Sensors: Intelligent Multi-Sensor Modeling, Identification, and Data Fusion for Automated Manufacturing

摘要本研究的主要目标是开发一种创新的方法，用于在灵活的制造工作单元环境中对来自多个异构信息源/传感器的数据进行建模、识别和融合。本研究的主要贡献将包括将人工智能技术（如遗传神经模糊方法和贝叶斯估计）与信号处理和统计技术（如扩展卡尔曼滤波）以及参数估计方法（如极大似然和期望最大化）相结合。本研究项目将集中于以下三个主要组成部分：1)精确的传感器建模，包括获取传感器和相关噪声的分析和概率模型，了解它们的能力和局限性；2)从多个传感器中提取实时数据，并在一个共同的处理平台上解释数据；3)制定策略，将来自这些传感器的无噪声数据结合起来，以消除不确定性，并获得制造系统运行环境的准确模型。这些创新的理论将在杜克大学的柔性制造车间进行测试和验证。由于该研究将强调在制造过程中使用各种传感器，因此制造业将成为主要受益者。

Intelligent Multi-Sensor Modeling, Identification, and Data Fusion for Automated Manufacturing(NSF Proposal No. 0427597)AbstractThe primary goal of this research effort is to develop innovative methodologies for modeling, identifying and fusing data from multiple heterogeneous information sources/sensors in a flexible manufacturing workcell environment. Major contributions resulting from this research would consist of integrating artificial intelligence techniques, such as Genetic-Neuro-Fuzzy method and Bayesian estimation, with elements of signal processing and statistical techniques such as Extended Kalman filtering, and parametric estimation methods such as Maximum Likelihood, and Expectation Maximization. This research project would focus on the following three major components: 1) precise sensor modeling which would include obtaining analytical and probabilistic models of sensors and associated noises, understanding their capabilities and limitations, 2) extracting real-time data from multiple sensors and interpreting the data on a common processing platform, and 3) developing strategies to combine the noise-free data from these sensors to remove uncertainty and to obtain accurate model of the environment the manufacturing system operates in. The innovative theories developed would be tested and validated via experiments conducted at Duke University's flexible manufacturing workcell. Since the research would emphasize the use of a variety of sensors in manufacturing processes, the manufacturing industry would be a major beneficiary.