Tasks, deliverables, and milestones
Task 2.1 Architectures and methodologies for feedback control over wireless sensor and actuator networks [CNRS, EECI, ETHZ, INRIA, RUB, RUG, TUB, TUDELF, TUE (lead), UKS, UNIPD, UNIPI, UNITN, UNIVAQ, TUDO, US]:
When a control system is implemented over a network, we must take into consideration communication imperfections such as quantization errors in the signals transmitted over the network, packet drops, variable sampling/transmission intervals, variable communication delays, as well as communication constraints caused by multiple nodes sharing the network and the fact that only one node may be allowed to transmit its packets per transmission (scheduling protocols determine which node is updated at each network access). We propose to investigate the design of control algorithms that are resource aware, i.e., that can perform within a given set of specifications even in presence of non-ideal communication networks. Recently proposed hybrid model predictive control strategies tailored to wireless sensor feedback will be investigated. In all the control approaches, the importance of state observation cannot be overemphasized. In particular, observers that consider network-induced imperfections such as packet drops, variable sampling/transmission intervals, and variable communication delays will be developed.
Further we propose to develop control algorithms that are optimized with respect to the characteristics of the implementation platform. In particular, we will investigate the following questions: 1) is the characterisation of the system describing the platform computationally affordable in practical situations?, 2) can the characterization be made “adaptive” (i.e., the application estimates the platform while it executes), 3) can these techniques co]exist with platform level adaptation of the scheduling parameters, 4) can the workload generated by event triggered (or state triggered) control applications be characterized to determine if they can be safely executed with other applications?
Finally, to allow systematic methods for co-design of control algorithms and network configuration (e.g. defining scheduling, routing, transmission power, etc), a clear formulation of the effect of network configuration on control performance is needed. We propose approaching control design over networks by providing a formal mathematical framework, where control algorithms can be synthesized in conjunction with constraints coming from the communication protocol parameters as well as from software and hardware at the implementation level, which can also take into account non-idealities coming from the communication network. Such a general formulation will unavoidably turn into technical difficulties (e.g. undecidability) in finding a closed solution for the co-design problem: we will address this problem using abstraction and automatic verification and design tools coming from the computer science scientific community.
Task 2.2 Network resource allocation and adaptation for control systems [KTH, RUG, UNIPI, UNITN (lead), UNIVAQ]:
The increasing emphasis on the control components embedded in high-end systems (e.g., automotive systems) and severe cost limitations are triggering an important paradigm shift: hardware resources are aggressively utilised and shared amongst multiple tasks. In this context, the resource scheduler plays a prominent role. This component mediates between concurrent computation and communication requests and has a profound impact on the Quality of Service (QoS) experienced by the different applications. We propose the use of adaptive resource allocation, e.g. use the platform parameters as actuators to maintain the QoS within specified bounds. Cross-layer adaptation can be an effective mean to adjust the network to the need of the control applications. In protocol design for networked control, cross-layer adaptation can be used to choose optimally the variables of the protocol at various layers, such as the radio transmit powers, the medium access control and routing.
Cross-layer optimization can be posed as an optimization problem where the decision variables are associated to the communication layers and the constraints include requirements from the control applications, the computation resources, and the fundamental limitations of the communication channel. The solution of these optimization problems is often performed off-line or via numerical algorithm based on decomposition techniques and the Lagrange duality theory. However, this approach requires often a global knowledge of the state of the network at some central point, which is quite difficult, if not impossible, to obtain. We will investigate how an online solution of cross layer optimization problems couldbe achieved.
Task 2.3 Protocol design and optimization for networked control [KTH (lead), TUE, TUDO]:
Energy-efficient and reliable protocols are essential for wireless sensor networks used in scenarios where source information must be available for control applications. However, widely accepted protocols for control and actuation applications are still missing. The adoption of wireless technology further complicates the design of networked control systems. Two strategies are proposed to tackle the problem of designing protocols for wireless sensor networks used in control applications: 1) development of protocols from scratch, or 2) optimization of existing protocols (such as IEEE 802.15.4, and Wireless HART). The first approach will require substantial system level design, where the complex requirements of control and communication are simultaneously considered to design new protocol stacks. The second approach will be based on the optimization of the numerous protocol parameters that existing solutions, such as IEEE 802.15.4, offer. In either approach, a modelling of the protocol behaviour as a function of the communication and control parameters is essential.
Task 2.4 Networked control systems in the application domains [CNR, ETHZ, INRETS, TUB, TUDELF, UNIPD, UNIPV, UNITN, UNIVAQ (lead), US]: This task will investigate the need for novel methods in networked control systems and the impact of those methods in transportation systems (AD1), energy (AD2) and biological and Medical systems (AD3). In particular, we will investigate the following:
Intelligent tires: The use of “intelligent” (also called smart) tires with sensors that are embedded in the tread to provide direct tire strain measurements, allows precise measurement of friction, and hence increases the efficiency of active attitude control systems. It also reduces realization costs, it increases flexibility, and it eases maintenance, debugging and diagnostics. Problems regarding wireless technology, e.g., packet losses, fading effects, and synchronization losses, need formal models and an analytic approach that takes into account all non-idealities that are typical of Networked Control Systems (NCSs).
Wireless Technology for Energy efficient intelligent buildings: Commercial buildings are responsible for a significant fraction of the energy consumption and greenhouse gas emissions worldwide. If one could achieve more than 50% reduction in energy consumption in new and existing buildings, respectively, the impact on carbon emissions will be significant. Energy efficiency can be achieved by deploying a complex distributed control system consisting of heterogeneous components connected both wirelessly and with wires. The case for wireless technology in this application domain is compelling: a building evolves, mutates, grows or restructures during an average lifetime of about 100 years. When planning and designing a new building, the wiring is done anticipating some near term needs that are bound to grow and change with time in an unpredictable way. Wireless technology may be the only way of introducing new functionality. In addition, in the case of events such as fires or earthquakes, wired connections can be made inoperable and a reliable wireless communication infrastructure provides a major benefit. The data transmitted in an intelligent building range from simple sensor readings (for example, temperature) to complex images especially for rescue operations and people identification. Hence, energy efficient intelligent buildings provide a challenging environment to develop and test networked control systems concepts.
Electrical stimulation: While embedded control is the current paradigm in electrical stimulation and neuroprosthesis, coupling of these technologies with smart environments is conceivable in the near future with the goal of improving patient benefits. Interaction with the environment will require a substantial change in the control layer that most likely will be composed by multiple regulators connected through wireless and wired channels. To this aim an early assessment of the impact networked and distributed control can have on the future of electrical stimulation will be pivotal in guiding research.
Task 2.5: New control paradigms and architectures, network resource adaptation and protocol design in novel application areas [KTH, RUB, UNIVAQ (lead), US]: During the last year of the project, efforts will be devoted to exploring the needs for new networked control methodologies for emerging application domains not considered during the main part of the project.
To read the available deliverables you must be a Level 1 Partner.
To upload a deliverable you must be a WP Leader.
Please log in to have access to deliverables and to download the template.