Proposé par : Serena Ivaldi (email@example.com, www.loria.fr/~ivaldi)
The team MAIA, and its probable sequel LARSEN, participates to the European Project CoDyCo, whose aim is to advance the current control and learning techniques for whole-body motion of robots interacting with humans and environments, subjects to multiple contacts. CoDyCo is proposing methodologies for performing coordinated interaction tasks with complex systems, combining planning and compliance to deal with predictable and unpredictable events and contacts. The proposed algorithms are validated in real-world interaction scenarios with the iCub humanoid robot engaged in whole-body goal-directed tasks, such as balancing while reaching for a distant object, or leaning on a soft chair.
In this context, one of our objectives is to learn the prioritization of elementary tasks. The core element of the CoDyCo controller is the intelligent combination of prioritized tasks, which allows covering a large variety of possible scenarios while only requiring a small number of elements. Nevertheless, the control architecture requires a meaningful prioritization scheme that tells the systems which tasks to activate and how certain tasks can overrule each other. While it is possible to devise such prioritizations for complex tasks manually, the automatic generation from data is much more desirable. Our goal is to investigate how a prioritization can be obtained from observing tasks, similar as in imitation learning, and how it can be self-improved. The relative importance of the tasks imposed by the prioritization can be changed during execution by the learned prioritization based on the current context. A further improvement is to enable the generalization to novel situations. This activity will involve a collaboration with the partners of CoDyCo in the Technical University of Darmstadt and the University Pierre Marie Curie in Paris.
Alain.Dutech AT loria.fr