Automated control of the U.S. Navy’s submarines — a great technological challenge, is the goal of a new DARPA contract for a submarine operational automation system (SOAS). The contract will begin Phase 2 of an effort to use advanced computer software, including so-called artificial intelligence and neural networking techniques, to achieve a “new dimension” of automated control of submarine systems and subsystems. It is aimed ultimately at helping submarine commanders cope with an increasingly complex tactical environment.
Unlike the UUV program which will prototype existing technology in a new application for near-term transition to the Navy, the SOAS program reflects DARPA’s traditional approach of focusing on cutting-edge, high-risk technology that may offer new payoffs far into the future. With its decision to focus on a “framework for the future” that would look beyond today’s technology, DARPA rejected arguments of experts in submarine technology and advanced software, both in industry and government, that a variety of currently operational systems can meet the Navy’s need for a totally integrated command decision support system.
The initial step towards SOAS development was DARPA’s award — in January 1989 – of Phase 1 contracts for a technology development plan and demonstration of capabilities, on the BSY-1 control system for newer LOS ANGELES-class attack submarines, and a closer integration of weapons and sensors aboard both attack and OHIO-class TRIDENT ballistic missile submarines.
Automating Submarine ControL
DARPA’s approach to SOAS is based on the recognition that the integration of the hardware, software, and the skills of the human crew of the submarine is of paramount importance. Automating submarine control is necessary in order to enable submarine commanders to get “better decisions quicker,” to cope with the vast complexity of both external and internal submarine operations. Overall submarine control requires a level of situation awareness beyond that needed by any other weapons system. A submarine is like a giant factory, consisting of more than 400 subsystems within the specific areas of acoustic and electrical systems, machinery, hydrodynamics, control surfaces, weapons control, and weapon systems that are monitored by many kinds of sensors. Systems employing highpressure compressed air, nuclear power, and sophisticated electronics operating within the closed atmosphere of the submarine, and the normal isolation of submarines on patrol, out of reach of emergency assistance, underline the urgency of making the commander’s job easier.
Additionally, thousands of elements of so-called “historical data” on ship systems, sea conditions, geography, and other areas that reside in sensors scattered throughout the submarine must be reconciled with “emergent” data that are received continuously by acoustic, communications, and other systems.
Tactically, submariners must be able to respond to threats posed by quieter, faster Soviet submarines with more capable weapons as well as to rapid changes in the tactical environment caused by more sophisticated U.S. Navy weapons and sensors. Sub commanders and their crews must be aware of weapon and ship system status while attempting to evade the enemy and be prepared to operate in a degraded status in case of flooding, fire, and equipment failures. Combat is an added load on the commander, who must simultaneously both “fight the ocean” and be aware of all continuous changes in the condition of internal systems.
Meanwhile, continuous improvements in the access to data, for example, by the Navy’s development and deployment of: the BSY-1, (the BSY-2 is planned for next-generation, SEA WOLF-class attack submarines), the CCS Mk 2; and other incremental improvements in system control — result in a dramatic increase in data available, which nonetheless may still be ambiguous in the submarine environment. Two of the primary challenges of submarine control that form the basis for the SOAS requirement are the uncertainty of the data provided to the submarine crew and the need to develop a precise understanding of the external and internal “worlds.”
Access to massive amounts of information — often incomplete or ambiguous — from many different sources, requires the commanding officer and his team to “paint a picture” of the composite tactical and system situation by assimilating the output of multiple sensors and computers. The task of assimilating data is made more challenging by the distribution of submarine control responsibilities among a command structure — the CO’s team — whose members are at stations in different parts of the ship.
Among the long-range goals of SOAS are a tenfold reduction in tactical scene description errors, a two-thirds reduction in target classification time, and a doubling of weapons effectiveness.
Several similarities exist between decision-support systems for aircraft pilots and submarine officers, such as the requirement for artificial intelligence-based expert systems that integrate data of subsystems and provide the pilot – or sub commander — with sets of options he may elect to take. Three key differences exist:
• The level of uncertainty is greater in a submarine than in an aircraft;
• The time frame is more compressed for the pilot than the submarine; and
• While a decision support system will serve as a “ghost backseater” to the pilot, the sub commander has a team of skilled officers with whom he consults on decisions.
In an AI (Artificial Intelligence)-based decision support system, the sub commander would have access to– in addition to the tactical and ship system data provided by dedicated computers and displays — a comprehensive, real-time analysis of the tactical situation and all ship control operations that provide decision choices based on available data. The commander will decide whether or not to take the system’s “advice” — which automatically will be revised accordingly.
DARPA declines to comment on the precise scope of the submarine automation system which remains in a concept exploration stage. However, officials say, the system could initially be introduced “in parallel” with existing systems and augmented with an “advisory” function. Rather than follow the conventional approach to developing an AI-based automation system, in which experienced sailors are quizzed about their operational requirements by software engineers who then try to transform the “real life” anecdotes into Al-based programs, the company teaches AI principles to the submariners, who then are able to produce programs that incorporate a greater depth of experience, and a truer reflection of their requirements.
A company developed a series of software algorithms that translates AI developed rules into geometric forms which permit real-time calculations of submarine system data, including operational data on range, bearing, and motion, and display decision options to the submarine commander in the form of geometric images.
Submarine Automation Systems “could employ a significant AI element to help with decision-making.” It could also include more modest goals, such as speeding up the operations of individual systems by making them more “user-friendly- -for example, by assisting crewmen in loading and presetting weapons.
Another “near-term” program submitted as a SOAS candidate is a so-called “generic shell” that is designed to serve as the foundation for a variety of command decision support systems.
As a generic system, analogous to a computer “spreadsheet” program that can perform a wide variety of tasks, the architecture “represents a way you can cast a large number of [command decision support] problems.” The shell “would be applicable across the Navy’s command hierarchy and across warfare areas.” Several elements of the system — tactical picture generation, tactical forecasting, and part of the situation assessment function — are “up and running.”
“Evidentiary” reasoning is the process of drawing conclusions based on available data, or evidence on specified problems such as the status of ship systems or tactics. Rule-based reasoning is the reflection of established Navy doctrines of operations — for example, “If the enemy takes a certain action, shoot him.” Spatial rules are based on locations of objects, such as ships, and distances between them. Statistical reasoning employs conventional mathematical or probability calculations.
A key technology challenge is development of an “explanation facility” that will provide a coherent rationale for the decision recommended by the decision support system. ‘”The commander isn’t going to be satisfied with The computer says so’ as the explanation.”
A Long-Term Program
DARPA’s development approach to SOAS is to “catve off the hard parts and do them first.” In the initial year of the contract, focus will be on designing the high-level architecture of the decision support system, which will manage such taskoriented functions as system assessment, tactical planning, mission planning, and mission execution, among others. Subsequently, in the out-years, individual sub-systems such as weapons and sensors will be simulated in order to test the design concept and performance of the system architecture. Additional subsystems will be added in a step-by-step process, and subsystems also will be “coupled” within the system.
Initially, the DARPA SOAS program is expected to last four to five years, allowing the government and industry to “make inroads” on the problem. MeanwhiJe, the complexity of the submarine, the threat environment, and the demands on manpower and training all are changing. The entire scope of the program is still in a conceptual study stage. DARPA is looking into the future, at the hard, long-term problem. “The program is basically software – hardware is changing too fast.” But “you don’t do automation for the sake of automation. You do it in order to gain something in terms of performance. DARPA [with SOAS] isn’t interested in upgrading existing systems — it’s trying to establish a context for future thinking for a total system.”
