Parallel Processing -
A Personal Tour from Simulation via Image Processing to Robotics

Thomas Bräunl
Univ. Stuttgart, IPVR
Breitwiesenstr. 20-22
D-70565 Stuttgart, Germany
braunl@informatik.uni-stuttgart.de

ABSTRACT

Programming methodologies for SIMD and MIMD systems are the basis for most parallel applications. Modula-P and Parallaxis are parallel programming languages developed by the speaker. Modula-P is designed for MIMD processing with explicit processes and synchronization via Semaphores, Monitors, and messages. Parallaxis, however, is a data-parallel language with a new concept for defining PE connectivity. Both languages can be used for simulation of parallel processing on sequential systems, as well as for programming real parallel systems.

Data-parallelism is sufficient for many compute-intense application areas. This holds for simulation applications, e.g. traffic simulation, as well as a whole library of basic image processing operations.

The use of cameras as sensors for mobile robots explains the need for fast image processing systems in this area. Robotics is an ideal test bed for applying computer science techniques to practice. All kinds of parallel processing, asynchronous, synchronous, and pipeline, are required for programming mobile robots. Some system architectures explicitly demand parallel processing. There are currently two robot projects at Stuttgart. Project COMROS (Cooperative Mobile Robot Systems Stuttgart) follows the more traditional engineering approach with large and accurate mobile robots. Project AutoLife (Autonomous Mobile Robots for Artificial Life) uses small and inexpensive mobile robots for Artificial Life research.

Colloquia Series page.