KNOBS: An Experimental Knowledge Based Tactical Air Mission Planning System and a Rule Based Aircraft Identification Simulation Facility
Proceedings of the Sixth International Joint Conference on Artificial Intelligence. .
© 1979 International Joint Conferences on Artificial Intelligence
Abstract
KNOBS (from Knowledge Based ~stem), proper, refers to an experimenta~ sy~tem u~der development at the MTiRE Corporation to explore the applicability of ar~ific1al ~ntell~ge~ce techniques to the implementation of an automated, extremely flex1ble tact1cal.m1ss1on planning consultant. !his pap~r will discu~s tha~ system and then another wh1ch supports the rule based simulat1on of a1rcraft ident1ficat1on strategies.
- Garvey, T. D. , Fischler, H. A., Hartin, R. A. , and Sinko, J. W., "Machine Intelligence Based Multi-Sensor ESH System", Interim Technical Report, SRI International, Menlo Park, July 1978.
- Pretty, R. T., (Ed.), Jane's Weapon Systems 1978, Franklin Watts Inc., New York, 1977.
- Roberts, R. B. and Goldstein, I. P., "The FRL Manual", HIT AI Report 409, Cambridge, June 1977 .
- Rosenschein, S. J., "Structuring a Pattern Space, with Applications to Lexical Information and Event Interpretation." Doctoral Dissertation, University of Pennsylvania, Philadelphia, 1975
References were extracted automatically from the PDF and may contain errors.
Full text extracted automatically from the PDF; figures, tables and equations may differ. The PDF is the version of record.
1. MISSION PLANNING:
Tactical mission planning may be posed as the intelligent utilization of one's I available tactical air resources. Projections of possible theater level tactical conflicts indicate an expectation of very intense, complex, and time sensitive planning requirements as well as a requirement to improvise and implement novel tactical doctrine in response to unforeseen modes of weapon systems employment by the enemy. It is our thesis that knowledge based techniques represent the most promising approach towards the achievement of such trainable software. Work has concentrated on the problem of planning interdiction attacks, particularly in the choice of attack aircraft, munitions, and the electronic protection required to gain those aircraft safe passage. Currently the system is undergoing intense detailed design, while parts have been implemented in INTERLISP. These latter are sufficient to serve frequent "advocacy" demonstrations of the flexibility, human legibility and system self-explanation capabilities associated with knowledge based systems.
1.1 DATA · BASE: There are basically two data bases; one for resources, the other for targets. Targets are to be stored in a hierarchical data base represented as list
This work was sponsored by the Directorate of Mathematical and Information Sciences of the Air Force Office of Scientific Research, the Information Sciences Division of the Rome Air Development .Center, and the MITRE Corporation Independent Research Program.
structure in the INTERLISP virtual storage. Individual targets are to be stored as FRL (3] frames. At this writing, we are involved in the translation of FRL from MACLISP into INTERLISP. The individual target frames are in fact depositories of simple record type information, with the procedures to be invoked when such values are needed or changed being derived by inheritance chains through more generic "target class" frames.
1.2 KNOWLEDGE REPRESENTATIONS: The system will employ two forms of knowledge representation; production rules and frames. The rules are chained backwards in a deductive manner to manage such generic choices as aircraft, weapons, support, and electronic counter measures. Software has been implemented for the interpretation, the translation (in both directions) betw~en the internal form of the rules and the user legible/writable form, the explanation of deductions (in terms of the entire chain of rules and data employed in a deduction), and the creation and editing of rules.
A typical (MYCIN-like) rule for the selection of munitions is:
--TR17--
IF:
1: THE target IS PINPOINT
AND 2: THE CLOUD COVER OF THE target IS LESS THAN 2/8
AND 3: THE time on target IS WITHIN DAYTIME
AND 4: THE AAA DEFENSE OF THE target IS MORE THAN LIGHT
AND 5: PAVE KNIFE IS AVAILABLE AT time on target
THEN:
1: THE BEST MUNITIONS FOR THE target IS ONE OF: MK-82-LG, MK-83-LG, MK-84-LG
Lower case indicates variables. The justification for this rule can be inferred from the discussion in [2), p. 782.
