SART (Semantic Aware Real-Time scheduler)
Version 0.9, July 2010
Copyright (C) 2010 Ali Paikan, Antonio Sgorbissa,
Università degli Studi di Genova
Via Opera Pia 13, 16145,
Genova, Italy.
Despite many decades of work in robotics, the research community is
still lacking a well-accepted common middleware which is adequate to
deal with complex and real-time intelligent robotic applications. Most
research groups still write their own systems based on standards of
software engineering and using conventional real-time operating sys-
tems. Although some of them provide libraries to develop real-time
applications, they are not still suitable to support the development of
any complex robotic application, especially when time constraints are
crucial. Moreover, they are often inadequate to deal with the com-
plexity of intelligent systems, where it is important to schedule tasks
not only on the basis of their timing constraints, but also to choose
between available tasks depending on the current context.
Semantic Aware Real-Time scheduling (SART in short), has been thought
for complex and real-time intelligent robotic applications. SART addresses
the problem of selecting, depending on the current context, a subset of
tasks to be scheduled by the operating system in order to achieve the
user’s objective, by meeting real-time and additional constraints.
Towards this end, it can be easily interfaced with existing middleware
for robotics to provide them with extended functionalities.
SART has been designed and implemented in software as a three layered archi-
tecture: through the SART Abstraction Layer, it can be easily extended to work
with virtually any possible real-time operating system. Currently, SARTAL has
been implemented for XENOMAI, OSEK/VDX and ETHNOS and available for public
download from SoureForge project.
All experiments performed have shown that SART produces a significant increment
in performance, facilitating management of different tasks in complex intelligent
system, thus being suited for complex robotic applications in which the set of
tasks to be scheduled varies dynamically in a significant way.
The promising future of this notion gives us enormous encouragement to crave for
any chance to contribute to pushing the frontiers of the theoretical and practical
aspects of SART in pursuit of creating more powerful semantic-aware rea-time scheduler
and expanding this idea on distributed real-time systems in future.
Further information and SART API documentation can be found in 'doc'
folder from source code.
Please refer to the following link for updated and newer version of SART.
http://sourceforge.net/projects/sartos/
Ali Paikan
ali.paikan@gmail.com
