The idea of the reactable workgroup is to create a toolkit in Ptolemy II. that enables construction of applications like the reactable (see http://www.youtube.com/watch?v=0h-RhyopUmc ) with the objective of requiring much less infrastructure. I.e., it should work with a standard video camera on a laptop for example.

The underlying goal of the reactable project is to use video processing as a driver for research in to models of computation and cyber-physical systems. In particular, when a periodic model of computation such as Synchronous Dataflow (SDF) interacts with an asynchrounous data source such as a video stream, then there are a number of possibilities. Among them are:

  • Return the last video frame
  • Hang and wait for a new video frame
  • Periodically poll for a new video frame
Each of these designs has tradoffs and is illustrative of design choices that must be made when integrating between a physical process and computation.

The Reactable uses OpenCV, which has a number of Java interfaces. During our review of these interfaces, we asked, "What is the underlying model of computation of the interface?" and "How appropriate is the underlying model of the interface in an actor oriented system?"

These questions have prompted us to use OpenCV as a driver for a discussion about Dataflow with Marc Geilen (Eindhoven University of Technology) about buffer copying and optimization.


©2002-2018 Chess