Skip to main content

Neural Adaptive Control Technology

Objective

The project aims at : the fusion of adaptive control and neural networks into Neural Adaptive Control Technology; the development and consolidation of the theory of NACT; the delivery of transparent and constructive NACT design procedures for engineering systems; and implementations of the resulting multiple computing agent system in industrial practice.

APPROACH AND METHODS

- Classification of nonlinear plants to be controlled in order to clearly establish the scope of problems the NACT controllers have to tackle.
- Survey of nonlinear control techniques relevant to these control problems and their feasibility for NACT.
- Consolidation and development of adaptive control with feedforward neural networks. Fundamental research on the theory of recurrent neural networks in the context of adaptive control.
- Generation of transparent and constructive design procedures and real-life verification of the procedures on existing industrial test-beds.

POTENTIAL

The expected results will have a significant impact on a wide range of industries. The resason for this is that adaptive control is a universal technique dealing with dynamic systems. This means that if a design procedure works in general it will work on a specific real system.

The Project is intended to contribute to the long-term emergence of universal adaptive control technology based on neural networks. The topic chosen is currently believed to offer prospects for working design methods for real-world nonlinear plants. It is unlikely that the Project will result in a "control panacea"; that goal, if at all attainable, is beyond the timescale and resources of the proposal. The scientific and technological output of the Project should, however, be immediately exploitable. The scope of the project coincides with the future interests of a whole range of industries in the trend to ever more efficient and safer high performance control. This cannot be achieved unless nonlinear problems are attacked at their heart by nonlinear adaptive models and/or controllers. The outcomes of the Project will contribute answers to these issues and may be expected to be rapidly assimilated by industry and exploited by it.

Coordinator

DAIMLER-BENZ AG
Address
Alt-moabit 92B
10559 Berlin
Germany

Participants (1)

UNIVERSITY OF GLASGOW
United Kingdom
Address
University Avenue
Glasgow