The Project aims at:
- bundling expertise within computer vision and robotics across the Danish-German border and supplying a competence center that links to industry as well as research and educational centers in the region
- producing scientific knowledge about the manipulation of non-rigid bodies, so that such products can be handled in manufacturing processes in agriculture, meat production and cloth manufacturing, amongst others
- developing and using a framework for adaptive robot vision systems that are able to adapt and to learn robot actions in the domains described above
- developing a prototype robot vision platform in which different kinds of non-rigid objects (clothes, agricultural products, etc.) can be handled
- making the robot/vision platform accessible to companies (small and large and within different market segments) that have a demand for handling non-rigid bodies.
The progress achieved in robot handling within the last decade has mainly had its focus on rigid and geometrically well defined parts and objects.
A very large part of the production and handling of parts and objects are categorized as non-well-defined geometric objects, called ‘non-rigid bodies’ or ‘flexible objects’, and they are found within business areas such as garden farming of fruit and vegetables, laundries and cloth manufacturing etc. These significant industries have a demand for new technology. The development of a combination of accessible soft- and hardware is within reach, so a successful combination has been made possible.
In this project our aim is to develop a robot platform which is able to grasp and manipulate non-rigid objects and to act in relation to experience achieved by memorizing.
This will be achieved by:
- extracting 3D information in real-time and guiding the extraction process through modeling dynamic structures of non-rigid objects
- connecting grasping movements to these dynamic structures
- offering a multiple handling option, where the functionality are to be controlled by visually or tactile feed-back
- an efficient storage of data that the robot has learned
The platform will then be made accessible to companies that have a demand for handling of non-rigid bodies.
