A structure for maintaining a shared world model in a dynamic environment between differentiated embedded systems and their interaction with human supervisors (AES.5414)

There is a growing demand for the deployment of embedded autonomous systems in distributed environments. Applications are in public safety, monitoring and control of traffic and environmental conditions, assistance and clean-up work in disaster areas (e.g. [11]). The complexity of the tasks encountered almost always necessitates the collaboration of multiple embedded systems with a functional differentiation. One can think of functional differentiation on (i) ability to move, (ii) the sensor and actuator capabilities and (iii) decision level. Collaboration must be based upon a collective world model. In order to allow for robust operation and graceful degradation when communication is disrupted or a system is lost, a realization of such a model requires that the systems involved can communicate and have a locally maintained view of the collective world model. The degree of agreement between the locally maintained world-models, i.e. the level of uncertainty that can be tolerated, depends on the level of coordination required.

The situations faced by these systems are characterized by a dynamic nature. The strength of networked embedded autonomous systems is that they can adapt to these changing situations.

In many situations, interaction of the autonomous systems with humans for critical decisions and monitoring activities is of vital importance. There is a trade off between (the limited) decisions which can be taken by an autonomous system, human supervision and interaction. In particular in dynamic situations where decisions are time-critical the trade offs are important. An example of a decision wheresuch a trade off is involved, is where to go for obtaining additional information on both the world model and the supervisor's understanding of the situation.

This proposal focuses on the development of a structure for maintaining a shared world model in a dynamic environment between differentiated embedded systems and their interaction with human supervisors.

For results, see http://staff.science.uva.nl/~vlassis/ and http://carol.wins.uva.nl/~mtjspaan/.

Seven companies and one other university are involved in this project.