An Evidence Accrual Data Fusion Technique for Situational Assessment

Abstract

Systems have become significantly more complex as more sensors, actuators, and processing can all be contained within a single system. Such systems now process ever increasing amounts of data that can be used to better understand the environment and issues that concern the system. Whether the system is autonomous or involves an operator in the decision loop, processing the data into useful information is of paramount importance. The process of combining this data into a clearer picture is often referred to as data fusion. These systems are more complex than have been developed in the past. This includes an increase in the number of sensors that monitor not only the external environment of the system but also the current operational state of the system. Systems such as unmanned aerial vehicles (UAVs) and robots may operate according to a set script of actions or be teleoperated but still must have a set of autonomous capabilities to overcome changes in operational conditions, avoid obstacles and other hazards, and be able to identify potential threats or mission criteria. Often, a single sensor cannot provide all the information that necessary to recognize the issue, event, or target of interest. With multiple sensors, more data is available, but the resulting streams of measurement data produced create a number of issues. The concept of data fusion is applied in these cases to sift through the often voluminous amounts of data, determine which data is important, associate the important data that is related, and combine the data into information that provides an improved understanding of internal or external environment. Data fusion is the ability to combine information from various sources, e.g., sensor systems, databases, individual perspectives, etc., into a coherent picture that improves the overall understanding of the events of interest. In some cases, such as target tracking or image understanding, specific types of algorithms can be used to fuse the data into a clearer picture. Data in those cases are often of a similar format and can be easily combined. For other problems, such as classification of entities or events, situation assessment, or impact assessment, the data come from a variety of nonrelated sources. In this chapter, a situation assessment data fusion technique for mechatronic systems is developed where the development of relationships of events or objects are determined. The examples that will be addressed with this type of fusion include a condition-based monitoring system