed self-regulating event triggers adapt communication parameters and control gains to meet global control-cost and resource constraints.The design uses distributed optimization and adaptive MDP techniques, with a dual price mechanism adjusting thresholds according to total transmission rate.
  • Self-triggered and mixed sampling: Self-triggered sampling predicts future level crossings, allowing sensors to sleep and reducing energy consumption and contention delay.Mixed self-triggered and event-triggered schemes combine prediction with event-based activation; MPC-based approaches jointly choose control signals and waiting times.

VIII. EXPERIMENTAL TESTBEDS

The survey presents WNCS testbeds and simulation infrastructure for evaluating wireless protocols, control systems, and their interaction. These platforms include campus WSN deployments, an HVAC building testbed, and a coupled-tank apparatus.

  • Testbed and simulation motivation: Small-scale WNCS experiments often cannot capture delays and losses in realistic large wireless networks, motivating representative testbeds and integrated simulation tools.The WCPS federated architecture combines Simulink for plant and controller dynamics with TOSSIM for wireless-network simulation and can use experimental wireless traces.
  • WSN testbed: The Washington University WSN testbed uses a server network manager and TinyOS protocol stack on TelosB nodes equipped with MSP430 microcontrollers and IEEE 802.15.4-compatible CC2420 radios.The deployment is shown across Bryan Hall and Jolley Hall; the testbed contains 79 nodes.
  • Building automation testbed: The KTH HVAC testbed spans five rooms with sensors and actuators for HVAC control, plus corridor and outdoor sensing.The system supports experiments involving indoor and outdoor temperature, humidity, CO2, light, occupancy, and door or window events.
  • Building automation testbed: The HVAC architecture lets users design experiments through LabVIEW, connect remotely to the testbed, and download experimental data through a web browser.A database logs HVAC data in real time, while the experimental application interacts with data logging and supervisory control modules.
  • Process-control testbed: The coupled-tank apparatus demonstrates process-control experiments using a pump, water basin, and two uniform-cross-section tanks.The setup represents liquid-level and inter-tank-flow control problems relevant to process industries.
  • Open challenges: The survey identifies unresolved challenges in WSN and NCS research after reviewing existing results and experimental platforms.This motivates the presentation of further research problems beyond the testbeds described in the section.

A. Tradeoff of Joint Design

The survey frames WNCS design as a tradeoff between control performance, network cost, and design complexity. It identifies open challenges involving system dynamics, communication abstractions, correlated losses, energy constraints, and emerging low-latency applications.

  • Joint communication-control design can improve control performance but increases design complexity, creating scalability and tractability concerns.The survey emphasizes quantifying both performance benefits and complexity costs.
  • Adaptive sampling can improve control performance, but its benefit depends on control-system dynamics and may increase stability and implementation overhead.Real-time adaptation is more relevant for fast dynamics and may add complexity for slow systems.
  • Feasible message-delay and loss requirements can satisfy control cost while producing different energy-consumption and robustness costs.The survey calls for tools that compare network requirements and their resulting design costs.
  • Joint-design approaches need efficient communication abstractions and should incorporate variable transmit power, rate, and time slots to reduce energy consumption.Existing approaches often assume constant physical-layer power and rate, although variability has been shown to improve communication energy consumption.
  • Simplified Bernoulli packet-loss models overlook the temporal and spatial correlation of wireless links, including the effect of consecutive losses on control performance.The survey identifies integrating packet-loss dependencies as an open requirement for interactive and joint design.
Loading 1708.07353v1…