Networked Control
Emerging networked control systems such as robotics, smart energy grids, sensor networks, and autonomous vehicular systems are all characterized by multiple feedback control loops that are closed over a communication channel. The communication infrastructure is shared among those feedback control loops and potentially other applications. This poses novel challenges for the communication and control system design to support such coupled systems with stringent real-time requirements.
Networked Control Systems are inherently very complex to be analyzed due to, first, their large scale, and second because of couplings between local control loops resulting from either physical interconnections or shared communication medium. To take these challenges into account, there has been a shift of paradigm from classical centralized control to distributed control. The employment of distributed control in such systems is motivated by: splitting computational load between distributed entities, decreasing the amount of communication exchange, information privacy, robustness against single point of failure, geographical distribution of computing resources and limited computational power of existing devices. Thus, our research focuses on developing novel concepts of distributed control for networked and cyber physical systems such that the real time requirements of both control systems and the communication system are satisfied.
Current topics:
- Optimal Co-Design of Wireless Resource Management and Multi-Loop Networked Control
- Large-scale Cyber-Physical Systems with Resource Constraints
- Optimal Information-constrained Control of Distributed Systems
- Cost-Efficient Consensus of Networked Multi-Agent Systems
Related projects:
Current topics:
Optimal Co-Design of Wireless Resource Management and Multi-Loop Networked Control
Researcher: Dr.-Ing. Mohammad H. Mamduhi
Description:
Emerging cyber-physical systems (CPS) such as smart industrial production lines, smart energy grids, and autonomous vehicular systems are characterized by multiple feedback control loops that are closed over a shared communication channel. This poses new challenges for the communication and control system design to support such networked control systems (NCS) with stringent real-time requirements. Network design has to move from traditional throughput-oriented optimization of network resources to real-time orientation to support NCS. Control systems in turn need to become aware of the changing conditions and opportunities of the network infrastructure. A strictly separate design is known to lead to high conservatism and thus to a low quality of control and to low efficiency and high cost in the context of resource usage. So, future CPS design needs to address control and networking jointly to efficiently fulfill the tight control performance requirements. Today, there exists no systematic approach for the joint design of the control and communication protocol. The objective of this project is to develop a framework for the co-design of communication and control, concretely for the optimal co-design of the wireless resource management in terms of medium access control and networked multi-loop control.
Research challenges:
- deep understanding of the relationship between the achievable control performance, communication network parameters and control system parameters based on analytic models
- novel optimal control and scheduling designs under resource constraints
- novel mechanisms for medium access in multi-hop wireless networks with real-time requirements
- optimal layering architecture and co-design for wireless communication and control over wireless multi-hop networks
Approach:
We take the scenario of multiple independent control loops accessing one shared wireless communication channel as our starting point. In particular, we consider control systems with heterogeneous linear time invariant processes and a typical wireless network medium access control based on a slotted aloha principle. We investigate the wireless network resource management and control performance trade-off, propose new models for a joint consideration and explore their fundamental limits in terms of scale and performance. Based on the decomposition of the overall optimization problem, we will derive a novel approach for the control and scheduling design as well as new network resource management schemes to support a co-design. We will provide a first optimal system co-design through the solution of the decomposed optimization problem. A key focus is put on finding fundamental results. Preliminary findings of a joint work of the two proposers’ collaborating research groups suggest a high benefit from the joint modeling and co-design strategies.
Expected results:
- fundamental joint modeling and analysis of control processes & medium access mechanisms
- analysis and decomposition of the optimization problem
- design of event-triggered control with awareness of communication medium
- design of new control-aware resource allocation mechanisms
- control and communication co-design based on joint optimization variable set
Selected publications
- Try-Once-Discard Scheduling for Stochastic Networked Control Systems. International Journal of Control, 2018, 1-15 more… BibTeX
- Prioritized contention resolution for random access networked control systems. Conference on Decision and Control (CDC), IEEE, 2017, 6689-6695 more… BibTeX
- Adaptive Decentralized MAC for Event-triggered Networked Control Systems. 19th International Conference on Hybrid Systems: Computation and Control (HSCC) , 2016 more… BibTeX
- Decentralized Event-triggered Medium Access Control for Networked Control Systems. 55th IEEE Conference on Decision and Control (CDC), 2016 more… BibTeX
Large-scale Cyber-Physical Systems with Resource Constraints
Researcher: Precious Ugo Abara
Motivation
This project will consider large scale distributed energy systems that are characterized by a high number of interacting components. Control of such systems over communication networks is an important topic with many application domains such as infrastructure systems and smart grids. Efficiency and reliability requirements necessitate a tighter real-time coordination of the individual elements with wireless sensing and actuation providing the supporting technologies. Traditionally, a control algorithm is realized on an embedded platform by performing updates on sensor/control/actuator data at a fixed (often high) sampling rate. Recent research shows considerable computational and communication resource savings by updating the control (sensing/actuating) data only at the occurrence of certain events. This is known as event-triggered control. The research focus is on the design of efficient and scalable distributed event-triggered control and scheduling algorithms for large-scale systems under resource constraints. One key challenge is the simultaneous consideration of physical constraints (through the dynamic interaction of the sub-systems) and cyber-constraints (resulting from limited communication and computation resources) which is the case in large-scale systems.
