Abstract:
Legged animals show extraordinary versatility in navigating complex environments and overcoming unexpected obstacles.
This agility appears promoted by mechanical compliance intrinsic to their limbs.
Extensive research has been conducted on the importance of elastic compliance for legged locomotion, demonstrating numerous benefits such as improved energy efficiency, enhanced shock absorption, and facilitated power generation.
However, elastic compliance can also produce unwanted vibrations, reduce control bandwidth, and cannot counter energy disturbances.
Although rarely implemented in legged robotics, mechanical damping is a promising solution to these limitations.
Interestingly, muscle fibers exhibit viscous-like damping behavior when suddenly stretched.
This capacity seems dependent on muscle excitation level, indicating that tunable mechanical damping plays a functional role in biological locomotion.
In particular, tunable mechanical damping could facilitate a quick response to sudden perturbations, mitigating errors and delays in sensorimotor information.
This doctoral dissertation investigates principles of tunable mechanical damping in fast perturbed legged-locomotion.
Through quantitative methodologies, it expands previous knowledge of how muscle fibers exploit tunable mechanical damping around touchdown.
Central to this research was the force-velocity relation, a phenomenological function describing viscous-like capacities within muscle contraction dynamics.
Additionally, this thesis explores technical solutions to achieve tunable mechanical damping in legged robotics, combining numerical analysis and hardware experiments.
The content of this dissertation relies on five manuscripts (four peer-reviewed journal articles and a pre-print), the fruit of collaborative research during my doctoral project.
In two computational studies, we could confirm that the force-velocity relation grants muscle fibers tunable mechanical damping during the earliest response to step perturbations (i.e., the preflex phase).
However, we found that current interpretations of this phenomenon require revision.
Without feedforward neuronal modulation, muscle-produced mechanical damping played a minor contribution in regulating the preflex response.
Large impact velocities during reference hopping could further compromise such contribution.
In contrast, tunable mechanical damping produced by the force-velocity relation became a dominant regulating factor when feedforward stimulation was allowed.
In particular, we observed more adjustment of touchdown force and preflex work in response to step perturbations and a simultaneous increase in hopping stability.
Since our computational studies relied on Hill-type muscle models, which are ansatz approximations of real muscle contraction, we conducted an in vitro investigation with realistic boundary conditions to validate our simulations.
This study confirmed activity-dependent damping-like properties in biological muscle fibers.
However, the results showed an initial short-range stiffness phase, which Hill-type models should include for better predictions.
In vitro, the muscle fibers' viscous-like response developed more smoothly and later than what was observed in our simulations, suggesting the need for a better characterization of the force-velocity relation's eccentric side.
Inspired by the biological observations of viscous-like properties within the muscle fibers, we investigated the advantages of incorporating viscous dampers to provide tunable mechanical damping in legged robots.
Using numerical simulations of a robotic leg, we found that viscous damping consistently outperforms Coulomb friction damping in rejecting potential energy disturbances caused by step perturbations.
In contrast, our empirical hardware experiments revealed that damping rate control of a viscous damper that is directly connected to the knee joint fails to generate mechanical damping that can be fine-tuned.
In a follow-up study, we overcame this limitation with a slack-damper mechanism.
This device uses a cable with adjustable slackness to connect the viscous damper to the knee rotation.
Using hopping experiments with various terrain disturbances, we demonstrate that slackness control could effectively and intuitively adjust energy dissipation in a leg prototype, indicating that tunable mechanical damping was achieved.
These experiments also confirmed that embedded mechanical damping causes a trade-off between locomotion energy efficiency and robustness. We argue that the tunability of our slack-damper mechanism and its perturbation-triggered nature make this trade-off more favorable.
The findings of this dissertation support previous evidence that tunable mechanical damping is beneficial for legged locomotion.
We demonstrate that tunable mechanical damping can naturally occur in biological systems due to an intricate interaction between neuromodulation, inner muscle mechanics, and environmental conditions.
In legged robotics, our slack-damper mechanism shows that simple technical solutions are sufficient to implement tunable mechanical damping effectively.
Force-velocity relation
Print-on-Demand