Title: Comparison of Fuzzy and Probabilistic Methods for Sensor
-----------------------------------------------------------
Validation, Sensor Fusion, and Diagnosis
-----------------------------------------------------------
Speaker: Kai Goebel
University of California at Berkeley
Department of Mechanical Engineering
6102 Etcheverry Hall
Berkeley, CA 94720-1740
email: goebel@ton.Berkeley.EDU
Abstract:
This talk will compare fuzzy and probabilistic techniques for prediction,
sensor validation and fusion as well as diagnosis which have been
recently developed in our lab. The fuzzy techniques make use of fuzzy
time series for predictions, validation gates for fusion and abductive
inference for diagnosis. Because of less limiting assumptions, these
fuzzy techniques have advantages under certain circumstances over probabilistic
techniques. The latter encompass probabilistic data association, which is Kalman
Filter based, Bayesian fusion, and influence diagrams for diagnosis.
It will be shown with data simulating a dynamic environment with high demand for
sensor data integrity from PATH, how these models behave in the
noisy and fast changing environment of moving vehicles in both open
and closed loop fashion.
This talk will include work done with Alice Agogino and Satnam Alag
+ ==========================< Yutaka Hata >============================+
| *** 4/11/95 - 3/31/96 **** | *** from 4/1/96 *** |
| BISC Visiting Scholar | Associate Professor |
| 327, Soda Hall, CS Division | Dept. of Computer Engineering |
| University of California, Berkeley | Himeji Institute of Technology |
| CA , 94720-1776, USA | 2167, Shosha, Himeji, JAPAN |
| Tel.510-642-8774 Fax. 510-642-5775 | (+81)792-66-1661(Tel),-8868(fax) |
| http://www.cs.berkeley.edu/~hata | hata@comp.eng.himeji-tech.ac.jp |
+------------------------------------+----------------------------------+