Research questions
The research focus is on the design of efficient and scalable distributed control and scheduling algorithms for large-scale CPS with resource constraints, which represents a widely open research problem. Modelling and design of the various aspects in these domains involves many challenges, some of which will be addressed by this research proposal:
- distributed control strategies such that the overall system maintains stability or meets/optimizes performance,
- control strategies that can cope with asynchrony between local loops,
- event-triggered control and scheduling of multiple event-triggered loops, aiming to reduce expensive and time consuming information exchange between control loops.
Approach: In our current set-up (see Figure) we investigate:
- a physical graph with interconnections representing coupling (possibly large-scale),
- a control graph, where each control unit bases its decision on distributed information,
- a scheduler attach to each plant that decides whether or not to send most recent measurement the its local controller,
- a shared communication channel between all agents of the network (capacity constraints or minimization of expected communication cost).
Key results and achivements:
- Derivation of a necessary and sufficient condition on the control network such that the separation property can be guaranteed, i.e. we give conditions and constraints on the information sharing pattern between the controllers of the subsystems such that we can apply the separation property in solving the optimal control problem.
- Derivation of sufficient condition for ergodicity of physically coupled NCS with linear dynamics when the interconnections are described by a DAG. We show that each subsystem has to satisfy a local condition in order to be ergodic.
Selected publications
- Quadratic Invariance for Distributed Control Systems with Intermittent Observations. 57th IEEE Conference on Decision and Control (CDC), 2018 more… BibTeX
- Information-constrained Optimal Control of Distributed Systems with Power Constraints. European Control Conference (ECC), 2018 more… BibTeX
- A Decentralized Consistent Policy for Event-triggered Control over a Shared Contention-based Network. 57th IEEE Annual Conference on Decision and Control (CDC), IEEE, 2018 more… BibTeX
- An Optimal LQG Controller for Stochastic Event-triggered Scheduling over a Lossy Communication Network. 7th IFAC Workshop on Distributed Estimation and Control in Networked Systems (NecSys), IFAC, 2018 more… BibTeX
- Separation in Coupled Event-Triggered Networked Control Systems. 20th World Congress The International Federation of Automatic Control, IFAC, 2017 more… BibTeX
- Separation Principle in Event-triggered Interconnected Networked Control Systems. International Symposium on Networked Cyber-Physical Systems 2016 more… BibTeX
Optimal Information-constrained Control of Distributed Systems
Researcher: Vedad Causevic
Motivation
The control systems are typically developed independently from the architecture of the underlying communication system. Also, current control systems’ design has not yet explored the possibility of in-network processing and active network components, since this is not yet a feature of today’s communication systems. In our research ''in-network control'' is a novel paradigm to be developed by pushing (parts of) the control as close to the sensors and actuators as possible, resulting in reduction of unnecessary unreliability and time-delay within control loop.
Research questions
- A novel framework for control partitioning on a given communication and computation infrastructure including appropriate analysis tools for performance evaluation
- Characterization of communication infrastructures resulting in a nicely decomposable control task
- Modeling of network infrastructure with active computational elements using graph theory
Approach
- As we need to gain a fundamental understanding how the spatial delay and loss distribution among the links of the communication infrastructure affect the stability and other control performance , in our approach, it is assumed that a controller is partitioned into components which are implemented on the communication infrastructure, i.e. within the cloud, which is modeled by a delay and reliability graph.
