DC: Small: A Programming Model for Distributed Data Fusion

随着传感器的范围扩展到智能手机，不断捕获和/或传输各种类型的数据，到处都在实时生成数据。许多数据以不同的语义形式出现，需要加以区分。例如，推文和即时消息携带不同类型的信息，尽管两者都基于文本。这种分布式数据承载着丰富的语义，需要捕获和处理这些语义，以便近乎实时地确定信息的状态，这带来了许多挑战。这项工作提出了一种基于语言(Java)的方法和一个基于移动代理的系统来利用数据源的属性来捕获数据的重要属性，如及时性、通用性、不完备性、一致性和自主性。此外，由于数据源是完全自治的，信息流策略规定了对非本地数据的访问控制。策略对数据访问和移动构成了限制，在这种情况下存在多种处理可能性。实时地进行这样的数据融合是非常具有挑战性的，使用Java移动代理框架是一种设想。编译器分析的作用对于使运行时变得智能以减少不必要的开销以控制分布式实时处理需求至关重要。建议在交通和导航系统等领域的几个应用程序上测试该软件基础设施。

Data is being generated everywhere in real-time since the scope of sensors has extended to smartphones that continually capture and/or transmit a varied type of data. A lot of data comes in different semantic forms that needs to be distinguished. For example, tweets and instant messages carry different types of information although both are based on text. Such distributed data carries rich semantics that need to be captured and processed for determining the state of the information in near real time and this poses many challenges. This work proposes a language-based (Java) approach along with a mobile agent based system to capture important properties such as timeliness, currency, incompleteness, consistency, and autonomy of data utilizing the properties of their sources. Moreover, since the data sources are quite autonomous, information flow policies dictate access control of the non-local data. Policies pose constraints on data access and movement and multiple processing possibilities exist in such a scenario. Performing such data fusion in real-time is very challenging and it is envisioned to use Java mobile agents framework. The role of compiler analysis is critical to make the runtime smart to reduce undue overhead to get distributed real time processing needs under control. It is proposed to test this software infrastructure on several applications from the domains such as the transportation, and the navigation systems.