MIT Lincoln Laboratory Support to Unmanned Aircraft Systems Integration into the US National Airspace

MIT Lincoln Laboratory 244 Wood Street, Lexington, MA 02420-9108 MIT Lincoln Laboratory Support to Unmanned Aircraft Systems Integration into the US National Airspace MIT Industrial Liaison Program Research and Development Conference This work is sponsored by the Department of the Army and the Department of Homeland Security under Air Force Contract #FA8721-05-C-0002. Opinions, interpretations, conclusions and recommendations are those of the author and are not necessarily endorsed by the United States Government. Rodney E. Cole 15-16 November, 2011 1

Outline Background Sense and Avoid overview Sense and avoid development Summary 2

Growing Demand for UAS Operations in the National Airspace Crew Training CONUS Basing for International Missions Oceanic and Littoral Surveillance Coastal Surveillance Border Surveillance Environmental Monitoring (Law Enforcement, Disaster Relief) Diverse Diverse operational operational requirements requirements for for UAS UAS in in the the national national airspace airspace 3

Growth in UAS Operations UAS Flight Hours 2009 is the first year the AF will train more groundbased UAV operators than 'fighter jockeys' and bomber pilots. - Gen. Stephen Lorenz, head of Air Education and Training Command Base Locations Navy USMC USAF Army Spec Ops Com DHS 2010:146 Units, 66 Locations 2015: 197 Units, 105 Locations Steady increase in operations will outgrow DoD airspace 4

UAS Airspace Operations Today Lengthy COA process Mean time to complete 150 days Requires advanced ATC coordination planning prior to operations See and Avoid surrogate outside Class A and restricted airspace Chase aircraft or ground observers Conduct operations above 18,000 ft if no chase aircraft Visual flight conditions File and Fly Desired Future Same process as manned aircraft No advanced coordination required Seamlessly integrate with traffic in all airspace classes See and Avoid performed by UAS Conduct operations at any altitude No constraints on visual conditions Need incremental progress toward open access to the national airspace 5

Barriers to UAS Airspace Integration* Barriers Current (Today) Near (< 18 mo) Time Mid (18 mo FYDP) Far (>FYDP) Pilot Qualifications Basic UAS Qualification Levels Airworthiness Airworthiness Standards Update UAS Specific Criteria Performance Standards Ops Stds & Procedures Night Ops Multiple UAS Lost Link Procedures Update Service Publications Equipage Link Security Bandwidth & Spectrum Certified Equipment Sense-and-Avoid (SAA) Performance Standards Ground-Based Airborne Lincoln focus Sense and avoid is the principal technical hurdle to airspace integration * JUAS COE Airspace Integration Capabilities Based Assessment, 2009. 6

Outline Background Sense and Avoid overview Sense and avoid development Summary 7

Sense and Avoid Overview SAA is the capability of an unmanned aircraft to remain well clear from and avoid collisions with other airborne traffic * Primary sense and avoid system functions: - Self-separation strategic maneuvering to maintain Well Clear - Collision avoidance tactical, last-minute maneuver to avoid a collision Principal sub-functions are: Airspace surveillance Threat detection Threat avoidance maneuver Self Separation Volume Collision Volume Avoidance Maneuver Threat Aircraft Surveillance Threat Detection Avoidance Detect Track Evaluate Prioritize Declare SAA Functional Execution Time-Line Determine Command Execute 8 *FAA Sponsored Sense and Avoid Workshop Final Report, 2009.

Sense and Avoid Concepts Ground-Based Airborne Unmanned Aircraft Operational Volume Threat Aircraft Surveillance Volume Unmanned Aircraft Threat Aircraft UAS Ground Station Operating Base Ground-Based Radar Array UAS Ground Station Air Traffic Services Rapid deployment enabled by leveraging existing ground-based surveillance and tools The same surveillance and support hardware supports a diverse range of platforms Operational volume limited by surveillance coverage Unconstrained operations enabled by surveillance volume fixed to platform Longer timelines associated with developing and certifying airborne components Smaller platforms may not have the size, weight, or power to support airborne hardware Path forward: GBSAA ABSAA Hybrid 9

