Entracte a recu le Grand Prix de l'ANR.
The ENTRACTE project addresses the problem of motion planning for anthropomorphic systems, and more generally, the problem of manipulation path planning. The associated trajectories are living in well- identified mathematical spaces, composed of multiple leaves made of locally parallel submanifolds, i.e. the so-called foliations. Moving from a leaf to another one is not directly possible. This structure raises a new order of complexity for planning algorithms: a complexity coming from the discrete structure of the leaves is added to the classical complexity coming from the dimension of the exploring space. No generic algorithm is nowadays able to handle this complexity. For pragmatic resolution, a priori information is added to the planner to specifically steer the resolution in some well-identified situation. However, there is no generic solution, neither to formulate this a priori added knowledge nor to automatically build it.
On the other hand, human beings implement neurocognitive structures able to solve such problems very efficiently. These structures are based on sensory-motor loops, which are closely related to the basic models of artificial motion used to describe the foliation structure of the movement space.
ENTRACTE proposes to study in parallel both the mathematical foundation of artificial motion and the neurocognitive structures used by humans to quickly solve motion problems. The primary objective of the project is to provide pragmatic planning solutions, which would then be applied on humanoid robots and virtual avatars. In a broader view, ENTRACTE addresses problems of planning in complex topological spaces, where the generalization of the structures encountered in humans can give a new viewpoint. In a long-term view, ENTRACT aims at promoting a geometrical approach of the movement neurosciences, by proposing a new theory of embodied intelligence.