Description du projet
Cognitive Systems and Robotics
The goal of TOMSY is to enable a generational leap in the techniques and scalability of motion synthesis systems. Motion synthesis is a key component of future robotic and cognitive systems to enable their physical interaction with humans and physical manipulation of their environment. Existing motion synthesis algorithms are severely limited in their ability to cope with real-world objects such as flexible objects or objects with many degrees of freedom. TOMSY propose to solve these problems by learning and exploiting appropriate topological representations and testing them on challenging domains of flexible, multi-object manipulation and close contact robot control and computer animation. The results of this project provide the necessary key technologies for future robots and computer vision systems to enable fluent interaction with their environment.
The goal of the proposed research is to enable a generational leap in the techniques and scalability of motion synthesis systems. Motion synthesis is a key component of future robotic and cognitive systems to enable their physical interaction with humans and physical manipulation of their environment. Existing motion synthesis algorithms are severely limited in their ability to cope with real-world objects such as flexible objects or objects with many degrees of freedom. The high dimensionality of the state and action space of such objects defies existing methods for perception, control and planning and leads to poor generalisability of solutions in such domains. These limitations are a core obstacle of current robotic research.We propose to solve these problems by learning and exploiting appropriate topological representations and testing them on challenging domains of flexible, multi-object manipulation and close contact robot control and computer animation. Topological representations describe motion in terms of more abstract, more appropriate, and better generalizing features: for instance, an embracing motion can better be described, controlled and planned in coordinates quantifying the `wrappedness' of arms or fingers around the object (as opposed to joint angle coordinates). Such topological representations exist on different levels of abstraction and reduce the dimensionality of the state and action spaces. This proposal investigates existing topological metrics (similar to the mentioned `wrappedness') and uses data driven methods to discover new mappings that capture key invariances. Given topological representations, we will develop methods for sensing, control and planning using on these representations.This proposal, for the first time, aims to achieve this at all the three levels of sensing, representation and action generation -- by developing novel object-action representations for sensing based on manipulation manifolds and refining metamorphic manipulator design in a complete cycle. The methods and hardware developed will be tested on challenging real world robotic manipulation problems ranging from domains with many rigid objects to articulated carton folding or origami and all the way to full body humanoid interactions with flexible objects. The results of this project provide the necessary key technologies for future robots and computer vision systems to enable fluent interaction with their environment -- as well as provide answers to the basic scientific question of the `right' representation in sensorimotor control.
Champ scientifique
Appel à propositions
FP7-ICT-2009-6
Voir d’autres projets de cet appel
Régime de financement
CP - Collaborative project (generic)Coordonnées du coordinateur
Coordinateur
100 44 Stockholm
Suède