Humans appear noisy in their motor capabilities, but are not uniformly so. There is emerging evidence of underlying structure to this noise. Humans exhibit lesser variability in parameters that directly affect the outcome of the task than in task-irrelevant parameters. Can the nervous system directly alter the covariance of motor noise across multiple muscles? Until our study, it was not possible to rule out other factors such as task mechanics and fatigue driven trial-to-trial variability. Using fine-wire electromyograms recorded from all muscles of the index finger, for a static finger pressing task, we found that there is indeed neurally modulated structure to the variability of muscle output. Muscle combinations that contributed to the instructed force direction had lesser variability than those that did not. For both machines and humans, open questions remain on whether reducing variability in actuation comes at a cost, energetic, performance or otherwise. In tasks that involve working near boundaries of instability, actuator noise can play a key role in monitoring whether the system is close to failure or not. Noise reduction is therefore not always beneficial.
Ongoing projects include:
- Bifurcation detection for tactile manipulation.
- Noise mitigation in jumping.
- Strategy selection to minimize noise propagation.