Sense and Avoid Elements System Elements Data fusion and Tracking COTS ground based 3D radar, FAA Threat detection and Maneuver Airborne sensor development Algorithms Sensors Decision Support Aids Sense and Avoid System ) t f ( h t r o N P(NMAC relative position) 10000 0.05 8000 6000 0.1 4000 0.2 2000 0.5 0-5,000 Modeling & Simulation 1-2500 0 East (ft) 2500 Standards 5,000 Airspace Characterization Safety Case Enabling Activities 10 Open Architectures Testbeds

Outline Background Sense and Avoid overview Sense and avoid development Standards Threat detection and maneuver guidance Modeling and simulation Testbeds to demonstrate capabilities Summary 11

Sense and Avoid Standards Development FAA-sponsored Sense & Avoid Workshops Defined sense and avoid Defined key essential elements of any SAA system OSD Target Level of Safety (TLS) Workshop Define DoD s understanding of target level of safety RTCA Special Committee 203 Industry, FAA, and DoD committee to define minimum SAA standards MIT/LL is Strongly Engaged in UAS Community and Regulatory Standards Development 12

Standards Development: Well Clear Well Clear Definition FAA-sponsored workshop requirement for SAA system to perform self separation Ability to stay well clear of air traffic Well clear concept from regulatory language, but is not defined Analytical Well Clear Boundary Well clear threshold Derived based on unmitigated probability of NMAC from random encounters ) t f ( h t r o N 8000 6000 4000 2000 0-2000 P(NMAC Relative Position) 0.2 0.5 1 0.1 0.05 MIT LL is establishing a risk-based separation standard based on our safety assessment framework Provides quantitative, analytical definition as the basis for system evaluation -4000-2000 0 2000 East (ft) Definition based on risk of collision after crossing the separation boundary Near Midair Collision (NMAC) Volume Laboratory is playing key role in developing and defining standards to support UAS airspace integration 13

Standards Development: Target Level of Safety Increasing risk Target Level of Safety Increasing safety Unmitigated level of collision risk Mitigations to reduce hazard likelihood Resultant risk level Candidate Noncooperative* Target Levels of Safety (collisions/hr) 10-6 10-7 10-8 10-9 Observed general aviation collision rate RVSM/ADS-B TLS (lowest previously used) FAA acceptable rate of hull loss from equipment hazard Required target level(s) of safety is currently being determined by the FAA/DoD/NATO * Aircraft not under ATC control 14

Lincoln ACAS-X based Logic Development Efforts FAA Army/Air Force GBSAA DHS ABSAA 3-D Wide Area Candidate Sensors Weather Other Sensors Sensor and Decision Support Services Sensor Fusion Conflict Detection Ground Based Radars FAA/DoD Radars And Data Bases CIWS EO/IR Airborne Sensor Cuing Resolution Advisory Logic Sensor Network Adapters UAS Sense and Avoid DDS Bus UAS Network Adapters UAS Telemetry UAS Operator Displays Common Network Services Messaging Alert Notification Data Recording Simulation And Build -in-test Directory Services Database s Airborne Collision Avoidance System (ACAS X) 4 year FAA effort to replace TCAS Designed to be highly adaptable to changes in sensors, airspace and aircraft performance FAA prototype effort flight test in FY13 Existing TCAS surveillance, new threat logic developed by MIT LL Maneuver algorithm adapted from FAA ACASA-X development Self separation Collision avoidance Conflict avoidance algorithms being evaluated in CASSATT Integration in Army testbed at Dugway Proving Grounds Prototype low cost, size, weight and power SAA sensor Electronically scanned phased array Applicable to Pred-B, Fire Scout, and other mid-sized UAS Flight tests of sensor in FY12 on surrogate manned aircraft Conflict avoidance studied off-line 15 MS-54062C

First Use of Lincoln Maneuver Algorithm Approach Internally funded in FY08 Airborne Collision Avoidance for Unmanned Aircraft Initial application was to unmanned aircraft (in collaboration with MIT CSAIL) Vertical collision avoidance Autonomous (no pilot delay) Different sensor systems, dynamics, and performance constraints Consistent approach applied (build model, choose cost metric) Led to the FAA funded effort to build next generation collision avoidance algorithms Modality Range Azimuth Elev. FOV Range Traffic TCAS Radar EO/IR Measurement Accuracy Coverage S. Temizer, M. J. Kochenderfer, L. P. Kaelbling, T. Lozano-Pérez, and J. K. Kuchar, Collision Avoidance for Unmanned Aircraft using Markov Decision Processes, in AIAA Guidance, Navigation, and Control Conference, Toronto, Canada, 2010. ADS-B 16 MS-54062C

