< Back to previous page

Project

Motion recognition based on an invariant motion description for the full three dimensional rigid body motion.

Recently, robots are operating in human-populated environments in a wide range of applications including personal and service robots, intelligent factories with human-robot cooperation and robot-assisted rehabilitation. In many of these applications, it is of key importance  that the robot understands the intention of the human operator. An important step towards understanding the human intention is to  recognize the operator's motion. Beside recognizing the operator's gesture or pose, useful information can be obtained from the three dimensional motion of the object handled by the operator, such as a tool. This project focusses on the recognition of the three dimensional motion of rigid objects such as the operator's tool.

Different demonstrations of the same motion vary in starting position and orientation of the object, the amplitude of the motion, the average execution speed, the motion profile, apart from inter- and intra-personal variations. This variability complicates the recognition of these motions. The impact of the variability is eliminated or at least attenuated by transforming the recorded motions to an invariant motion description before applying algorithms for motion recognition. This transformation will facilitate motion recognition as it leads to a large reduction in search space.

In the proposed project we develop motion recognition approach based on an invariant motion description for the full three dimensional rigid body motion, including both translation and rotations. We will show how the use of such an invariant representation facilitates robotic applications such as programming by human demonstration and human-robot interaction.

Date:1 Jan 2013 →  31 Dec 2016
Keywords:Bewegingsherkenning, Bewegingsbeschrijving
Disciplines:Metallurgical engineering