Distributed Collaborative Intelligence: A Tactical Offset Strategy

Distributed Collaborative Intelligence: A Tactical Offset Strategy

https://www.darpa.mil/attachments/OFFSET_Prop_Day_ARL-2.pdf

• 3 rd Offset : Undersea and extended range air • Tactical Offset : Unmanned systems to extend the reach and situational awareness for the Soldier across complex worlds Tactical Offset Extreme Terrain DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Tactical Intelligent Systems for 2025 and Beyond Underlying Assumptions • Large #’s of agents – 10s to swarms • Heterogeneous mix – including Soldiers • Highly collaborative systems • Highly distributed systems • Access to knowledge sources and increased perception and awareness • Increased cognitive behaviors and real-time optempo adaptable operations • Operation in complex and contested environments – peer capabilities Payoff • Extended reach, situational awareness, and operational effectiveness against dynamic threats in contested environments • Technical and operational superiority through intelligent, resilient and collaborative behaviors Vision • Highly distributed and collaborative heterogeneous teams of intelligent systems DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Axes of Complexity These factors limit the operational capability for a given autonomy technology suite. 1. Complexity of the environment(s) 2. Available infrastructure 3. Operational tempo 4. Number of agents 5. Degree of heterogeneity of the agents 6. Agent behavior complexity and adaptability 7. Degree of interaction and communication among the agents (both machine and human agents) The Army challenge: complex unknown environments, little or no infrastructure, and high operational tempo. DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. ARL Intelligent Systems Center Decision-Making beyond Human Op-Tempo Adaptable to New Environments Comprehensive World Model Common Sense Reasoning Dynamic Learning Self-organizing Robot Teams Distributed Computing & Autonomous Networking Robust Physical Agents Real-time Planning Socially Cognizant Behaviors & Adversarial Reasoning Desired Attributes of Army Intelligent Systems Mobility & Manipulation | Perception | Adaptive Control • Cognitive Architectures • Artificial Intelligence • Reasoning/Knowledge Engineering • Natural Language • Semantics • Big Data Analytics • Machine Learning • Distributed Computing (HPC) • Game Theory • Process & Task All DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Provide fundamental science underpinnings of autonomous systems for the Army From micro-systems to combat vehicles Soldiers/Unmanned System Teaming: • Combat multiplier • Team member • Heterogeneous groups • Following commander’s intent ARL Intelligent Systems Center Robotics CTA MAST CTA Internal ARL Research ARO DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Micro Autonomous Systems and Technology (MAST) Collaborative Technology Alliance (CTA) – 2008-2017 Enhance tactical situational awareness in urban and complex terrain by enabling the autonomous operation of a collaborative ensemble of multifunctional, mobile microsystems – Rapid and Mobile ISR for the Dismounted Soldier Caves, Strategic Bunkers, Subterranean Jungles and Under Canopies Megacities and Urban DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. MAST CTA Communications, Navigation & Coordination Mobility, Control, & Energetics Sensing, Perception, & Processing Three Cross-Cutting Research Thrusts and a Joint Experimentation Thrust • Indoor/Outdoor GPS-denied Navigation • Dealing with Uncertainty • Heterogeneous Collaborative Systems • Aggressive Maneuvers & Collaborative Behaviors • Robust Multi-Spectral Comms Solutions • Novel systems – flapping wing/cyclocopter • Perching & Grasping • Gust Mitigation • Novel Power Source • Understanding Ambulation over Complex Terrain • Hair-like Arrayed Sensors for Gust and Acceleration Sensing • Robust State Estimation • Vision and SWaPP Constrained StateEstimation • 5g 220 GHz Radar • Stereo/Optic Flow Sensor Fusion Joint Experimentation • Integrate and extend lower-level capabilities from core research tasks • Evaluate results on platforms performing elements of missionlike scenarios • Involve collaborations among multiple institutions, centers, and ARL DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. MAST CTA & the ARL Enterprise • Bio-Inspired Sensors and Controls Team •(SEDD-Materials) - Sensor Integration, StateEstimation, Human In-The-Loop Controls, PNT • Microsystems Mechanics Team • (VTD-Sci Mvr) - Aeromechanics, Soft Materials, Fluid Dynamics, Controls, Flight Dynamics • Manipulation and Mobility Team • (VTD-Sci Mvr) - Self-righting, grasping •Asset Control and Behavior Branch •(CISD-Info Sci) - collaborative behaviors, network aware communications, GPS-Denied nav • Tactical Network Assurance Branch •(CISD-Info Sci) - Networked