Descriptions:

Imitation learning has a wide range of applications due to its user-friendly programming without requiring specialized programming skills. Most of the imitation tasks nowadays focus on position or force data separately. However, as the interaction between the robot and its environment becomes more extensive and important, independent learning of position or force limits the realistic reproduction of human-environment behavior.

 

In this work, we would like to investigate learning of constrained manipulation tasks. Based on observed position and force data, human arm impedance should be estimated and used next to motion data to learn tasks like screw tightening out of multiple demonstrations. The aim is to reproduce this process in a more human-like way. Algorithms like Recursive Least Square(RLS), Gaussian Mixture Model(GMM), Gaussian Mixture Regression(GMR) should be investigated to estimate, learn and reproduce the constrained manipulation tasks. Ultimately, the performance of the reproduction procedure should be evaluated in laboratory experiments.

 

Tasks:
• Data gathering using teleoperation

• Impedance parameter estimation

• Impedance parameter and trajectory learning and reproducing with GMM & GMR

• Implementation on real robotic system

• Performance evaluation

 

Requirements:

-Good programming skills in C++ and Matlab

-Basic knowledge in robot control

 

Helpful but not required:

-Basic knowledge in Teleoperation

-Basic knowledge of modeling

-Statistical learning methods like GMM, GMR

Bibliography:
[1] Haddadi, A. and Hashtrudi-Zaad, K. Online contact impedance identification for robotic systems. In Proc. IEEE/RSJ Int. Conf. Intelligent Robots and Systems(IROS), 2008
[2] Calinon, S. and Guenter, F. and Billard, A. On Learning, Representing, and Generalizing a Task in a Humanoid Robot. In IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 37(2):286-298, 2007

 

Supervisors: M.Sc. Shushu Ma