Contact Person: Fredrik Heintz
For autonomous robots such as unmanned aerial vehicles (UAVs) to successfully perform complex missions, a great deal of embedded reasoning is required at varying levels of abstraction. Both deliberative and reactive parts of a robotic architecture also require access to information about the environment in which an autonomous robot operates. Gathering the required information about the environment is a complex process, often involving the combination of a multitude of information sources and processing techniques at symbolic as well as sub-symbolic levels.
A partial high-level view of the incremental information and knowledge processing required for a UAV traffic surveillance scenario.
DyKnow is a stream-based knowledge processing middleware service that helps organize the required processing in a coherent network of processes connected by streams. A stream-based approach is used since it captures the incremental nature of the information available from sensors and the continuous reasoning process of making inferences with minimal latency necessary to react to rapid changes in the environment. Streams represent aspects of the past, current, and future state of a system and its environment. Input can be provided by a wide range of distributed information sources on many levels of abstraction, while output consists of streams representing objects, attributes, relations, and events. In addition to providing conceptual support, DyKnow is fully implemented and serves as a central component of the UASTech UAV architecture.
DyKnow supports evaluating spatio-temporal formulas in a combination of RCC-8 and the metric temporal logic MTL over streams of information using formula progression. DyKnow further supports chronicle recognition, detecting complex events that are defined in terms of temporal patterns of primitive events. Object Linkage Structures support object classification, where hypotheses about object types and identities can be formed and continuously validated or rejected.
DyKnow Federations is an extension which allows controlled knowledge sharing between agents, such as robots, where agents may dynamically join or leave a federation to share some of its knowledge with the other agents. Specific services are provided for finding agents in possession of a particular piece of information, or agents capable of providing streams of such information. Agents in the DyKnow Federations framework comply with the FIPA ACL standard. Speech acts in the ACL language are used to set up communication channels, after which stream communication is handled through efficient lower level channels.
A conceptual view of DyKnow Federations, with two participating platforms.
DyKnow has for example been used for an execution monitoring functionality based on the incremental evaluation of temporal logical formulas, anchoring symbols to sensor data and a flexible and reconfigurable diagnosis framework (FlexDx).