Experiments for the online closed-loop control of neural prosthetics require feedback within 100ms. In a typical neurophysiology laboratory with local computing machines, a majority of this time is spent on acquiring and analyzing the neural signals and a minority (i.e. less than a millisecond) is actual data transfer among machines on local- or campus-area networks. However, the local computing machines may not offer the computational resources necessary for running complex algorithms or scenarios that have been recently proposed. While scientists can take advantage of remote computing resource providers, wide-area networks present much larger latencies that can affect an online experiment. This work presents a split modeling approach that allows the execution of a controller on the neurophysiology resource and the execution of computationally intensive modeling and adaptation algorithms on a remote datacenter, even with the inevitable network latency. Simulation results are presented to quantify how the accuracy of the controller is affected by the split modeling approach in the presence of delays, and to demonstrate that scientists can take advantage of remotely available massive resources.
[1]
José Carlos Príncipe,et al.
Brain-Machine Interface Engineering
,
2006,
Brain-Machine Interface Engineering.
[2]
Weifeng Liu,et al.
Kernel Adaptive Filtering: A Comprehensive Introduction
,
2010
.
[3]
B. Widrow,et al.
Adaptive noise cancelling: Principles and applications
,
1975
.
[4]
S. Haykin,et al.
Adaptive Filter Theory
,
1986
.
[5]
Renato Figueiredo,et al.
Model development, testing and experimentation in a CyberWorkstation for Brain-Machine Interface research
,
2010,
2010 Annual International Conference of the IEEE Engineering in Medicine and Biology.
[6]
Nicholas G Hatsopoulos,et al.
Incorporating Feedback from Multiple Sensory Modalities Enhances Brain–Machine Interface Control
,
2010,
The Journal of Neuroscience.