Data-driven Optimal Control of Robots

Optimal control provides a systematic approach to control robots. However, optimal controllers computed using inexact model information lead to significant performance degradation. One way to avoid this limitation is to completely discard model information. However, model-free optimal control approaches are computationally expensive (even inexact models can significantly speed up the optimal control computation).

The goal of this research is to offer a novel data-driven optimal control formulation that uses inexact models to speed up the computation while at the same time avoids significant performance degradation due to the use of inexact model information. Experiments on a three link direct drive torque controlled robot demonstrate the benefits of one such hardware-in-the-loop optimal control method, see Chen and Braun IEEE TRO 2019.

Y. Chen and D.J. Braun, Hardware-in-the-Loop Iterative Optimal Feedback Control without Model-based Future Prediction, IEEE Transactions on Robotics, 2019.

Y. Chen, L. Roveda and D.J. Braun, Efficiently Computable Constrained Optimal Feedback Controllers, IEEE Robotics and Automation Letters, 2018.

F. A. Yaghmaie and D.J. Braun, Reinforcement Learning for a Class of Continuous-time Input Constrained Optimal Control Problems, Automatica, 2018.

Using optimal control, we also investigate how impedance control strategies emerge from first principles of optimality, and how these strategies can be applied to variable impedance actuation. We utilize computational optimal control to devise controllers that exploit the intrinsic compliance and the natural dynamics of robots to achieve better performance. Experiments on the DLR Hand-Arm System developed at the German Aerospace Center demonstrate the benefits.

D.J. Braun, F. Petit, F. Huber, S. Haddadin, P. van der Smagt, A. Albu-Schaffer and S. Vijayakumar, Robots Driven by Compliant Actuators: Optimal Control under Actuation Constraints, IEEE Transactions on Robotics, 2013.