Skip to main content

Multiple model architectures for intelligent modelling and control

Objective



This project aims at the further development of learning systems which allow the developer to introduce a priori knowledge. Such approaches bring structure into the learning task, improving performance and robustness, as well as leading to more interpretable trained systems. The project will combine existing approaches such as Local Model Networks and Markov Mixtures of Models into a single framework which will then be made available to other researchers in the form of a MATLAB toolbox, including visualisation tools. This toolbox will be open to a number of techniques including fuzzy logic and belief networks.
The project will involve partners from Germany (Daimler-Benz research, DLR Braunschweig), who will provide data to test the methodology for real world applications from autonomous robotics and helicopter modelling. Other partners in Britain (Univ. of Glasgow) and Norway (SINTEF research) will collaborate on the theoretical aspects of easing the combination of learning from data while introducing human knowledge and insight.

Funding Scheme

RGI - Research grants (individual fellowships)

Coordinator

Danmarks Tekniske Universitet
Address

2800 Lyngby
Denmark

Participants (1)

Not available
United Kingdom