Collaborative Unmanned Aircraft Systems
The practical use and acceptance of unmanned aircraft systems (UASs) is dependent on a verifiable, principled and well-defined foundation for interactions between human operators and autonomous systems. This interaction is going to be mixed-initiative in nature. Humans will request help from autonomous systems and autonomous systems will request help from humans when collaborating to achieve complex missions in unstructured and challenging environments. In developing a principled framework for such sophisticated interactions, a great many interdependent conceptual and pragmatic issues arise and need clarification both theoretically and pragmatically in the form of demonstrators.
In this project we are targeting a triad of fundamental, interdependent conceptual issues: delegation, mixed-initiative interaction and adjustable autonomy. These are used as a basis for developing a principled framework for collaborative unmanned aircraft systems. These concepts can be used to clarify, validate and verify different types of interaction between UASs and human operators. One of our main platforms is the LinkQuad quadrotor system which is used as a testbed for multiagent systems.
Autonomous robots interacting with the real world often face complex and uncertain environments. As the complexity of a problem increases, it becomes less feasible to model exactly in advance. One approach to tackle this is to give the robot some capability to learn from experience. Robot learning is a growing area of research at the intersection of Machine Learning and Robotics that study such problems. If the environment can change over time, which is often the case in the real world, such learning may be required for the robot to carry out its task.
The complexity of developing deployed architectures for realistic
collaborative activities among robots that operate in the real world
under time and space constraints is very high. This complexity will
be tackled by working both abstractly at a formal logical level and
concretely at a system building level. More importantly, the two
approaches will be related to each other by grounding the formal
abstractions into actual software implementations. This guarantees the
fidelity of the actual system to the formal specification.