APPROVED FOR PUBLIC RELEASE UNCLASSIFIED AI-enabled Real-time Situational Understanding at the Tactical Edge ARO Adversarial ML Workshop 14 September 2017 Tien Pham, PhD Senior Campaign Scientist Information Sciences Campaign The Nation’s Premier Laboratory for Land Forces UNCLASSIFIED 1
UNCLASSIFIED Topics Background: AI & ML Essential Research Area (ERA) AI-enabled Situational Understanding Collaborative Research Programs & Facilities with AI & ML The Nation’s Premier Laboratory for Land Forces 2 UNCLASSIFIED
UNCLASSIFIED Essential Research Areas (ERAs) Distributed / Cooperative Cyber & EM Engagement in Contested Human- Technologies Environments Agent for Complex Teaming Environments Artificial Intelligence/ Machine Manipulate Learning Failure Physics for Robust Materials Tactical Unit Energy Independence Manufacturing at the Point of Need Accelerated Learning for a Ready Force Discovery The Nation’s Premier Laboratory for Land Forces 3 UNCLASSIFIED
UNCLASSIFIED Research Context Unified Unified Lan Land Ope d Operation tions s Pr Prevailin vailing in a Co g in a Comple mplex W x Wor orld ld Lar Large ge-scale, c scale, clutter luttered, ed, contest contested ed urban en urban envir vironment onment Highly-dispersed team of human & robot agents accessing highly heterogeneous information sources Learning in new Dynamic in-flight environments with learning & re-planning at the deception from Speed of the Fight persistent threats Decide Faster High Operational Tempo Asymmetric Manned-Unmanned Vision Teaming Improved Enhanced Mobility Situational Understanding The Nation’s Premier Laboratory for Land Forces 4 UNCLASSIFIED
UNCLASSIFIED AI & ML Research Challenges AI & ML Research Gaps AI & ML with small samples, dirty data, high clutter Lear Le arning ning in in AI & ML with highly heterogeneous data Comp Comple lex x Da Data ta En Envir viron onmen ents ts Adversarial AI & ML in contested, deceptive environment Distributed AI & ML with limited Res esou ource ce-co const nstrai aine ned d communications AI AI Pr Proc oces essing sing AI & ML computing with extremely low size, at t the the Point oint-of of-Nee Need weight, and power, time available (SWaPT) Explainability & programmability for AI & ML Gen Gener eraliz alizable ble & Pred & Pr edic icta table ble AI AI AI & ML with integrated quantitative models Goal: To research and develop artificially intelligent agents (heterogeneous & distributed) that rapidly learn, adapt, reason & act in contested, austere & congested environments The Nation’s Premier Laboratory for Land Forces 5 UNCLASSIFIED
UNCLASSIFIED Potential Future Capabilities Envisioned Comba Combat t Ca Capa pabilities bilities Auto utono nomou mous s Aug ugmen mented ted Rea eali lity ty for or UAV V Sw Swar arms ms Multi Multi-fac acete eted d Pictu Picture e Precision Engagement ISR, force protection, operational environment, friend-foe Non-kinetic Engagement BDA, network healing locations, activities, threats Squad Sensors Cogniti Cogn itive e EW EW & SIG & S IGIN INT Squad Autonomy Per erson sonal al Prote Pr otection ction SIGINT/EW SIGIN T/EW Pr Prec ecisi ision on Inter Interne net t of of counter cyber Enga En gage gemen ment sense Battl Ba ttlefield efield or electronic adversaries, Thing hings s attack, signature evade, jam management Wea earable ble Elec Electr tron onics ics biosensors, threat locators, sensors Human-Mac Human Machine hine Coll Collabo boration tion Cogn Cogniti itive e Netw Networ orks ks for or En Enha hanc nced ed Net Net-en enabled bled Human Human-Rob obot ot Decision Mak Decision Making ing Network that perceives Se Semi mi-au auton tonomo omous us Comba Combat t Tea eaming ming conditions, maintains memory, Wea eapo pons ns & adapts (e.g., Man Un-Manned Teaming) The Nation’s Premier Laboratory for Land Forces 6 UNCLASSIFIED
