Combining structuring, learning and modelling

Methods and mechanisms to observe human manipulation in a way that the observed actions enhance the robot’s skills

Probabilistic abstract representation for manipulation activities, including actions and symbolic goals, to be used for planning and verification

New kinematic model of the human hand

Taxonomy of bimanual grasping and manipulation

Representation and decomposition at the high abstraction level of the manipulation activities including goals and strategies

Scene analysis and task reasoning component

Reasoning over coordination schemes resulting in a decision based on global task-specific quality measures including hand, arm, object and even tool features and capabilities

Grasp and path planning

Scheduling and monitoring goals and different strategies during execution

Automatic selection of the appropriate control parameters

Smoothness during transitions

Hybrid discrete-continuous control (task coordination and feedback control)

New actuators

Combined sensing

Robot hand prototypes