This project is focused on researching about learning techniques and manipulation of objects and the development of a set arm-hand that allows robots to get progressively closer to handling capacities offered by the arm-hand set of human beings.
The design of a arm-hand set does not allow to reach the capacities expected by itself because its effectiveness requires even more progressively advanced perception methods and also advanced learning and planning methods. The objectives proposed in this project are:
- Development of a light arm-hand set highly anthropomorphic that allow to be used in a large number of handling operations.
- Design and development of characterisation methods of the affordances developed by human being when grasping and handling objects and simple tools.
- Design and development of learning methods that perform learning or capturing human hand gestures to reproduce them afterwards on similar objects and conditions.
- Design and development of planning methods for manipulation that allow to combine learned skills, activate them in precise moments and react to obstacles and unpredicted situations detected by the sensory systems of the robot.
The project is divided on two well differentiated parts:
- Development of very lightweight manipulator that enables to attach an advanced hand in a way that the whole arm-hand set has a weight around 12 -13 Kg with a large manipulation capacity. That involves the development and construction of a lightweight 7 DOF manipulator in which an advanced commercial hand similar to human being and light will be attached.
- This project will tackle the development of learning and planning methods that permits to use in the most efficient way the arm-hand set to handle and manipulate objects effectively. This arm would be installed in the mobile manipulator MANFRED-2.