FEIT Research Project Database

Real-time optimisation

Project Leader: Jonathan Manton
Primary Contact: Jonathan Manton (jmanton@unimelb.edu.au)
Keywords: optimisation; signal processing
Disciplines: Electrical & Electronic Engineering
Research Centre: Nonlinear Signal Processing Lab

Most problems in signal processing reduce to optimisation problems. They are generally solved as if they were "once-off" optimisation problems whereas great benefit can be gained by treating them instead as "real-time" optimisation problems. The crucial distinction is that once-off approaches consider the geometry of each individual cost function (eg, is the cost function convex?) whereas our approach (called Optimisation Geometry) is to study the geometry of the class of cost functions as a whole. This project develops signal processing algorithms based on this real-time optimisation viewpoint.