Author:

Abstract:

More and more robot applications are deployed in more human-like production lines with a large variability and uncertainty in the task, or production lines manipulating or processing natural products. Robots co-exist with humans in the same workspace, or even more directly physically interact with humans. Developing robot software that can deal with these disturbances requires a significant effort. However, we can’t afford these development costs because typically these same production lines deal with smaller product series. A shift from plan-sense-act to planning in terms of behaviors or skills can help resolving these conflicting goals. A constraint-based task specification and control framework is used to design these skills in a model-based way. We argue that for rapid and scalable development, a clear separation is needed between specification of the skill and controlling the skill.