|
|
유비쿼터스
|
| 1. Motivation
Why intelligent techniques for ubiquitous
systems?
- Ubiquitous environments are highly
dynamic, noisy and unpredictable.
- Inference of high-level contexts from
various information sources (sensors, user profile, web information)
is needed because the high-level contexts are not directly sensed
- Predefinition of service for all
configurations is impossible and generation of high-level services from
low-level components in real-time is very important
Figure 1.
Context-Awareness between human and ubiquitous robot
The goals of this research are as follows.
- Developing an adaptation scheme in the
context of middleware and applications
- Developing appropriate information
fusion methods for ubiquitous systems
- Developing context-aware system for
ubiquitous systems
- Developing basic theories and
algorithms of the advanced intelligent models for ubiquitous environment
2. Proposed Methods
- Building Bayesian networks for
context-awareness of service robots
- Developing advanced methods for
Bayesian networks such as refinement of Bayesian networks, decision
networks, and dynamic Bayesian networks
- Writing our own customized Bayesian
network API (learning and inference)
- Detection of system faults (system
crush, malfunction of applications ...) before they happen using
probabilistic models
- Flexible service generation mechanism
based on behavior-based approach
Figure 2. Proposed
context-awareness system based on multiple Bayesian networks and rule-based
selection scheme
3. Experimental Results
- Scenario: Greeting service of service
robot
- Building 15 different Bayesian networks
for a number of different situations
Figure 3. Greeting
scenario in the early morning
Figure 4. Bayesian
networks for inference of user's intention in the early morning
4. Publications
[1] S.-J. Han, and S.-B. Cho,
"Evolutionary neural network for anomaly detection based on program's
behavior," IEEE Transactions on Systems, Man and Cybernetics-Part B,
2005.
[2] K.-J. Kim and S.-B. Cho, "Robust
inference of Bayesian networks using speciated evolution and ensemble,"
Lecture Notes in Artificial Intelligence, vol. 3488, pp. 92-101, 2005.
[3] K.-J. Kim and S.-B. Cho, "Construction
of Bayesian networks for context-aware service robot," The 2nd
International Conference on Ubiquitous Robots and Ambient Intelligence,
2005.
[4] J.-O. Yoo, K.-J. Kim and S.-B. Cho,
"Inference of high-level context using fuzzy Bayesian network for music
recommendation of service robot," The 2nd International Conference on
Ubiquitous Robots and Ambient Intelligence, 2005.
[5] M.-C. Jung, and S.-B. Cho, "A novel
method based on behavior network for web service composition," IEEE
International Conference on Next Generation Web Services Practices,
2005.
|
|
|
