Utilizing Natural Dynamics for Reliable Legged Locomotion

ended

Despite the significant progress made in the field of humanoid robotics over the last 10-15 years, bipedal locomotion in robotics is still far from human performance in terms of speed, versatility, and robustness. The design of most humanoid robots nowadays is dominated by the aim at high rigidity and position accuracy in the motor units. In contrast, the NatDyReL project aims at a fundamental shift of paradigm in the design and control of humanoid robots, towards a new generation of intrinsically compliant robots that can adjust their open loop actuator impedance in real-time to the task. We believe that the maturing technology of variable impedance actuators in combination with novel control approaches for the intrinsically elastic dynamics has the potential of bringing humanoid locomotion and multi-contact motions to a new level in terms of energy-efficiency and execution speeds more similar to the human archetype. However, to fully utilize the ultimate benefits promised by variable impedance actuators, i.e. to store and release energy as well as to provide physical protection against shocks caused by impacts, it is necessary to exploit the natural compliant whole body dynamics on all levels of the system design, planning and control hierarchies. This project follows two scientific tracks for achieving (a) energetically efficient and high performant legged locomotion and (b) robust and dynamic contact transitions and in-contact motions for whole body locomotion in uncertain and confined spaces. As a strong basis to the mentioned application oriented objectives, we also aim at fundamental contributions on the control challenges related to novel variable impedance actuator technologies. The project is expected to make a strong impact on bipedal humanoid locomotion. Moreover, the developed methods will be sufficiently general such that they can also be transferred to other morphologies such as e.g. multi-limbed walking or climbing robots.