behaviors, VHF and Optical Comms • Smart Swarm Munitions Team • (WMRD-Sci for Lethality & Prot) – Vehicles, propulsion, networking, perception •New infrastructure, labs, and test facilities DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. The RCTA Vision Making unmanned systems an integral part of the small unit team Systems that: • Understand the environment • Learn from experience • Adapt to dynamic situations • Possess a common world view • Communicate naturally • Conduct useful activity • Can act independently, but within well prescribed bounds Through research to enable and advance: • Abstract reasoning • Learning by example • Reinforcement learning • Semantic perception • Communication through language • Human behavior modeling • Agile 3-D mobility at operational tempo • Human-like manipulation From tool to teammate DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Merging Cognition & Control From tool to teammate The unique hybrid architecture combining cognitive upper and geometric lower layers – bridging the gap between artificial intelligence & control • Introduction of abstractions to facilitate humanrobot communication – behavior specification based on structured language • Algorithms to permit semantic labeling of objects, behaviors and their relationships • A focus upon learning, to include learning without reliance on large sets of training data • A broad description of the environment that goes beyond placing objects in fourdimensional space to include concepts such as object compliance and signature • Natural modes of communication – voice and gesture • Exploration of unconventional mobility modes • Human-scale manipulation Coupled with enhancements to facilitate teaming DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Thrust 1 – Op-tempo Maneuvers in Unstructured Environments Robustness Op-tempo Collaboration Complex Environments Thrust 2 - Human-Robot Execution of Complex Missions Situational Awareness in Unstructured Environments Distributed Mission Execution Trusted Execution of Verified Missions Thrust 3 – Mobile Manipulation Complex 3D Environments RCTA: Key Challenges DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Operational Scenario Intelligent System Components Autonomy / Swarms Network Experts Sensors Knowledge Bases Autonomy for Networking : Networking for Autonomy DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Autonomous Networking Autonomous Networking • Blend networking and autonomy to enhance both and provide resilient seamless services • Hybrid multi-radio multi-wavelength PHY • Autonomy as mobile infrastructure • Cognitive, self-healing radio • Seamless embedded PNT • Distributed Beamforming & EW DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Research Topics: Distributed Intelligence: Establish the theoretical foundations of multi-faceted distributed networked intelligent systems combining autonomous agents, sensors, tactical super-computing, knowledge bases in the tactical cloud, and human experts. Heterogeneous Group Control: Develop the theory and algorithms for control of large autonomous teams with varying levels of heterogeneity and modularity across sensing, computing, platforms, and degree of autonomy. Adaptive and Resilient Behaviors: Develop theory and experimental methods for heterogeneous multi-agent groups to carry out tasks in the physical world. Expected Payoff: • Extended reach, situational awareness, and operational effectiveness against dynamic threats in contested environments Distributed and Collaborative Intelligent Systems and Technology (DCIST) New Collaborative Research Alliance (CRA) Opportunity • Distributed and Collaborative Intelligent Systems and Technology (DCIST) • Program Announcement Feb 2017 - See official announcement for programmatic and submission details • Will be announced through FedBizOpps • Opportunities for collaboration with ARL researchers, leveraging ARL facilities, and an on-site presence Vision: Develop the underpinning science to extend the reach, situational awareness, and operational effectiveness of Intelligent System/Soldier teams against dynamic threats in complex and contested environments and provide technical and operational superiority through fast, intelligent, resilient and collaborative behaviors. DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. Tactical Offset Summary • Merge cognition & control for distributed intelligent systems • Fundamental new science of human – autonomy teaming • Leap ahead in rapid deployment & autonomy projection • Focus deep learning on Army intelligent system scenarios • Autonomy for Networking : Networking for Autonomy Drive new tactical offset strategy DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. End Dr. Brian M. Sadler Senior Research Scientist, Intelligent Systems Army Research Laboratory brian.m.sadler6.civ@mail.mil (o) 301-394-1239