Model Predictive Control (MPC) on a Chip

http://www.ntu.edu.sg/home/ekvling/MPC-on-chip/index.htm

Contact Info: Dr LING Keck Voon || Email: ekvling@ntu.edu.sg || Tel: +65 6790 5567

Project Summary Report

The development of Mirco Electro Mechanical Systems (MEMS) has enabled mirco miniaturization of machines which traditionally occupy a large volume space into chip scale level. There will be increasing demands for sophisticated control systems to be embedded into these micro machines to achieve truely integrated ``laboratory on a chip'' or even ``factory on a chip'' devices. A case in point is the concept of a biochip for biological and chemical analysis. One specific example of a biochip is the uPCR (micro-Polymerase Chain Reaction) bioreactor. This project aims to develop embedded control technology for miniaturized devices, We aim to develop tools and modules to exploit the vast computational power and flexibility of FPGA for embedded control of miniaturized devices, and enable quick and flexible deployment of advanced control strategies to meet the demands of these applications. One motivation for using FPGA as a prototyping platform for embedded control of miniaturized devices is that the resulting design would contain the blueprint for the fabrication of the control systems. This provides a direct path towards the fabrication of tightly integrated miniaturized devices and their control systems. The uPCR bioreactor will be our targeted application which will serve to validate the proposed solutions and as a high-profile demonstrator of the technology. Model Predictive Control, or MPC as it is sometimes known, is our candidate control technology to be encapsulated in suitable FPGA modules.
About MPC MPC is a generic control technology which is more powerful than the basic ``Proportional + Integral + Derivative'' (or PID) control which is routinely used for process control. It is a suitable technology for control problems which are too complex for a PID controller to solve satisfactorily. For example, problems in which several variables which interact with each other are to be controlled, problems in which the performance specification is unusually high, problems in which significant ``dead-times'' occur, and --- above all --- problems in which there are significant constraints that cannot or should not be violated by the manipulated or the controlled variables. Temperature control of a uPCR chip is such a problem. The chip in fact contains many uPCR's, which are close enough to each other for large interactions to exist, the performance specification is quite exacting, and there are constraints in the form of power limitations, and temperatures which must not be exceeded if the samples are not to be destroyed. MPC remains relatively easy to tune, compared with most other ``advanced control'' proposals, and this is a major reason for proposing it as a generic technology, to complement existing PID technology. Another reason is that recent developments indicate that the formalism of MPC can be applied to so-called hybrid systems, namely systems containing switchings, continuous and discrete states, with design specifications often expressed in terms of dynamics response, logical statements and constraints. This opens up an extremely wide and important range of applications in all industrial and commercial sectors (manufacturing, chemical and process, automotive, marine, aerospace, etc). MPC relies on solving constrained optimization problems on-line. The computational load for such problems is much greater than for traditional control algorithms. Some open issues such as the packaging of the algorithm for an effective and flexible device-application interface need to be studied. Algorithmic enhancement of the control design techniques to allow a proper tradeoff between computational load and the complexity of implementation will be investigated.
About PCR PCR was developed in 1985 by Kary B. Mullis, who was awarded the 1993 Nobel Prize in chemistry for his work. It is used in DNA fingerprinting and in medical tests to identify diseases from the infectious agent's DNA. In forensic use, the test can be used to compare two samples of DNA. PCR is also used in classification to help show evolutionary relationships between organisms on the molecular level. The uPCR chip promises improved temperature uniformity and reduced cycling time, together with decreased sample and reagent volume consumption. Hence it has the advantages of being able to be used even when only very small samples are available, of faster operation, and of improved accuracy. This would provide healthcare workers with portable and powerful devices which they can carry to the field for timely detection of diseases with quick response to life threatening situations. The use of semiconductor chip fabrication technologies to fabricate such chips in large quantities would also result in radical cost reduction. The realization of an embedded control for the uPCR bioreactor serves to demonstrate the applicability of the ``MPC on a chip'' concept. The current approaches to control system design for uPCR reactor rely on empirical and heuristic techniques, based on a combination of ingenuity and experience. We propose to develop a more analytical model-based approach using techniques from control theory, system identification and signal processing. These are theories that have been successfully applied in the aerospace and process control application. However, they have not been used in uPCR reactor.

Presentations and Papers

• KV Ling, SP Yue and JM Maciejowski, Model Predictive Control on a Chip, to be presented at ACC’06, June.
• KV Ling, JM Maciejowski and BF Wu, Multiplexed Model Predictive Control, 16th IFAC World Congress, Prague, July 2005. [presentation slide]
• MH He and KV Ling, Model Predictive Control on a Chip, The 5th International Conference on Control & Automation, Budapest, Hungary, June 26-29, 2005.
• “Model Predictive Control (MPC) on a Chip”, presented at the Regional Symposium on Embedded Hybrid Systems, 12 October 2004, Suntec Singapore International Convention & Exhibition Centre. [pps file (4MB), poster (1MB)]
• KV Ling, MH He, BF Wu, Y Zhang, An Augmented Model Predictive Controller for Multirate Cascade Systems, American Control Conference, Boston, USA, pp.1575-1579, 30 June - 2 July 2004. [slide, paper]

Related Work

• KWAN Kin Hoong, SIM Cherng Yi Sabrina, CHIN Kok Shyang and LING Keck Voon, A Modular FPGA Design For Servo Application, Design Competition sponsored by Micron Semiconductor Asia Pte Ltd, 4^th-7^th prize, 2003.
• Chin Kok Shyang and Ling KV. Re-configurable Embedded Control System Design, 9th National Undergraduate Research Opportunities Programme (NUROP), 13 September 2003, Singapore
• KV Ling, P. Falugi, JM Maciejowski and L Chisci, Robust Predictive Control of the Furuta Pendulum, 15th IFAC World Congress, Barcelona, Spain, July 2002

Projects Completed in 2005

Project Title	Principal Investigator	Funding
Model Predictive Control (MPC) on a Chip See also A*STAR EHS Booklet	LING Keck Voon	ASTAR (Ref:  022 106 0044)
Automatic Code Embedment for Manufacturing Prognostics	LING Keck Voon Goh Kiah Mok (SIMTech)	NTU-SIMTech project (Ref: U04-E-110B (Apr 2004 Apr 2006)

Projects Completed in 2004

Project Title	Principal Investigator	Funding
Embedded Control System for a Micro-PCR Bio-chip	LING Keck Voon CH'NG WEI LUEN LEE CHUNG HONG, ANDY	FYP4039

Projects Completed in 2003

Project Title	Principal Investigator	Funding
MPC on a chip	LING Keck Voon	ASTAR (Ref: 012 106 0055)
Investigation of FPGA Technology for Embedded System Applications	LING Keck Voon	FYP