Control and Signal Processing Lab

2012 Seminars

Control and Signal Lab Seminar - Some Topics on Mean Field Game Theory

Date: 2:00pm Wednesday 5 December 2012

Venue: Richard Newton Room, Electrical and Electronic Engineering Building 193

This talk focuses on Mean Field Game (MFG) theory with applications to consensus, leader-follower and major-minor agent systems. The MFG methodology addresses a class of dynamic games with a large number of minor agents in which each agent interacts with the average or so-called mean field effect of other agents via couplings in its individual cost function and individual dynamics. A minor agent is an agent which, asymptotically as the population size goes to infinity, has a negligible influence on the overall system while the overall population’s effect on it is significant.

In this talk three topics will be briefly discussed: (1) applications of the MFG methodology to large population consensus and flocking behaviour, (2) extension of the mean field linear-quadratic-Gaussian (MF LQG) framework so as to model the collective system dynamics which include large population …

Control and Signal Processing Lab Fortnightly Seminar Series

Date: 2:00pm Wednesday 21 November 2012

Venue: Richard Newton Room, Electrical and Electronic Engineering Building 193

Certified System Identification Towards Probability-Free Results: System identification is the science of constructing models from data. A model is never an exact description of reality, and it is desirable that the identified model comes accompanied by certificates of quality able to describe the level of precision of the model and its domain of validity. This talk is about certified system identification. Our contention is that data contains much more information than traditional identification methods can exploit, and, by looking at classical identification problems with new eyes, methods can be developed carrying precise quality guarantees which are valid under general assumptions. Taking the challenge of developing these methods may lead to a paradigm shift in many contexts in which identification is applied. SHORT BIO: Marco Claudio Campi is a Professor of Automatic Control at the …

Control and Signal Processing Lab Seminar Series - Wednesday 10th of October

Date: 2:00pm Wednesday 10 October 2012

Venue: Richard Newton Room, Electrical and Electronic Engineering Building 193

Dr. Alireza Farhadi, Research Fellow, Electrical and Electronic Engineering Department, The University of Melbourne will present Performance, Information Pattern Trade-Offs and Computational Complexity Analysis of a Consensus Based Distributed Optimization Method

This talk is concerned with a cross-disciplinary approach for the convergence of computer science and control in large scale networked control systems. In this talk a simple consensus based distributed optimization method is presented which approximates the solution of a linear quadratic optimal control problem subject to constraints using distributed decision makers subject to non-classical information pattern. This method can be seen as a mechanism for distributing the computational load of the centralized control to distributed decision makers, which work together in parallel and approximate the optimal solution. …

Control and Signal Processing Lab Seminar

Date: 2:00pm Wednesday 19 September 2012

Venue: Richard Newton Room, Electrical and Electronic Engineering Building 193

What has Instrumental Variable (IV) methods for system identification to offer? Instrumental Variable methods can consistently identify plant models of systems operating in open- or closed-loop while relying on simple linear (regression) algorithms. IV methods are attractive when dealing with complex processes that are high-dimensional with many inputs and outputs, since they do not require non-convex optimization. In addition to this computationally desirable property, IV methods have the potential advantage that they can provide consistent plant estimate even when the noise model is mis-specified. The development and performance of IV methods will be analyzed in two frameworks: linear closed-loop system identification and open-loop system identification of LPV (Linear Parameter Varying) models. An application to rainfall/flow modeling will be discussed.

The main important …

Control and Signal Processing Lab Seminar - Acceleration of Distributed Optimisation Methods

Date: 2:00pm Wednesday 22 August 2012

Venue: Richard Newton Room, Level 5, Electrical and Electronic Engineering Building, Building Number 193

In this talk we consider the settings where an optimisation problem is to be solved distributedly. Particularly, we address the problem of choosing the parameters of the algorithm in a way to achieve the best convergence properties. Initially, we explore the application of multi-step methods to networked optimisation and develop techniques to formulate accelerated primal and the dual algorithms for network-constrained optimization of strongly convex objective functions with Lipschitz continuous gradients. Given the condition number of the Hessian of the cost function and the structure of the underlying network, we determine the algorithm parameters that guarantee the best convergence factor and characterise scenarios when significant speedups can be obtained over the standard gradient descent. Moreover, we study the effects of uncertainty in problem parameters …

Discussion on Transforming Tools of Emerging and Converging Technologies for Societal Benefit

Date: 2:00pm Wednesday 8 August 2012

Venue: Richard Newton Rooms, Level 5, Electrical and Electronic Engineering Building, Building number 193

This Control and Signal Processing Laboratory breadth series talk aims to promote discussion and look at 'the bigger picture' of research in the international context. Students are especially encouraged to attend as the discussion may give perspective to their own research plans. In particular, Professor Manton is seeking feedback on the ideas he will be presenting at a workshop that is part of an international study coordinated by NSF in collaboration with the Chinese Academy of Sciences and CNSF, the Department of Science & Technology in India, and science and technology organizations in Australia.

The workshop describes itself as follows: Some important themes pervade science, mathematics, and technology and appear over and over again, whether we are looking at an ancient civilization, the human body, or a comet. They are ideas …

Convergence of Stochastic Consensus Algorithms over Switching Noisy Networks

Date: 2:00pm Wednesday 1 August 2012

Venue: Richard Newton Rooms, Level 5, Electrical and Electronic Engineering Building, Building No. 193

Associate Professor Minyi Huang will be presenting a seminar as part of the Control and Signal Processing Lab seminar series in the Department of Electrical & Electronic Engineering.

Associate Professor Huang's work considers consensus problems with delayed noisy measurements in switching networks and stochastic approximation algorithms with decreasing step sizes applied. Since the averaging weights must be selected in a distributed manner, the double stochasticity condition widely used in the literature does not hold in general and so the Lyapunov function based approach is difficult to apply. His work develops a new approach to prove consensus, which is achieved by establishing ergodic theorems for degenerating stochastic matrices. …

Building a Smart Water Grid for the Indus River Basin: Challenges & Opportunities of Cyber Physical Systems

Date: 2:00pm Wednesday 18 July 2012

Venue: Richard Newton Rooms, Level 5, Electrical and Electronic Engineering Building, Building No. 193

Dr Abubakr Muhammad invites you to learn about the Lahore University of Management Sciences (LUMS) work in mathematical modeling, sensing, information processing and control that aims to realize efficient management of hundreds and thousands of networked irrigation channels in an integrated manner. Their methods are currently being deployed for testing and comparison by provincial irrigation departments and local water management institutes in Pakistan. In the wake of a worsening water crisis and in full realization of the social and economic impacts of disruptive automation, they strive to find out the limits of 21st century engineering to save a complex 19th century irrigation system from total collapse. …

Contact Us

Prof Jonathan Manton

Director, Control and Signal Processing Laboratory