| Flexible Situation Assessment |
|
Organization responsible: UvA, TRT-NL People involved:
Project description: The proposed research addresses several theoretical as well as practical aspects of robust information fusion in real world applications. In particular, we focus on fusion systems that support automated interpretation of large amounts of heterogeneous and uncertain information. Such system can assist human decision makers with automated situation assessment and represent indispensable components in complex intelligent systems. In general, knowledge about the situation, i.e. the state of the system and its environment, is a basis for adequate planning and action selection. We introduce a simple example in order to illustrate relevant problems and explain contributions of the proposed research. Let’s assume that several vehicles are involved in an accident in a tunnel on a busy highway, which is a regular route for tanker trucks transporting toxic chemicals. The accident results in a fire resulting in fumes that are released into the environment through the tunnel portals and the ventilation system. Due to a high chance of the presence of chemical tanker trucks at the incident site, the crisis managers must find out as quickly as possible whether a toxic gas has been released into the populated area nearby. However, devices for reliable detection of toxic gases are usually not available. In order to avoid false alarms or ignorance of serious situations, we must infer the presence of the gas through interpretation of large quantities of heterogeneous observations. Such observations could be obtained from sensors installed in the tunnel and it surroundings and through human reports about typical symptoms such as smell, haziness, irritation, etc. In addition, UAVs equipped with sophisticated sensor suites could provide valuable information on the gas concentration in the tunnel’s vicinity. Obviously, quick and reliable situation assessment requires an appropriate infrastructure that can efficiently integrate various information sources into meaningful information filtering systems. In addition, situation assessment often involves reasoning about the facts which cannot be observed directly, such as the presence of toxic gases. Consequently, we must infer such hidden events of interest through appropriate interpretation of patterns of observable symptoms, such as uncertain chemical sensor measurements, different health problems, etc. Clearly, the accuracy as well as the efficiency of such interpretation is crucial for adequate decision making and optimal control where misleading or delayed state estimation can have devastating consequences. The main focus of the proposed research is on theoretically rigorous techniques that support:
|