3. This is an attractive solution for unmanned vehicles
requiring a high level of reliability, very low SWAP
and the ability to fail to a safer, fail-silent condition
following most first failure conditions.
For vehicles requiring higher reliability, another
concept was developed by combining the enhanced
rapid start concept (Figure ) and the minimal system
with self-test
4.
5. SELECTED CONCEPT AND
APPLICATIONS
The conclusion of the system concept
development process was that several
solutions can be built from the Rapid
Start/Restart SCP building block (RSCP)
shown in Figure .
6. FURTHER DEVELOPMENT
These concepts look promising but much
needs to be done to make them a reality. In
the next phase of this research project it is
intended to build a demonstration system.
This will be used to demonstrate rapid start-
ups in response to simulated fault conditions.
The demonstration will require both hardware
and software development
A prototype RSCP building block will be
designed and built
7. For the final system, the RSCP building block hardware
must be very carefully designed to avoid introducing any
single points of failure that might cause it to not fail-silent.
Electrical fault containment regions and methods used to
prevent the propagation of faults may be required. The Navy6
Sea wolf ship control and the X-38 NASA crew return
vehicle fault tolerant computer [5] are examples of solutions
previously developed by Draper that utilize these fault tolerant
electrical design principles
8. Demonstration
Injecting simulated faults in the powered
RSCP and demonstrating that the
unpowered back-up
RSCP can start-up rapidly and assume
control.
Injecting simulated common cause faults
and
demonstrating that restarts will allow the
system to
recover
9. CONCLUSIONS
The use of rapid starts and restarts looks very
promising as
the means to reduce the SWaP of highly reliable
systems
and to make them more robust to transient faults
and
common cause faults
Developing such systems will be challenging
because of the fast start-up times required and the
difficulties associated with saving and restoring
state data