The syntactic pattern matcher employed by the inference mechanism is quite general. It includes synonym and spelling correction capabilities, as well as provision for restrictions on variable instantiations defined by arbitrary multivariate predicates. This matcher is also employed to implement a flexible command-query system.
Frames will be employed to represent both the data base and, at a higher level, missions and support sub-missions. Frames are particularly attractive with respect to the time sharing of support missions, (e.g., reconnaisance, refueling, and SAM suppression) among several attack missions.
- EXAMPLE: The following dialog (user lines start with "&&") is possible in the current system:
&& WHAT IS THE HARDNESS OF TARGET 2? THE HARDNESS OF TARGET 2 IS UNKNOWN && HOW ABOUT TARGET 1? TARGET 1 IS VERY HARD && WHY?
BY TR3: TARGET 1 IS VERY HARD SINCE:
1: DATA: TARGET 1 IS A BRIDGE
&& IS TARGET 3 AN AIRBASE?
NO, TARGET 3 IS AN SA-4
&& CHOOSE MUNITIONS FOR IT
THE BEST MUNITIONS FOR TARGET 3 IS SHRIKE
&& WHY?
BY MR4: THE BEST MUNITIONS FOR TARGET 3 IS SHRIKE SINCE:
1: BY TR1: TARGET 3 IS AN EMITTER SINCE:
1: BY TR4: TARGET 3 IS A SAM SINCE:
1: DATA: TARGET 3 IS AN SA-4
For more information about SA-4's, SHRIKEs and why one is bad for the other, see (2] pp . 77 and 150.
2. AIRCRAFT IDENTIFICATION:
A critical\ contemporary military need is for a method of sure, safe identification of aircraft, capable of reliable distinction between friend and foe (IFF). We shall concentrate here on approaches to the IFF
problem based primarily on "cooperat i ve" techniques . It seems possible today to build electronic query and response systems tha t almost solve the problem electronically. Difficulties, though, stem from the facts t hat in certain theaters it is ant i cipated t hat an aircraft would be required to pass thousands of such query-response tests -- and pass all of
them to survive, that equipment can fa il completely, that a spread spectrum signal mi ght be detected if it happens to be near enough to and in the direction of an enemy rece i ver which is afforded sufficient integration time, and that jamming is in fact quite often possible . In addition, tomorrow's war must be as silent as possible , for in querying an aircraft fo r identification, the inquirer exposes both himself and the plane, should he be a friend . Therefore one should not normally query without the intent to fire. It seems likely that the query should also transmit (crypto-securely) information such as the position and the weapon the inquirer is contemplating firing . The aircraft may then use that information to decide whether or not to respond on the bas i s of all the threats he has perceived or inferred. The logic of the IFF problem thus becomes one of evaluating procedural strategies in a collusive, n-person game . Such evaluation is proving error-prone as more complex IFF "solutions" are being proposed and mo r e thorough analyses demanded.
- WHAT SHOULD BE DONE : Our first response was to initiate the design of a rather genera l knowledge based IFF game definition and execution system. In fact, there a re are a number of "electronic warfare" problems tha t share this same "gaming" characteris ti c : electronic counter and counter-counter measures, the ~xploitation or defeat of the enemy's surveillance or intelligence structure , the maintenance of survivable communicat i ons and/or surveillance networks, or the penetration by an interdiction aircraft of the enemy's defenses. For a knowledge based approach to this last, see Ll]. While we ar e stressing the IFF problem in this paper , it was our intent that the game definition/execut ion mechanism be sufficiently general to suppor t study of the logic of many of these "electro magnetic warfare" problems.
