Feedback Control using a Wireless Network
Supervisors: Prof G. W. Irwin and Dr W. Scanlon
This is an interdisciplinary project which combines both communications and control in an interesting and new area of research.
In a Networked Control System (NCS) the plant, sensors, controllers and actuators are interconnected by a communication network. Replacing point-to-point wiring with a communications network has important practical advantages in terms of reduced weight, simpler installation and maintenance, better reliability and improved flexibility, all at a lower cost.
Despite this, industrial NCS is characterised by a lack of standardisation, which contrasts sharply with the computer networking field. For example, the emergence of global standards for commercial wireless networking technology, that cross both physical and application domain boundaries, has facilitated the growth in low cost, low power, high performance network devices with an extremely high degree of interoperability.
The analysis and design of conventional computer control systems generally relies on a number of simplifying assumptions such a constant sample rate, synchronised control and non-delayed sensing and actuation. The insertion of a data network in the feedback path adds complexity to the control analysis and design. The main communication impairment is unpredictable delays (latency) and variation in delay ("jitter") caused by routing congestion or medium access delays in network nodes.
Current research is focused mainly on packet-switched wired (e.g. Ethernet LAN and inter-networked) solutions. Wireless NCS, using technology such as IEEE 802.11, the most general wireless standard is an exciting prospect offering corrosive environment traversal and inherent mobility advantages over wired solutions. Unfortunately, wireless links are subject to relatively high levels of packet error rate and, depending on the protocol used, medium access contention. Although both of these effects could be simply considered as "delay", their time-varying statistical nature and the operational benefits of wireless warrant serious study.
One of the main omissions in NCS research to-date is the determination of quality of service (QoS) requirements for closed-loop control systems. QoS enhancements are now available for both wired and wireless network protocols, although admittedly for multi-media services, and not for automatic control. It is therefore crucial to be able to definitively specify QoS performance for the widespread adoption of wireless NCS. Furthermore, the subject demands an integrated communications and control approach to achieve the required quality of performance (QoP) needed for automatic control, an application domain uniquely characterised by hard real-time operation.
Our previous work has investigated adapting the sample rate using QoS measures (congestion, length of delays or channel quality) for a fixed controller design. Closed-loop stability has been proved theoretically, along with scientific simulation studies to quantify the effects of channel errors and channel contention on a wireless NCS application. This new PhD project will build on this work to include varying the control law to maintain closed-loop performance in the presence of varying channel conditions. Several alternative approaches are envisaged:
- Unlike hard-wired computer control, the availability of the communications network allows a sequence of measurement and control samples to be transmitted at each instant. The most timely one can then be used within a predictive control framework to handle the time-varying time delays and packet losses introduced by the wireless NCS.
- Model-reference adaptive controller where the parameters are adjusted online, using suitable QoS measures, to maintain closed-loop control performance.
Both fixed and variable-sampling rates will be considered in each case and the methodology will involve both theoretical developments and application simulation studies.