Report from International Visit in the USA, fall 2001

By Anton Cervin

Supported by a travel grant from ARTES, I spent five weeks in the United States in September and October of 2001. I had the opportunity to visit researchers at two leading universities, University of California at Berkeley (four weeks) and University of Illinois at Urbana Champaign (one week).

The main part of my time was spent in Berkeley, where I visited the Ptolemy research group, headed by Prof. Edward Lee, at the Department of Electrical Engineering and Computer Sciences. About fifteen people (PhD students, post docs, technical staff) are involved in the development of Ptolemy II, which is a tool for modeling and simulation of heterogeneous systems. Heterogeneous means that different models of computation can be used in the same model. To me, this was very interesting, because my own research deals with the interaction of two different domains: control and real-time scheduling.

The classical domains in Ptolemy include finite state machines (FSM), discrete events (DE), and synchronous dataflow (SDF). Recently, new models that relate to real-time control have been or are being developed: the continuous time (CT) domain and the real-time operating system (RTOS) domain (not yet completed). Using these new domains, it is for instance possible to simulate a complex real-time system with communicating computer nodes, real-time scheduling, continuous plant dynamics, etc. A Matlab/Simulink-based tool called TrueTime, developed in Lund by myself, Johan Eker, and Dan Henriksson, has similar capabilities, so it was very interesting to compare the different tools. While Simulink has very good support for continuous-time and discrete-time simulation, Ptolemy has several other useful domains and will soon also support code generation. The Ptolemy project is hosted by the Gigascale Silicon Research Center (GSRC) web site, so it is possible for external developers and evaluators (like me) to follow the developments remotely.

As a consequence of my visit, I was asked to present a paper about control systems development using Ptolemy II at the IFAC Conference on New Technologies for Computer Control in Hong Kong in November (http://starship.mech.hku.hk/ntcc2001), where I was presenting another paper already.

If you want to know more about Ptolemy, please have a look at http://www.ptolemy.eecs.berkeley.edu.

At Berkeley, I also had a chance to discuss with people from another very interesting real-time-related project called Giotto. Giotto is a time-triggered language for embedded control systems. It supports timely sampling and actuation, as well as control systems distributed to several nodes. Developed in conjunction of Giotto is the Embedded Machine, a simple virtual machine which executes E-code (embedded code). The idea is that a Giotto program should be translated into platform-independent code (i.e., the E-code) that retains the real-time properties, especially the scheduling of inputs and outputs in a controller. The Embedded Machine could possibly be used also to implement some of our own feedback scheduling ideas.

For more information about the Embedded Machine, have a look at http://www-cad.eecs.berkeley.edu/~cm

The last week of my trip was spent in Illinois visiting Prof. Lui Sha at the Department of Computer Science. Lui Sha is an external expert in my research project and one of the originators of the feedback scheduling ideas. One of his current projects is dealing with control of delays in web servers. Applying the feedback principle of automatic control to computer systems, it is possible to better deal with uncertainties such as workloads and processing times. During my stay we were also comparing our different simulation tools for feedback scheduling systems.

Both Berkeley and Illinois are very nice places to visit. Berkeley is especially interesting with its very special (I guess some would call it un-American) atmosphere and its location (close to San Francisco, great nature nearby, etc.). Both universities are highly ranked and have great computer science and electrical engineering departments. The only downside to visiting the USA at this time was the very expensive American dollar. I was very lucky in Berkeley to locate a small room for only $600/month in a house where I shared kitchen, etc., with five other international students.