Submarine Operational Automation System (Late '80s DARPA program)

[Grey Havoc]

2.1 CONCEPT OF OPERATION

Since the advent of machines, the role of the human has become increasingly
to manage the machines and monitor their performance. In the combat environment,
however, the human must not be required to manage and monitor all
submarine systems. Rather, his intent and actions should be monitored and
understood by the machine so that he can be truly supported in assessment,
planning, and plan execution during combat, tracking, and normal underway
operations.

The command centers of future submarines must face an increasingly technically
sophisticated enemy, in superior numbers. While advances in submarines,
sensors, weapons, and other systems address the problem, they also
increase the amount of data and information which the control room/attack
center personnel must assess. In addition, the management of these systems
creates an unacceptable workload at critical mission times. The results
are, potentially, more submarine losses or mission aborts through loss of
situational awareness, and fewer enemy destroyed because the submarine and
its capabilities are not used to their maximum capability and thus offensive
opportunities are lost.

2.1.1 Operational Considerations of Full-Scale System

The SOAS challenge is to assist the submarine commanding officer in the tactical
execution of his mission. This can be accomplished with a combination
of decision aiding, data integration, information display, and problem
focus. The submarine decision maker is typically faced with an excess of
data, much of which is ambiguous, from which he must focus on the critical
issues and make tactical decisions. SOAS will integrate and structure the
data to provide information, not data, to the decision maker; it will help
the decision maker focus on the critical issues through its unique concept
of "dominance"; it will alert the decision maker if routine conditions or
events become constraints; it will provide tactical planning and answer
"what If" questions, and it will anticipate future requirements by drawing
inferences from ownship and enemy options.


SUBMARINE OPERATIONAL AUTOMATION SYSTEM; PHASE I FINAL TECHNICAL REPORT May 31st 1989

6.1.6 Hardware

Reasoning functions in real-time computer processing present a major
challenge to SOAS hardware and software designers. Therefore, the Lockheed
team's approach has targeted the SOAS development toward parallel processing
architectures and high-performance VLSI components. Symbolics
workstations and IRIS graphics workstations were used in Phase I to
facilitate the transfer of the the similar PA architecture into the SOAS
environment for concept demonstration. The PA program, just like the SOAS
program, recognizes that the use of these workstations are inadequate for a
program with large amounts of functionality that runs in real time.

The SOAS choice for the next phases are Solbourne computers - SUN clones
that run in parallel. The Solbourne 4/804 is a fully binary-compatible
with the Sun-4; in fact it runs the latter's SunOS operating system.
Nonetheless, the Solbourne Series4/804 doesn't stop with the complete
emulation. This system adds one significant feature not available from any
Sun system: multiprocessing. The Series4/804 has four Cypress processors,
which function asymmetrical - that is, with one CPU handling the
distribution of tasks and I/O. The primary CPU distributes the workload
among the CPUs and handles any kernel-level tasks such as file activity and
process control or communications. Solbourne includes a fully supported X
Windows and X Windows debugger.

6.1.7 Language

The ease of LISP in quickly prototyping AI solutions has been very useful
in the PA and SOAS programs. However it is recognized that LISP has a
maintenance and performance problem and production problems, so that LISP
is not suitable for real time operational use. For these reasons it is
necessary to abandon LISP.

Candidate replacement languages considered the most seriously were C++ and
Ada. The C++ language offers many of the attributes that are required for
the next phases of SOAS without some of the risks involved in going directly
to Ada. With regard to Ada, based on a realistic assessment, it is
naive to expect that all the risks and problems associated with the
marriage of Ada and real-time artificial intelligence are going to be
resolved in the next few years. Therefore, the Lockheed team will take an
incremental approach using C++.

Progress has begun in the translation of some of the SOAS subsystems to
C++. The Pilot's Associate program has similar requirements to go to a
real-time system language. The SOAS SA subsystem is over 50% translated.
The KADET tool without the SOAS enhancements will be translated by the end
of June 89.


Ibid