A3-[Multibot] - Details

Mapping and Coordination with Heterogenous Mobile Robots


The A3 project addresses the problem of exploration with mobile robots and developed several approach to exploration mapping and localization for single and multi-robot systems. In the third phase of the project, we extend our exploration techniques towards cooperating mixed human-robots teams. In several scenarios, teams consisting of humans and robots can be expected to solve the task at hand efficiently and safely for the humans. This is, for example, the case in rescue missions.

Serious problems arise from the fact that the communication and coordination between humans and robots is non-trivial. Typically, the human has to adapt to the capabilities of the robots, especially for interaction. We therefore aim at developing a robotic system that is able to communicate and coordinate with humans in a more natural way. This includes a natural way of interacting, namely by speech, and the ability to share information about spatial concepts as well as objects in space. This capability will enable a robot, for instance, to execute commands such as "check if there is a person in the kitchen" even in unknown environments. Achieving such tasks requires the robots to acquire a model of the world which integrates geometrical maps and semantic classes such as kitchens, restrooms, etc. In order to do so, it needs to be able to identify typically objects that are found in prototypical environments.  We will extend our existing spatial representations so that we will be able to model the geometry as well as the semantics of the scene. This representation will also provide landmarks that can be used to construct route instruction which are meaningful to humans. Such a landmark could consist of one or more objects which are characteristic for a certain region. The joint human-robot team does not only require the understanding of human like descriptions about space and routes in the space. To talk to people, the robots also need to generate meaningful sentences. This will enable them to pose questions, guide a person, or resolve ambiguities.