Environmental Engineering Reference
In-Depth Information
6.3.6 Decoupling of Scheduling from Communications
Decoupling of scheduling from communications has been beneficial in sim-
plifying the Rossi X-ray Timing Explorer (RXTE) ground system's planning
and scheduling process and in reducing TOO response time. A similar decou-
pling could also be expected to make the job of onboard scheduling easier and
enable a more dynamic and autonomous scheduling process onboard.
6.3.7 Onboard Data Trending and Analysis
To support smart fault detection, diagnosis, and isolation, more elaborate
capabilities for onboard data trending and analysis will be required. As dis-
cussed previously, a statistical package is needed to perform standard data
analysis procedures (e.g., standard deviation, sigma-editing, chi-squared, etc.)
in order to evaluate the believability of new measurements and/or calculated
parameters and to extrapolate predicted values from past data. The statistical
package will also be needed to support planning of new calibration activities.
6.3.8 Ecient Algorithms for Look-Ahead Modeling
Conventional FSW typically is time-driven realtime software that tries to de-
termine the best course of action at the moment as opposed to the optimal
decision over a future time interval. This is an appropriate approach for stan-
dard FSW applications such as control law processing and fault-detection
limit checking, but is not likely to yield acceptable results in areas such as
planning and scheduling. For such applications, at least a limited degree of
look-ahead modeling will be required to capture the complete information
critical to ecient decision making.
6.4 AI Enabling Methodologies
To implement the Remote Agent functionality discussed in the previous sec-
tion, more sophisticated modeling tools and logic disciplines will be required
than the simple rule-based systems that have been employed in FSW in the
past. In particular, a smart fault detection, diagnosis, isolation, and correc-
tion agent could also utilize a wide variety of AI products to generate and
interpret its results, including state modeling, case-based reasoning, and neu-
ral nets. “Intelligent” constraint-evaluation algorithms used in planning and
scheduling, as well as the data monitoring-and-trending agent, will also need
these AI enabling methodologies. The following discusses onboard operations
that would be enabled through the use of collaborative Remote Agents.
