Research Initiation Awards: Spatial-temporal Information Fusion and Real-time Sensor Data Assimilation Using Sequential Monte Carlo Methods

该研究启动奖项目题为-时空信息融合和实时传感器数据同化使用顺序蒙特卡罗方法-的首要目标是开发易于处理的方法，时空信息融合和真实的时间传感器数据同化使用最先进的概率技术的基础上顺序蒙特卡罗（SMC）方法。将开发新的算法和方法，以提高大规模时空系统的信息融合和数据同化的有效性和效率。一个特别的重点是有效地使用实时信息和传感器数据，从地理上分布的数据源，更准确的状态推断使用SMC方法。该项目的具体目标包括：开发创新算法，以利用系统的时空状态进行更有效的采样/恢复和SMC方法的收敛;开发先进的数据集成方法，以支持从分布式数据源吸收实时信息和传感器数据;开发基于并行环境的新的计算方法和基础设施，以提高大型计算机的性能，该项目有望对时空信息融合和真实的时间传感器数据同化的理论和实践方面产生影响。新开发的方法将增加无缝集成真实的时间数据流的能力，以支持更好的分析和预测以及真实的时间决策。该项目涉及本科生，通过培训他们以逐步的方式进行研究，扩大他们的网络，并为研究生学习或劳动力做好准备。该项目将通过远程访问格鲁吉亚州立大学的集群系统，加强Voorhees学院校园内计算机科学课程的教学，学生可以在集群系统上运行并行和分布式程序。

The Research Initiation Award project entitled - Spatial-temporal Information Fusion and Real-time Sensor Data Assimilation Using Sequential Monte Carlo Methods - has the overarching goal to develop tractable approaches for spatial-temporal information fusion and real time sensor data assimilation using state of the art probabilistic techniques based on sequential Monte Carlo (SMC) methods. New algorithms and methods will be developed to enhance the effectiveness and efficiency of information fusion and data assimilation for large-scale spatial temporal systems. A special focus is to effectively use real-time information and sensor data from geographically distributed data sources for more accurate state inference using SMC methods. Specific objectives of this project include: developing innovative algorithms to exploit the spatial-temporal state of a system for more effective sampling/resampling and convergence of SMC methods; developing advanced data integration methods to support assimilating real-time information and sensor data from distributed data sources; and developing new computing methods and infrastructure based on parallel environment to enhance performance of large-scale information fusion and data assimilation.The project promises to have an impact on both theoretical and practical aspects of spatial-temporal information fusion and real time sensor data assimilation. The newly developed methods will increase the capability of seamlessly integrating real time data streams to support better analysis and prediction and real time decision making. The project involves undergraduate students by training them to conduct research in a stepwise manner, broadening their network, and preparing them for graduate studies or the workforce. The project will enhance the teaching of computer science classes on campus of Voorhees College by remotely accessing the cluster system at Georgia State University on which students can run parallel and distributed programs.