Have fun with Automatic Control

I recently kicked off a new project on the design of a Virtual Lab on Automatic Control. Some first experiments are already running online at vlab.mooo.info. Visitors of the web site will be able to play and experiment with closed and open loop systems involving a PID controller and a dynamical system with a time-delay. I have endeavoured to provide full configurability of the framework and allow the user to explore in depth the dynamics of a PID controller in a closed loop. The Bode and Nyquist Diagrams for the open loop system can be generated as well. The project is still in alpha phase (testing) but soon it will be released with more running virtual experiments (with animation and surprises). So... stay tuned (follow @isToxic on twitter for more news).

When Lyapunov's Stability Theorem is not enough...

There are cases where Lyapunov's stability theory is not enough in the sense that it can prove merely stability for equilibrium points that are actually asymptotically stable (after days and nights struggling to find a Lyapunov function for our system). The pendulum with linear friction term is the prime example of such a case: The total energy function of the system is used as a Lyapunov function. However, this absolutely natural choice fails to provide the expected asymptotic stability result. At that point La Salle's invariance principle chimes in to accommodate the aforementioned theoretical shortage.

Lyapunov on a stamp

Note on La Salle's invariance principle: 

Download the PDF file from http://users.ntua.gr/chvng/ac2/LaSalle_Invariance_Principle.pdf 
Read also: The Realm of State Space Systems

PBPK modelling and Predictive Control

Traditionally, drug administration scheduling is – at best – designed using average population data such as PK/PD profiles. This common practice yields suboptimal therapies and does not consider the distinct attributes of the patients treating them as a bulk. Outliers of the general distribution are likely to exhibit adverse effects due to violation of toxicity constraints or fail to retain the therapeutic levels. The lack of any feedback contributes even more to the probability of something going wrong. Nowadays, the grounds have shifted and the need for accuracy and efficiency calls for closed-loop practices introducing thus automatic control into the field. Computer-aided drug administration will provide an optimal solution to the problem, with the guarantee that all safety requirements are fulfilled.

State observer design for nonlinear systems

The Frobenius theorem of differential geometry provides necessary and sufficient conditions for a distribution to be completely integrable. This has important implications on the design of controllers and observers for nonlinear systems. In this presentation we go through all necessary definitions in order to state the Frobenius theorem. These are the notions of a Lie bracket, distribution, involutive and completely integrable distribution, codistribution and annihilator of a distribution. We then state the Frobenius theorem and explain its logical depth.

The realm of State Space Systems

These notes, made available at http://users.ntua.gr/chvng/en, cover a wide range of topics on automatic control of linear time invariant systems. We start with an introduction to state space models and offer a rudimentary classification of control systems (Linear Time Invariant, Linear Time Variant, Input Affine and more) and we juxtapose the Laplace transform approach. 

Notes on Automatic Control