
Contact Us
Company History
Application
Summaries
Mailing List
Neural
Net Reference
Customer Quotes
Customer Forum
Search

|
CONCLUSIONS We believe this Phase I SBIR project was highly
successful. We met and exceeded all of our primary goals.
In particular, we accomplished the following major tasks:
- We designed, developed, and
integrated a software and hardware environment to
implement and test adaptive inverse control for
ventilation. We built a skeleton ventilator, an
electronic interface to the ventilator, and
integrated this with an off-the-shelf
microcontroller. We implemented the ventilator
interface software in the microcontroller, the
neural primitives required to test the AIC
methods in the microcontroller, and the windows
OLE component necessary to remotely control
NeuroSolutions and communicate with the
microcontroller. This environment can be expanded
and commercialized to produce a general purpose
neural control package.
- We simulated 7 different control
architectures and implemented 4 different control
architectures on our ventilator test setup. For
the better models we also implemented a
disturbance canceling mechanism that can handle
the breath-to-breath differences common in
patients. The ventilator test setup is the same
as that used by physiologists studying
spontaneous breathing. The control architectures
included: standard FIR AIC, gamma memory AIC,
Laguarre memory AIC (simulation only),
Physiologic model AIC, predictive switching
mixture of experts AIC, tap-delay neural network
AIC (simulation only), and the self-organizing
map based predictive switching multiple model
controller (simulation only).
- We tested the different control
architectures using a physiologically relevant
criterion (work of breathing) and against the
market-leading Puritan Bennett 7200 ventilator.
Both the Physiologic AIC and Predictive Switching
Mixture of Experts AIC methods outperformed the
PB7200.
- We are in the process of filing a
provisional patent on our physiologic AIC method,
and have already filed two provisional patents on
our goal of using these techniques for the
higher-level goal of setting many of the
ventilator settings automatically.
Ventilator
Home
|