Key results and achievements
- A framework for power-constrained optimization based on information decomposition is developed
- Algorithm for implementation of such controller within a cloud
- Mapping of SDN (Software-defined networking) and in-network concepts to appropriate control problems
Selected publications
- Finite-Time Distributed Topology Design for Optimal Network Resilience. IET CONTROL THEORY AND APPLICATIONS, 2019 more… BibTeX
- Human-Robot Team Interaction Through Wearable Haptics for Cooperative Manipulation. IEEE Transactions on Haptics, 2019 more… BibTeX
- Age-of-Information vs. Value-of-Information Scheduling for Cellular Networked Control Systems. International Conference on Cyber-Physical Systems, 2019 more… BibTeX
- Distributed Link Removal Using Local Estimation of Network Topology. IEEE Transactions on Network Science and Engineering, 2018 more… BibTeX
- Try-Once-Discard Scheduling for Stochastic Networked Control Systems. International Journal of Control, 2018, 1-15 more… BibTeX
- Optimal LQG Control under Delay-dependent Costly Information. IEEE Control Systems Letters 3 (1), 2018, 102-107 more… BibTeX
- Bounded Consensus of Linear Multi-Agent Systems with External Disturbances Through a Reduced-Order Adaptive Feedback Protocol. 37th Chinese Control Conference (CCC), IEEE, 2018 more… BibTeX
- Distributed Topology Manipulation to Control Epidemic Spreading over Networks. IEEE Transactions on Signal Processing 67 (5), 2018, 1163-1174 more… BibTeX
- LQG Control via Wireless Sensor Networks with Minimal Transmission Power. Workshop on Discrete Event Systems (WODES), 2018 more… BibTeX
- Information-constrained Optimal Control of Distributed Systems with Power Constraints. European Control Conference (ECC), 2018 more… BibTeX
- State-dependent Data Queuing in Shared-resource Networked Control Systems. 57th IEEE Annual Conference on Decision and Control (CDC), IEEE, 2018 more… BibTeX
- Covariance-Based Transmission Power Control for Estimation over Wireless Sensor Networks. European Control Conference (ECC), 2018 more… BibTeX
- Energy management in a multi-building setup via distributed stochastic optimization. American Control Conference (ACC), 2018, 2018 more… BibTeX
- Towards In-Network Industrial Feedback Control. ACM SIGCOMM, 1st Workshop on In-Network Computing (NetCompute 2018), 2018 more… BibTeX
- Event-Triggered Output-Feedback H∞ Control with Minimum Directed Information. Proceedings of the Conference on Decision and Control (CDC), 2017 more… BibTeX
- Value of Information in Minimum-Rate LQG Control. In Proceedings of the International Federation of Automatic Control World Congress, 2017 more… BibTeX
- Prioritized contention resolution for random access networked control systems. Conference on Decision and Control (CDC), IEEE, 2017, 6689-6695 more… BibTeX
- Event-Triggered State Estimation: An Iterative Algorithm and Optimality Properties. IEEE Transactions on Automatic Control 62 (11), 2017, 5939-5946 more… BibTeX
- On Three-dimensional Formation Control with Mismatched Coordinates. IEEE Transactions on Control of Network Systems 5 (3), 2017, 1492-1502 more… BibTeX
- Optimal Stationary Self-Triggered Sampling for Estimation. In Proceedings of the Conference on Decision and Control (CDC), 2016 more… BibTeX
- Optimal Information Control in Cyber-Physical Systems. IFAC Workshop on Distributed Estimation and Control in Networked Systems, 2016 more… BibTeX
Cost-Efficient Consensus of Networked Multi-Agent Systems
Researcher: Xianwei Li
Motivation
Multi-agent systems have potential applications in distributed optimization, unmanned vehicles, smart grids, sensor networks, animal flocking and swarming, etc., and thus have received considerable attention in the past years. Since each agent usually has limited energy and communication capacity, controllers for multi-agent systems (usually termed protocol) are expected to distributed (i.e., communicate between neighboring agents) and cost-efficient (i.e., easy to implement and less frequent communication). For the latter, there are basically two ways to achieve more cost-efficient: one is to increase spatial sparsity (e.g., reduce communication links and reduced-order protocols) and the other is to increase temporal sparsity (e.g., reduce sampling and communication frequencies).
Research questions
- Is it possible to remove controller interaction in a protocol while only retaining information exchange about agents? When does such a protocol exist?
- How to deal with uncertainties in agents for protocols with/without controller interaction?
- How to tractably parameterize reduced-order protocols under relative information sensing?
- How to increase sampling frequencies in event-triggered consensus protocols?
- How to design protocol gains without computing the graph Laplacian?
Approach
- Robust control theory such as H_{∞} theory and Small-Gain Theorem is employed to deal with the removal of controller interaction in protocols.
- Lyapunov method is applied to improve robustness when uncertainties are encountered.
- Adaptive mechanisms are introduced to compute protocol gains without knowing the graph Laplacian.
- Event-triggered mechanisms are designed to increase sampling/communication frequencies between agents.
Key results and achievements
- A unified robust control approach has been proposed to design protocols without controller interaction for both homogeneous and heterogeneous multi-agent systems. Existence conditions are established from the aspects of poles of agents, zeros of agents and/or the communication graph, respectively.
- A novel reduced-order protocol has been proposed and a tractable parameterization has also been presented. Compared with the existing reduced-order ones, the proposed protocol requires relative information between agents only.
Selected publications
- Fully Distributed Consensus Control for Linear Multi-Agent Systems: A Reduced-Order Adaptive Feedback Approach. IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, 2019 more… BibTeX
- Output-feedback protocols without controller interaction for consensus of homogeneous multi-agent systems: A unified robust control view. Automatica 81, 2017, 37-45 more… BibTeX
- Output-feedback H∞ consensus of linear multi-agent systems over general directed graphs. 2017 13th IEEE International Conference on Control & Automation (ICCA), IEEE, 2017 more… BibTeX
- Design of output-feedback protocols for robust consensus of uncertain linear multi-agent systems. 2017 American Control Conference (ACC), IEEE, 2017 more… BibTeX