PhD Defense : Bio-inspired locomotion assistance and symmetrization using adaptive motor primitives by Henri LALOYAUX
As life expectancy increases worldwide, more people are affected by gait pathologies, caused either by the natural decline in locomotor functions, or indirectly.
Last decades have seen emergence of promising technological solutions to assist their walk, such as lower limbs exoskeletons.
There exists a large diversity of strategies to control these devices, among which some are said to be bio-inspired as their purpose is mimicking control laws identified in living bodies.
This thesis presents one of these neural strategies, i.e. motor primitives.
These primitives have been identified as fundamental signals which reconstruct muscle activation signals during diverse locomotion tasks.
Two methodologies are presented to extract such primitives from a database spanning over several walking conditions.
Importantly, primitives are here restrained to Gaussian-like shapes, with potential asymmetry.
Besides leading to more biologically relevant patterns, such constraints should also simplify a later control strategy built on them.
We show that four to five primitives might be enough to reconstruct muscle stimulations and hip moments across several locomotion tasks.
In our first experiment, we validate the combination of simplified primitives and a musculoskeletal model for assisting healthy subjects with a hip exoskeleton.
This framework showed adaptation to the user’s gait for different slope inclinations and proved its relevancy for ensuring smooth transitions between tasks.
Other experiments are reported in this thesis, where a novel type of symmetrization algorithm is presented.
Leveraging on the low amount of parameters of the primitives-formalism, a hip exoskeleton is controlled through this strategy.
Both types of planar walking were tested, as well as the relevancy of locking the symmetrizing adaptation mechanism of this algorithm.
This was the opportunity to assess the potential of integrating a musculoskeletal model into this symmetrization algorithm, and led to contrasting results between both types of assistance.
Jury members :
- Prof. Renaud Ronsse (UCLouvain, Belgium), supervisor
- Prof. Grégoire Winckelmans (UCLouvain, Belgium), chairperson
- Prof. Paul Fisette (UCLouvain, Belgium)
- Prof. Philippe Lefèvre (UCLouvain, Belgium)
- Prof. Virginia Ruiz Garate (Mondragon Unibertsitatea, Spain)
- Prof. Maarten Afschrift (Vrije Universiteit Amsterdam, The Netherlands)
Video conference link : https://teams.microsoft.com/l/meetup-join/19%3ameeting_NGMxMDYyOTMtYzJiNy00MmMyLTg2YTItZGFkY2I2Mzc1MDAy%40thread.v2/0?context=%7b%22Tid%22%3a%227ab090d4-fa2e-4ecf-bc7c-4127b4d582ec%22%2c%22Oid%22%3a%2283a4aaf9-c63b-4923-afe4-863490bb58f0%22%7d