UNCLASSIFIED Topics Background: AI & ML Essential Research Area (ERA) AI-enabled Situational Understanding Collaborative Research Programs & Facilities with AI & ML The Nation’s Premier Laboratory for Land Forces 7 UNCLASSIFIED
UNCLASSIFIED Motivational Scenarios Unified Unified Lan Land Ope d Operation tions s Pr Prevailin vailing in a Co g in a Comple mplex W x Wor orld ld Lar Large ge-scale, c scale, clutter luttered, ed, contest contested ed urban en urban envir vironment onment The Nation’s Premier Laboratory for Land Forces 8 UNCLASSIFIED
UNCLASSIFIED AI-enabled Real-Time Situational Understanding Unified Lan Unified Land Ope d Operation tions s Pr Prevailin vailing in a Co g in a Comple mplex W x Wor orld ld Large Lar ge-scale, c scale, clutter luttered, ed, contest contested ed urban en urban envir vironment onment Focus on a commander of a small team operating in highly clutter denied environment and making decision with locally available information AI-enabled Real-time Situational Understanding for Decision Making The Nation’s Premier Laboratory for Land Forces 9 UNCLASSIFIED
UNCLASSIFIED AI-enabled Real-Time Situational Understanding Capabilities Focus: Enemy Estimates (SA) for Mission Command at the Edge in a distributed, complex, denied environment, with local resources AI & ML ERA Focus: Learning and Reasoning in Complex Data Environments & Resource-constrained AI Processing at the Point-of-Need AI & ML ERA AI that integrates adaptive learning inputs to generate tactically sensible Adaptive enemy estimates Learning & Reasoning at the Edge Contribute to real-time decision-making in adversarial, cluttered, distributed environments AI & ML Focused Efforts 1. Adversarial distributed ML 2. Robust inference & ML with conflicting sources 3. Adaptive online learning in real time 4. Adversarial reasoning integrating learned information 5. Resource-constrained adaptive computing for AI & ML The Nation’s Premier Laboratory for Land Forces 10 UNCLASSIFIED
UNCLASSIFIED Conceptual Roadmap Cumulative-Connected-Converged (1) A (1) Adversarial dversarial Distr Distributed ibuted ML ML ML that ML that is is Robu Robust st and and Resistant Resistant to D to Deceptive eceptive and Conf and Conflicting licting Input Inputs (2) Robust (2) R obust Infer Inference ence & & ML ML Reasoning about Reasoning about Enemy Enemy that that Incor Incorpor porat ates s Distr Distributed ibuted Learning Learning (4) A (4) Adversarial dversarial reasonin reasoning g integr integrating ating learned learned informa information tion AI for Gener AI or Generating ting Tactica acticall lly-Se Sensible ible Estimates f Estima tes for or Decision Decision Making Making at t t the he Edge Edge Di Distributed stributed, , Adv Adver ersariall sarially-robu obust, st, Reso esour urce ce-ad adaptiv ptive e (3) Adaptive (3) A daptive Online Online Learning Learning in in real real time time Adap Adaptive tive Real Real-time time Learning Learning with Constr with Constraine ained d Resource Resour ces (5) Resour (5) Resource ce-constr strain ained d adaptive adaptive computing computing The Nation’s Premier Laboratory for Land Forces 11 UNCLASSIFIED
UNCLASSIFIED AI-enable Capability Scenario “Fight’s Eyes” A Soldier supported by team of AI-enabled reasoning to provide agents in complex environment possible course of actions AI-enabled real-time Tactically sensible decision making estimates of enemy based on locally available information The Nation’s Premier Laboratory for Land Forces UNCLASSIFIED
Recommend
More recommend
