Learning Options for Compositional Motor Control with Adapter Banks
Abstract
Learning flexible motor primitives is a hallmark of skilled motor control. Recent neuroscience theory proposes that motor primitives may be implemented as low-rank perturbations of a shared recurrent network, but leaves open how such a system is learned. We translate this principle into a novel architecture for learning motor skills end-to-end: a shared recurrent core modulated by a bank of residual adapters, each selected by a discrete latent code. Trained on closed-loop biomechanical control, the adapters develop emergent low-rank perturbations of the recurrent dynamics despite no architectural rank constraint, placing task representations in disparate subspaces of the shared core network. A simple high-level policy over the learned options, optimized while the whole network is frozen, sequences the low-rank adapters to produce novel out-of-distribution movements. We demonstrate the ability to generalize to novel motor sequences within the closed-loop control setting, improving on the generalization error of a task-input-conditioned multitask baseline by upto order of magnitude.
Related
- Stabilizing Variables in Motor Control and Learning: A Control-Theoretic Account
- Neural basis of compositional control
- A Unified Neurocomputational Framework for Closed-Loop Motor Control and Sense of Agency
- Age-related shift towards feedforward motor control under uncertainty
- A pipeline with pre-processing options to detect behaviour from accelerometer data using Machine Learning tested on dairy goats
BibTeX
@misc{kumar2026learning,
title = {{Learning Options for Compositional Motor Control with Adapter Banks}},
author = {Sreejan Kumar and Marcelo Mattar and Lea Duncker},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.17042},
eprint = {2609.17042},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.17042},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Learning Options for Compositional Motor Control with Adapter Banks
AU - Sreejan Kumar
AU - Marcelo Mattar
AU - Lea Duncker
PY - 2026
DA - 2026/09/15
PB - arXiv
DO - 10.48550/arxiv.2609.17042
UR - https://arxiv.org/abs/2609.17042
AB - Learning flexible motor primitives is a hallmark of skilled motor control. Recent neuroscience theory proposes that motor primitives may be implemented as low-rank perturbations of a shared recurrent network, but leaves open how such a system is learned. We translate this principle into a novel architecture for learning motor skills end-to-end: a shared recurrent core modulated by a bank of residual adapters, each selected by a discrete latent code. Trained on closed-loop biomechanical control, the adapters develop emergent low-rank perturbations of the recurrent dynamics despite no architectural rank constraint, placing task representations in disparate subspaces of the shared core network. A simple high-level policy over the learned options, optimized while the whole network is frozen, sequences the low-rank adapters to produce novel out-of-distribution movements. We demonstrate the ability to generalize to novel motor sequences within the closed-loop control setting, improving on the generalization error of a task-input-conditioned multitask baseline by upto order of magnitude.
ER -