ACAS X Problem Statement NextGen procedures and surveillance necessitate changes to TCAS TCAS logic is heuristic and difficult to revise Challenges to collision avoidance logic development Uncertainties due to sensor noise and aircraft behavior Aircraft performance and operational constraints Need to maximize safety while minimizing unnecessary alerts Solution is a decision-theoretic approach to collision avoidance logic development Uses explicit models of sensor and dynamic uncertainty Optimizes logic according to objective performance measure Leverages advances in computation and algorithms 17 MS-54062C

Logic Development Comparison Legacy TCAS Development Cycle Model-Based Optimization Approach Logic (pseudocode) Encounter Model Simulation Performance Metrics Evaluatio n Models: Pilot response Encounter Aircraft response Performance Metrics Optimization Logic (table) manual pseudocode revision Human effort focused on pseudocode Time-consuming process Many parameters require tuning Unlikely to be optimal Human effort focuses on models Computers generate lookup table Optimal logic - Low false-alert rate - High degree of system safety Model-based approach is much more easily adapted to NextGen needs 18 MS-54062C

ACAS Logic Accounts For Uncertainty Previous methods use extrapolation of intruder path Leads to single point of closest approach Depending on situation uncertainty in point of closest approach can be high May lead to induced encounters if probabilities of other intruder maneuvers are not accounted for Solution is to determine actions taking into account uncertainties in intruder path Other probabilistic aspects of collision avoidance: UAS pilot response Aircraft performance State uncertainty (measurement error) 19 MS-54062C

Dynamic Programming Logic Development and Implementation Overview Offline Logic Optimization Performed on high-performance computers Real-time Logic Usage Executed onboard aircraft (once per second) Pilot, Intruder, Aircraft Dynamics Models Performance Metric Expected Cost Table Sensor Measurements Tracker Model Discretization State, inc Uncertainty Dynamic Programming Action Selection Expected Cost Table Action (no alert, climb, descend) Model-based, accounts for uncertainties 20 MS-54062C

Adapting ACAS-X Algorithm for UA SAA Ground-Based Sense and Avoid UAS Ground Station Unmanned Aircraft Operational Volume Operating Base Ground-Based Radar Array Threat Aircraft Key Differences with ACAS Different track error characteristics UA have different performance limits than manned aircraft UA sense and avoid must perform selfseparation and collision avoidance Two different levels of alert urgency Adaptation Steps Adapt aircraft dynamic model to UA Change action space from vertical maneuvers to horizontal for self-separation Same action space for collision avoidance Next Steps Performance modeling Develop coordination with manned collision avoidance (with NIEC) ACAS-X model-based algorithm modified to support UA SAA 21 MS-54062C

Safety Modeling and Simulation Framework Realistic Fast-Time Simulation: millions of encounters Raw radar data Tracking and fusion Feature extraction Encounter models Aircraft flight profiles and dynamics Fast-time simulation Surveillance model Collision avoidance system models (algorithms) Collisions per encounter Collisions per flight-hour Relative risk analysis Track database Density processing Density models Encounters per flight-hour Encounter rate estimation Target Level of Safety risk analysis Cooperative: Noncoop: Observed encounters per flight-hour Proportional to traffic density and airspeeds Encounter Rate Model 22

M&S Applied to Notional GBSAA 10 million encounters Simulation of close encounters 10 million outcomes Three radar system Horizontal Vertical Couple simulation results with airspace density Unmitigated collision rate (per flight hr) With mitigation collision rate (per flight hr) 25%-tile: 5x10-8 1x10-9 50%-tile: 3x10-7 7x10-9 75%-tile: 2x10-6 5x10-8 Estimated unmitigated collision rates CASSATT framework is being used to refine and analyze threat detection and maneuver logic 23 MS-54062C

