1 / 11

Deep Understanding of Courses of Action Paul Cohen University of Massachusetts

Deep Understanding of Courses of Action Paul Cohen University of Massachusetts. Premises Computers can understand courses of action well enough to serve as C2 assistants. Making C2 increasingly electronic will facilitate coalition operations Evidence

talib
Télécharger la présentation

Deep Understanding of Courses of Action Paul Cohen University of Massachusetts

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Deep Understanding of Courses of ActionPaul CohenUniversity of Massachusetts Premises • Computers can understand courses of action well enough to serve as C2 assistants. • Making C2 increasingly electronic will facilitate coalition operations Evidence • DARPA’s High Performance Knowledge Bases (HPKB) project, Control of Agent-based Systems (CoABS) and Capture the Flag

  2. bulldozer train tank can-be-used-as isa isa m1 m1 m1 What is shared understanding?Yes, it’s more than ontology y isa x In “conventional functional” systems, semantics are provided from without, by users (by “convention”), and machines have no independent understanding of anything. Besides ontologies, you need semantics (“axioms”) to help machines tell whether we would judge assertions to be true, partly true, false, or more or less probable. This is shared understanding between us and machines. Besides ontologies and axioms you need a representation of context.

  3. A common theory of understanding • Functionalism: We understand to the extent that we combine inputs with knowledge to give the right outputs. “The COA is not feasible because Blue 3 is assigned a task of which it is not capable. Blue 3 is an infantry brigade with a nominal speed of 4 mph, tasked to fix Red 1, a motorized rifle regiment with a nominal speed of 20 mph. A fixing unit should be faster than the unit it is fixing. Suggestions: Assign Blue 4 the fixing operation, or degrade the speed of Red 1 with deep fires.” Source: HPKB SAIC team

  4. Depth of Functional Understanding “The COA is not feasible because … a fixing unit should be faster than the unit it is fixing. Suggestions: Assign Blue 4 the fixing operation, or degrade the speed of Red 1 with deep fires.” Domain-specific “expert system” General If X can’t do a task but Y can Then assign the task to Y If to-be-fixed unit X moves faster than a fixing unit Then degrade X’s speed with deep fires If Y is supposed to maintain contact with X, and X moves faster than Y Then try to reduce Y’s speed

  5. General physical schemas develop early in life I want to maintain contact with Maamaa, and Maamaa moves faster than me So try to reduce Maamaa’s speed MAAAAMAAA!

  6. Clausewitz, an advocate forphysical schemas as a basis for understanding warfare "There is no higher and simpler law of strategy than that of keeping one's forces concentrated.” "The conduct of war resembles the workings of an intricate machine with tremendous friction, so that combinations which are easily planned on paper can be executed only with great effort. Consequently the commander's free will and intelligence find themselves hampered at every turn, and remarkable strength of mind and spirit are needed to overcome this resistance." Conjecture: If we understand warfare (and much else) in terms of simple physical schemas, then perhaps these schemas can serve as a core set of concepts in terms of which coalition partners can understand one another and computers can understand command and control.

  7. How physical schemas facilitateshared perspective Ontology + Context + Axioms = Reasoning If Y is supposed to maintain contact with X, and X moves faster than Y Then try to reduce Y’s speed “Though I am only a small scouting force I can harass Red and cause him to assume a battle formation, thus slowing him down…”

  8. How physical schemas facilitateshared perspective Ontology + Context + Axioms = Reasoning “Shared perspective is not just an ontology issue” -- Buster McCrabb If Y is supposed to maintain contact with X, and X moves faster than Y Then try to reduce Y’s speed “Though I am only a small scouting force I can harass Red and cause him to assume a battle formation, thus slowing him down…”

  9. Capture the FlagPlanning with Physical Schemas

  10. Shared perspective isn’t always symbolic representations distance Success Probability Narrow ledge between success and failure Distance from Flag Time

  11. Conclusion • Functional understanding can be achieved by very specific or very general knowledge – rules about deep fires or rules about slowing one’s opponent – general is better! • Physical schemas are general concepts learned in the cradle and widely applicable to course of action analysis, and more. • HPKB and Capture the Flag understand courses of action via general knowledge, esp. physical schemas • Computer-based C2 tools for coalition planning could provide a lingua franca if they communicate in terms like physical schemas

More Related