Co-Adaptive Control of Bionic Limbs via Unsupervised Adaptation of Muscle Synergies

Abstract

In this work, we present a myoelectric interface that extracts natural motor synergies from multi-muscle signals and adapts in real-time with new user inputs. With this unsupervised adaptive myocontrol (UAM) system, optimal synergies for control are continuously co-adapted with changes in user motor control, or as a function of perturbed conditions via online non-negative matrix factorization guided by physiologically informed sparseness constraints in lieu of explicit data labelling. Methods: UAM was tested in a set of virtual target reaching tasks completed by able-bodied and amputee subjects. Tests were conducted under normative and electrode perturbed conditions to gauge control robustness with comparisons to non-adaptive and supervised adaptive myocontrol schemes. Furthermore, UAM was used to interface an amputee with a multi-functional powered hand prosthesis during standardized Clothespin Relocation Tests, also conducted in normative and perturbed conditions. Results: In virtual tests, UAM effectively mitigated performance degradation caused by electrode displacement, affording greater resilience over an existing supervised adaptive system for amputee subjects. Induced electrode shifts also had negligible effect on the real world control performance of UAM with consistent completion times (23.91±1.33 s) achieved across Clothespin Relocation Tests in the normative and electrode perturbed conditions. Conclusion: UAM affords comparable robustness improvements to existing supervised adaptive myocontrol interfaces whilst providing additional practical advantages for clinical deployment. Significance: The proposed system uniquely incorporates neuromuscular control principles with unsupervised online learning methods and presents a working example of a freely co-adaptive bionic interface.

Simultaneous and Proportional Real-Time Myocontrol of Up to Three Degrees of Freedom of the Wrist and Hand

“…Certain academic efforts have been dedicated into tackling this issue [16]- [18] however, a clinically viable solution is yet to be found. At the same time, the induced controller instabilities lead to a lack of predictability which is paramount for establishing an engaging human-machine interface [19]- [21].…”

Unsupervised Myocontrol of a Virtual Hand Based on a Coadaptive Abstract Motor Mapping

“…Finally, they assigned muscle synergies to DoFs by observing which synergy was most active while activating each DoF. The work by Yeung et al [5] extends the calibration procedure described above, accounting for the evolution of muscle synergies over time due to the subject's familiarization with the myocontrol system and the displacement of electrodes, among other factors. They accommodate for those changes by employing an adaptive version of NMF with sparsity constraints and a forgetting mechanism that progressively discounts the contribution of old input samples.…”

“…The system adaptively decomposes muscular control inputs into sparse muscle synergies using a purposedly designed incremental NMF algorithm with sparsity constraints and a forgetting mechanism to discount the contribution of old input samples. Although derived independently, our formulation is similar (but not equal) to the NMF algorithm used by Yeung et al [5], which was published during our paper's final redaction. In contrast to their approach, this algorithm is used to implement an abstract motor mapping between the synergies' activations and a set of desired actions of the hand or wrist that may not be physiologically related to those activations.…”

Progressive unsupervised control of myoelectric upper limbs

“…Unsupervised myocontrol is a desirable alternative to supervised myocontrol, as it eliminates the need for hard-to-obtain labeled training data. Existing unsupervised myocontrol approaches derive lowdimensional approximations of the muscular input, corresponding to distinct muscle coactivation patterns, and employ them as control commands for the kinematic or kinetic variables of interest [19][20][21][22]. This is based on the neuromotor control principle that the human nervous system efficiently realizes movement by recruiting and coordinating nonredundant muscle synergies [23][24][25].…”

“…Yeung et al [21] designed an adaptive version of the paradigm by Lin et al [20], in which the factorization model was automatically updated during operation to account for changes in muscle synergies caused by the nonstationarity of sEMG and the user's adaptation to the myocontrol system. The same quasi-unsupervised calibration procedure was followed to build a myocontrol model for a prosthetic wrist, which involved performing specific actions in an unstructured manner and manually defining a biomimetic motor mapping between muscle synergies and wrist actions.…”