Testbed Efforts Army (GBSAA) Elimination of ground observers Reference architecture to support RFP First deployment to Dugway Proving Grounds summer 2011 Restricted airspace allows for unfettered testing Open architecture Radars Effort includes data fusion/tracking and maneuver algorithm development 3-D Ground Based UAS Telemetry FAA/DoD Radars And Data Bases UAS Netwo rk Ad ap ters Simulation And Build -in-test Wide Area Candidate Sensors Weather CIWS Sensor Network Adapters Other Sensors EO/IR UAS Sense and Avoid DDS Bus UAS Operator Displays Airborne Sensor and Decision Support Services Sensor Fusion Sensor Cuing Conflict Detection Resolution Advisory Logic Common Network Services Messaging Alert Notification Directory Services Data Recording Database s Air Force (GBSAA) Focus on consolidating ground observer task Cannon and Grey Butte Lincoln roles: Airspace characterization Algorithm development Modeling and simulation to support safety case Sentinel or ASR-11 STARS Lite DHS (ABSAA) Demonstrate flexible, scalable airborne radar based system customizable to a broad set of UAS Prototype a flexible low SWAP phased array radar for SAA surveillance Flight tests on surrogate manned aircraft Test with both JOCA and MITCAS maneuver logic MIT LL Lightweight Notch Array FAA Army testbed efforts briefed to FAA Research and Technology Development (RTD) office FAA RTD has expressed interest in the concept and potential application to support UAS research initiatives and NexGen technology implementation activities MIT and FAA RTD office to hold further discussions Discussions underway with TCAS program office to explore unmanned manned aircraft collision avoidance coordination 24

Army Ground-Based Sense and Avoid Program 3-D Wide Area Candidate Sensors Weather Other Sensors * Sensor and Decision Support Services Sensor Fusion Conflict Detection Ground Based Radars FAA/ DoD Radars And Data Bases CIWS EO/IR Airborne * Possible future extension Sensor Cuing Resolution Advisory Logic Sensor Network Adapters UAS Sense and Avoid DDS Bus UAS Netwo rk Ad ap ters UAS Telemetry UAS Operator Displays Common Network Services Messaging Alert Notification Data Recording Simulation And Build -in-test Directory Services Database s Lincoln Roles Define Open Reference Architecture for sense and avoid Self separation and collision avoidance requirements and algorithms Sensor fusion and tracking requirements and algorithms Assessments of system components: sensors, avoidance logic, display concepts Deployed to Dugway Proving Grounds summer 2011 25

Customs and Border Protection Southern Border Patrol Operations Requires Certificate of Authorization Advanced coordination with ATC Observers/chase aircraft < 18K ft Generally daylight, visual conditions Mission ops at 19,000 ft only 18,000 ft. UA Under ATC Control Chase Aircraft Desire ability to drop to 5,000 ft AGL for close-up observations Requires airborne sense and avoid 3000 ft. Ground Observer Ground Observer 26

DHS S&T Technology Demonstration Prototype Arrays 15 x 15 x 2.5 Surveillance System Key challenge is low cost, low power, lightweight, low profile active electronically scanned arrays (AESAs) (20kg, 200 W, $200K) MIT to build and demonstrate late FY12 Surrogate Aircraft Integrated Penalty Score Alerting Systems AFRL Automatic Jointly Optimal Collision Avoidance Modified by NG/Bihrle for Predator-B! 820 -ft Minimum Separation Follow TCAS/RA Right of Way Passive Ranging Maneuver EO-only Intruders w/ Range or Velocity Uncertainty Staying Well Clear (2460 -ft) Keep within EO / Radar Sensor Field of View Limits Staying within ATC Corridor Enabling Low-Cost Tile Based Arrays MIT LL ACAS-X based logic* Current results for collision avoidance only Alerting Options (Bank) 0-10 +10-20 +20 * Adapted from ACAS-X 27 MS-62602

Outline Background Sense and Avoid overview Sense and avoid development Summary 28

Summary Substantial progress in enabling UAS airspace integration is being made Laboratory roles: - Developing standards for UAS integration - Modeling and simulation to demonstrate safety - Sensor and algorithm development - Real-time architectures - Engaged with multiple stakeholders: DoD, DHS, FAA - Programs spanning airborne and ground-based sense & avoid - FAA engagement through the UAS Program Office and TCAS Program Office 29