ISBN 978-3-8439-5715-1

978-3-8439-5715-1, Reihe Informatik

Patrick Michael Vonwirth
Natural Control - Compliant Robot Control Featuring Neuromuscular Characteristics

283 Seiten, Dissertation Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau (2025), Softcover, A5

Zusammenfassung / Abstract

Bipedal locomotion is a fascinating field that connects biology with robotics. Despite significant recent advancements in stability and control, humanoid robots still do not match the agility, versatility, and adaptability of humans.

This research uses human gait and morphology analysis to develop more natural control methods for bipedal robots. It focuses on an anthropomorphic bipedal robot equipped with Series Elastic Actuators (SEAs), which provide the characteristic biological skeletal compliance of human bodies. Drawing inspiration from the human neuromuscular system, a new low-level control strategy is introduced that utilizes distributed hybrid impedance control to replicate natural antagonistic muscular control principles.

Additionally, this work introduces so-called Probabilistic Behavior Networks (PBNs) as an effective, suitable high-level control architecture. This architecture enables the robot to manage uncertain control policies and execute multiple tasks simultaneously, improving its adaptability and responsiveness to environmental changes and uncertainties.

Experimental results show that both technologies work well on cutting-edge robotic hardware in real-world scenarios. The proposed control strategy successfully mimics biological movements, representing a significant advancement toward more natural robot control.

By analyzing the achievements and limitations of the software and hardware used, this research provides valuable insights and recommendations to improve natural control in robotics. It lays the foundation for future innovations in natural, agile, and adaptable bipedal robots.