My mission is to reach human performance at fast changing, uncertain, rich and high-dimensional tasks with robots. I believe this goal can be achieved by developing Machine Learning methods, especially Reinforcement Learning, for soft muscular systems.
I pursued a PhD in Machine Learning and Robotics with Jan Peters and Bernhard Schölkopf at the Robot Learning lab within the Empirical Inference dept. During my PhD I interned at X, the Moonshot Factory (formerly Google X). Before, I received a MSc degree in Biomedical Engineering at the Imperial College London and a BEng degree in Information and Electrical Engineering from HAW Hamburg in conjunction with Siemens.
For information and detailed construction details for the 4-DoF pneumatic artificial muscle actuated robot arm please send me an email: dbuechler[at]tue[dot]mpg[dot]de.
Learning to Control Highly Accelerated Ballistic Movements on Muscular Robots
Video highlights the capabilities of robots actuated by pneumatic muscles to 1) generate highly accelerated motions and 2) prevent damage once such fast motions are executed on the real robot. We utilize this properties to tune controllers directly on the real system using Bayesian optimization without additional safety considerations. Data of unstable controllers (the racket motion is unstable in the video) can be incorporate rather than being avoided. Please also see the paper 
Video accompanying our ICRA 2016 paper 'A lightweight robotic arm with pneumatic muscles for robot learning'
Video accompanying our ICRA 2016 paper 'A lightweight robotic arm with pneumatic muscles for robot learning'. This video highlights the fast hitting motions that the four degrees of freedom robot is capable of generating and safely executing. This is realized purely using muscular actuation. This property is especially useful for Machine Learning approaches that explore. With our system we hope to enable exploration in fast motion domains and hence the application of Machine Learning in tasks like smashing table tennis balls.
High-speed and high-acceleration movements are inherently hard to control. Applying learning to the control of such motions on anthropomorphic robot arms can improve the accuracy of the control but might damage the system. The inherent exploration of learning approaches can lead to instabilities and the robot reaching joint limits at high speeds. Having hardware that enables safe exploration of high-speed and high-acceleration movements is therefore desirable. To address this issue, we propose to use robots actuated by Pneumatic Artificial Muscles (PAMs). In this paper, we present a four degrees of freedom (DoFs) robot arm that reaches high joint angle accelerations of up to 28000 °/s^2 while avoiding dangerous joint limits thanks to the antagonistic actuation and limits on the air pressure ranges. With this robot arm, we are able to tune control parameters using Bayesian optimization directly on the hardware without additional safety considerations. The achieved tracking performance on a fast trajectory exceeds previous results on comparable PAM-driven robots. We also show that our system can be controlled well on slow trajectories with PID controllers due to careful construction considerations such as minimal bending of cables, lightweight kinematics and minimal contact between PAMs and PAMs with the links. Finally, we propose a novel technique to control the the co-contraction of antagonistic muscle pairs. Experimental results illustrate that choosing the optimal co-contraction level is vital to reach better tracking performance. Through the use of PAM-driven robots and learning, we do a small step towards the future development of robots capable of more human-like motions.
IEEE Robotics and Automation Letters, Robotics and Automation Letters, 3(4):3161-3168, IEEE, 2018 (article)
Controlling musculoskeletal systems, especially robots actuated by pneumatic artificial muscles, is a challenging task due to nonlinearities, hysteresis effects, massive actuator de- lay and unobservable dependencies such as temperature. Despite such difficulties, muscular systems offer many beneficial prop- erties to achieve human-comparable performance in uncertain and fast-changing tasks. For example, muscles are backdrivable and provide variable stiffness while offering high forces to reach high accelerations. In addition, the embodied intelligence deriving from the compliance might reduce the control demands for specific tasks. In this paper, we address the problem of how to accurately control musculoskeletal robots. To address this issue, we propose to learn probabilistic forward dynamics models using Gaussian processes and, subsequently, to employ these models for control. However, Gaussian processes dynamics models cannot be set-up for our musculoskeletal robot as for traditional motor- driven robots because of unclear state composition etc. We hence empirically study and discuss in detail how to tune these approaches to complex musculoskeletal robots and their specific challenges. Moreover, we show that our model can be used to accurately control an antagonistic pair of pneumatic artificial muscles for a trajectory tracking task while considering only one- step-ahead predictions of the forward model and incorporating model uncertainty.
Our goal is to understand the principles of Perception, Action and Learning in autonomous systems that successfully interact with complex environments and to use this understanding to design future systems