From several informal scholar exchanges, National Center for
High-performance Computing (NCHC), Taiwan, and Artificial Intelligence
Applications Institute (AIAI), the University of Edinburgh have forged
an unique and ambitious research collaboration and becoming the first
British-Taiwanese collaboration in Grid Computing.
Recognising the needs for long-term ecological monitoring and
planning, NCHC has installed and managed Wireless Sensor Nets in
several national parks in Taiwan. The information collected are stored
in and made available through EcoGrid for access. The information
collected includes surveillance videos in the Fu-Shan National Park
covering the entire area for observing natural lives and protecting
them from potential poachers; audio recording of frogs of rare
species; under-sea coral reef and marine life observation stations and
more. Due to the continuous and non-intrusive methods deployed, such
monitoring and recording efforts have already made ecological
discoveries of significant importance that traditional methods
otherwise could not have made.
Continuous data collection in the EcoGrid, however, poses a great
challenge as how this data may be transformed into useable information
for the ecologists and in a timely fashion. For instance, one minute
of video clip typically takes 1829 frames and is stored in 3.72
Mbytes. That translates into 223.2 MB per minute, 5356.8MB per day and
1.86 Terabytes per year for one operational camera, and due to the
unpredictability of nature, one may not easily skip frames as they may
contain vital information. Based on our own experience, one minute's
clip will on average cost manual processing time of 15 minutes. This
means that one year's recording of a camera would cost human experts
15 years' effort just to perform basic analysing and classifying
tasks. Currently there are three under water cameras in operation and
this will cost a human expert 45 years' time just to do basic
processing task. This is clearly a hopeless situation and more
appropriate automation methods must be deployed.
Based on its expertise in workflow and planning, AIAI provides a
semantic-based workflow method that enables a self-generative
automation for carrying integration and analytical tasks. This method
not only allows us to describe what the initial execution steps in the
workflow, it also allows us to dynamically monitor and modify workflow
at run-time. Our initial work will be published in the conference of
CNNA in Taiwan in May 2005 [1].
AIAI's next aim is to provide a hybrid method that enables a
self-learned workflow enactment. Planning techniques are used as a
basis to determine which procedural steps is to take next according to
the environment and problems encountered, it would also learn from
past experiences and self-adapt when a chosen method does not work.
Publication
[1] Yun-Heh Chen-Burger, Fang-Pang Lin (2005) A Semantic-based
Workflow Choreography for Integrated Sensing and Processing. The 9th
IEEE International Workshop on Cellular Neural Networks and their
Applications, Hsin-chu, Taiwan, May 28-30, 2005.
[CNNA 2005 conference web site]
[Full paper]