Incorporating Artificial Intelligence into Multisensor Data Fusion Designs

• Main Authors: Ya-Jane Lung, 龍雅真
• Other Authors: Dah-Jing Jwo
• Format: Others
• Language: en_US
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/03288185108663815510

碩士 === 國立海洋大學 === 導航與通訊系碩士班 === 91 === Multisensor data fusion is the processing and synergistic combination of information gathered from multiple sources and sensors to provide a better inference of phenomenon. It is being applied to the wide range applications including military and non-military fields. For tracking monitoring, an Intelligent Connected Decentralized Network (ICDN) data fusion architecture, which embeds Fuzzy-Genetic Algorithm (FG) and Kalman filtering technique, is explored. The overhead communication problem in FCDN will be resolved by ICDN. At first, describing the FCDN consists of nodes (sensors), each with its own processing facility, which takes local observations and shares information with other neighbor nodes. It then assimilates the communicated information and computes a local estimate using Kalman filter. Second, based on a covariance matching technique, fuzzy-adaptive Kalman filter and Weighted Fuzzy Assessor (WFA) is designed respectively in order to estimate the appropriate measurement noise covariance matrix R and assign a weight that indicates a degree of the filter performance. Finally, the Genetic Algorithm (GA) acts as an important role in ICDN architecture. It compensates the insufficiency of fuzzy logic system designed and overcomes the communication limitation of FCDN. The result shows good performance of ICDN rather than that FCDN. Even best than Classical Decentralized (CD) architecture whose sensor or node doesn’t transmit information with each other.