- 2 WHAT WE HAVE DONE : KNOBS has been augmented by the addition of a forward cha i ni ng production system in which the conflict set resolution algorithm chooses the minimal set of rules that covers the max imal set of facts (as discussed in [4]) . This has been employed t o create a simulation of the world as i t mi ght be
seen by a single "Blue" surface-to-air missile (SAM) s ite. The SAM has a number of choices to make: whethe r to turn its radar on or off, whe the r or not to query each plane it has de t ec t ed, and whether or not to fire at such a pl ane. Our initial thought was to model the SAM '~ strategies for these choices by a set of produc t i ons. What we have done, in fact, is to wr i LP a numbe r of production sets, basi cally onp for ea ch ilPrativP step in the game, and to usP the production system interpreter as the mechan i sm for the entire simul a tion . This rathe r s imple idea turns out to be quite powe rful . The resulting simulations, in which not j ust the strategies but also the weapons sys t ems l imitations and even the "physics" are rep rPsented by human legible rules, yield a s t rong sense of cause and effect between hypo thes ized principles of behavior and game resul t s.
- ?~HPL~_ RULE : The following is a quite conse rvative rule meant to express a de termi na tion not to query an aircraft unless i t is pe r ceived as a possible direct threat to the SAM:
--QUERY1-
IF:
I :
AND 2:
AND 3:
AND 4:
AND 5:
AND 6:
RADAR IS ON
SAM liAS MISSILES LEFT
?A/C liAS NO~ BEEN FIGURED
? A/ C liAS NO'f BEEN QUERIED
?A/C IS LOW
?A/C IS FLYING WEST
THEN :
I: QUERY ?A/C
Identi f i e rs starling with "?" exi s tentially quantified variatles. represent
- 4 SAMPLE RUN: The listing below is an excerpt from an ac tual game run. We have prese rved just enough to trace the diss imilar f a t es of two pl anes, P0004, a Blue F-111F, and P0008, a Red HIG-21, with ra ther similar histor i es :
STARTING TURN # 1
AIRCRAFT P0004 OF TYPE F-111F HAS ENTERED RANGE
AIRCRAFT P0008 OF TYPE HIG-21 HAS ENTERED RANGE SAM QUERIES P0004 FIGURE P0004 IS RED Sl FIRED AT P0004
SAM QUERIES POOOR FIGURE POOOR IS RED SJ FIRED AT POOOR
STARTING TURN #2
STARTING TURN #3
MISSILE S1 HISSED
MISSILE S3 HISSED
S5 FIRED AT POOOR
STARTING TURN #4 P0004 --A F-111F PASSED THROUGH SAFELY
STARTING TURN # 5
MISSILE S5 HIT
P0008 -- A HIG-21 DESTROYED
By reference to the rules and some descriptive data, it is very easy to understand what happened. Both planes aroused the SAM's susp1c1on since they were flying WEST and LOW . Both therefore got queried. Neither responded: the MIG because it was Red and therefore did not know the code key and the F-111 because its IFF equipment was broken. Both therefore were figured Red and were fired upon . Both were saved from this round of missiles by their ECH . P0004 then passed out of range . Unfortunately for P0008, it stayed in range long enough to allow the SAM time for a second shot . This time the ECH was inadequate and the A/C was destroyed.
@inproceedings{ijcai1979p55,
title = {KNOBS: An Experimental Knowledge Based Tactical Air Mission Planning System and a Rule Based Aircraft Identification Simulation Facility},
author = {Engelman, Carl and Berg, Charles H. and Bischoff, Miriam B.},
booktitle = {Proceedings of the Sixth International Joint Conference on
Artificial Intelligence, {IJCAI-79}},
publisher = {International Joint Conferences on Artificial Intelligence},
editor = {Bruce G. Buchanan},
year = {1979},
month = {8},
url = {https://ijcai.org/proceedings/1979/55},
}
TY - CPAPER TI - KNOBS: An Experimental Knowledge Based Tactical Air Mission Planning System and a Rule Based Aircraft Identification Simulation Facility AU - Engelman, Carl AU - Berg, Charles H. AU - Bischoff, Miriam B. T2 - Proceedings of the Sixth International Joint Conference on Artificial Intelligence, IJCAI-79 PB - International Joint Conferences on Artificial Intelligence PY - 1979 UR - https://ijcai.org/proceedings/1979/55 ER -Copied
Machine-readable versions of this paper, for tools and AI agents: