21
Boeing Defense, Space & Security PhantomWorks Autonomous Systems: The Future in Aerospace Kevin A. Wise, Ph.D. Senior Technical Fellow, The Boeing Company NAE-AAES Convocation 24 April 2017

Autonomous Systems: The Future in Aerospace · Boeing Defense, Space & Security PhantomWorks Autonomous Systems: The Future in Aerospace Kevin A. Wise, Ph.D. Senior Technical Fellow,

Embed Size (px)

Citation preview

Boeing Defense, Space & SecurityPhantomWorks

Autonomous Systems: The Future in Aerospace

Kevin A. Wise, Ph.D.Senior Technical Fellow, The Boeing Company

NAE-AAES Convocation24 April 2017

BDS | PhantomWorks

A Disruptive Surge in Autonomy

• Automotive industry moving quickly• Predictions for 90% of Autos to be

autonomous by 2030• Reliable low cost sensor

development• Collaborative Autonomy

• New companies building autonomous aircraft at record pace

• Strong competition to be first to bring internet to new markets

• Ground/air package delivery

Autonomous Systems| 2

BDS | PhantomWorks

From Piloted Stable Aircraft to Autonomous Unstable Platforms: An Evolution

NonlinearAdaptiveControl

Mechanical Systems

Operated By Pilot

Autopilot Reduces

Pilot Workload

Fly-By-WireControls

Control Augmentation

Systems

1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

Autonomous Systems| 3

Mechanical To Digital

- Control Theory -The Enabling Technology

BDS | PhantomWorks

What was the first UAV?First engine-powered, heavier-than-

air aircraft capable of sustained flight Developed by Samuel P. Langley,

Secretary of the SmithsonianLangley Aerodrome No. 5 First flights - 6 May 1896 13 ft wingspan/tandem wing 3300 ft and 2300 ft (1000 m and 700 m)Langley Aerodrome No. 6 First flight - November 1896 4790 ft (1460 m) Launched from a riverboat catapultDemonstrators for manned attempt

Aerodrome No. 5

Aerodrome No. 6

Autonomous Systems| 4Ref: M. Hirschberg “American Attack Unmanned Aerial Vehicles,” AIAA 2003-01-3064

BDS | PhantomWorks

Some Examples – Early Boeing Programs

YQM-94A Compass Cope B1973-1974

Condor1986 -1988

90 ft. wingspan One of the largest remotely piloted

jet-powered UAVs for its era Endurance of 17 hr 24 min at

55,000 ft. altitude

All composite aircraft 200 ft. wingspan (B-52 185 ft.) Modular construction for transport to

remote sites 141 hr. flight test program Altitude record of 66,980 ft. for

piston-powered UAV Max duration of 2.5 days

USAF photo

Autonomous Systems| 5

BDS | PhantomWorks

Some ExamplesYPQM-149A or YPQM-150A

UAV- Short Range Sky Owl1989-1992

Length 4.12 m (13 ft 6.2 in) Wingspan 7.32 m (24 ft) Weight 566 kg (1250 lb) Speed 204 km/h (110 knots)

Parafoil Autoland

All numbers are approximate.

Autonomous Systems| 6

BDS | PhantomWorks

Some Examples

RQ-3A DarkStar Tier III Minus 1996 - 1999

Fully automated flight using GPS 69 ft. wingspan 500 nm range with 8 hr loiter time 45,000 ft. altitude Boeing/Lockheed Martin

31 flights with NASA Remotely piloted 28% scale of manned fighter Controlled with forward canard, split

ailerons and thrust vectoring Additional flights under USAF

RESTORE program. First demonstration of UAV adaptive flight control.X-36 Tailless Fighter Agility - Research Aircraft

1993-1998NASA photo

Autonomous Systems| 7

BDS | PhantomWorks

X-45 Joint Unmanned Combat Air System

Autonomous Systems| 8

Transformational ProgramEngineered To Support Full

Autonomous Operation

X-45A

49 ft.

F-16

Phantom Ray

F-117

36 ft.

Phantom RaySingle Pilot/

Operator Managing 4 Aircraft

BDS | PhantomWorks

Aloft for 10+ days 65,000 ft 250 ft wingspan Evolved from Condor ISR Telecommunications

HALE – High Altitude Long Endurance

Autonomous Systems| 9

Phantom Eye Demonstrator

BDS | PhantomWorks

Intelligent and Autonomous Systems

Sensing The Environment

Fusion, Perception, and Decision Making

Path ManagementNavigation,

Guidance, Control

• Interaction with humans• Dealing with contingencies• Trust

Autonomous Systems|10

Machine LearningArtificial Intelligence

BDS | PhantomWorks

Intelligent Autonomy – Evolutionary Change Producing Revolutionary Capability

Robust system operation with/without pilot/operator interaction

Mission goals achieved in the presence of system faults or contingencies

Force Level Planning/CAOC

Flight Package

Weapon/Sensor

Platform

Execution Chain

Real-time, Dynamic,Optimal

AssessEngageTargetTrackFixFindPlan

Communicate

“The whole is greater than the sum of its parts”

Autonomous Systems| 11

BDS | PhantomWorks

Integrated Battle Planning – “The whole is greater than the sum of its parts”Electronic Attack ISR

Strike

Integrated Battle Plan

OCA/DCA

Force Level Planning/CAOC

Flight Package

Weapon/Sensor

Platform

Autonomous Systems| 12

BDS | PhantomWorks

5K5K

5K5K

Hierarchical Control in Intelligent/ Autonomous Systems

P

1K

Uncertainties

Flight Control

Guidance/ Steering/

Navigation

MissionMgmt

Multi-Ship Mission Mgmt

Battle Mgmt

2K3K4K5K

HumanOperator/Pilot Interaction

Autonomous Flight

Contingencies and Uncertainties in All LoopsAutonomous Systems| 13

Machine LearningArtificial Intelligence

In Outer Loops

BDS | PhantomWorks

Contingency Management - Critical for Safe, Deterministic, Trustable Autonomy

Detection ReportAir Vehicle

GCS Operator

MCC

GCS

TOSC

MMS/MCS/FOCC

MissionPhase

&FlightModeLogic

Guidance FlightControl

OSMOutputSignal Mgmt

ISMInput

Signal Mgmt

RMRedundMgmt

Air VehicleMonitor

SensorData

ContModeLogic

SubsystemControl

• Air Data• ECS

• Propulsion• PDU

• FuelChannel A

Channel B CrossChannelData Link

SubsystemCmds

GPSNavigation

CategorizeCM Database

• Independent Action, Auto Response

• Operator Intervention and Control Plan

RespondSystem Response Plans (SRP)

Engine Out ••••Actuator Commands selected to be same value in bothchannels•Actuator failures will be detected to prevent force fights andactuator damage using a combination of monitors

• Actuator controller reports failures of position, RAM,ROM, EEPROM, power and communications errors

• Software model will be used to compare model positionto real actuator position

• Max rate of change for actuator position used to isolatefailures of position data

•Actuators in NAV processor will be shut off if Flight Controlprocessor in same VMS is failed•Actuators in both NAV and Flight Control processor will beshut off if a loss of CCDL is experienced but no loss inpartner channel is detected (determined by loss ofcommunications while still receiving synch interrupt discrete)as Navigation solution may be divergent resulting in actuatorforce fights

Bay Door Act •Actuator Commands selected to be same value in bothchannels•Actuator failures will be detected to prevent force fights andactuator damage using a combination of monitors

• Actuator controller reports failures of position, RAM,ROM, EEPROM, power and communications errors

• Software model will be used to compare model positionto real actuator position

• Max rate of change for actuator position used to isolatefailures of position data

•Actuators in NAV processor will be shut off if Flight Controlprocessor in same VMS is failed•Actuators in both NAV and Flight Control processor will beshut off if a loss of CCDL is experienced but no loss inpartner channel is detected (determined by loss ofcommunications while still receiving synch interrupt discrete)as Navigation solution may be divergent resulting in actuatorforce fights

• Deterministic & Predictable• Validated, Tested, Rehearsed

Autonomous Systems| 14

BDS | PhantomWorks

Why Is Trustable Autonomy Hard?

Battle Mgmt and C2• Integrated Control of Distributed Assets• Dynamic Resource Allocation• Task Assignment, Scheduling, Route Planning, SAA• Mission Execution and Monitoring• Vehicle and Sensor Management (Outer Loop, Inner Loop)• Subsystem Management

Top

LowestLevel

Hierarchical, distributed, optimal control of manned, semi-autonomous, and autonomous systems over wireless networks in adversarial environments

CommMgmt

Situational Awareness

Multiship Collaboration

Adaptive Networks

BMC2

InfoMgmt

ChallengesSW Hierarchy, Complexity, Curse of DimensionalityDynamic Environment, Not PredictableContingency ManagementAppropriate Models and SimulationsNon-unique BehaviorsCyber Security

Autonomous Systems| 15

BDS | PhantomWorks

Adaptive Control Motivation: Reduce Pilot Workload In Controlling Damaged Aircraft

F-15 Mid-Air CollisionA-300 Missile Attack B-747 Engine Failure

• F-15 Accident Revives Interest In Adaptive Control• NASA IFCS (Indirect Adaptive)

• Adaptive Control Successfully Applied To Open Loop Unstable Systems• USAF RESTORE (Direct Adaptive)

Autonomous Systems| 16

Closed-LoopReference

Model

BDS | PhantomWorks

USAF/NASA/Boeing Strong Partners In The Development of Adaptive Control

–Improved Performance, Mission Reliability–Reconfigurable, Damage Adaptive–Retrofits Onto Existing Flight Control Laws–Flight Proven

96 97 98 99 00 01 02 03 04

MK-82 L-JDAM

Reconfigurable Control For Tailless

Fighters (AFRL-VA/Boeing)

X-36

MK-84 JDAM

Adaptive Control For Munitions

(AFRL-MN/GST//Boeing)MK-84

05

Robust Adaptive Control

Technology Transition Timeline

X-45CX-45A

J-UCAS

06 07 13

Boeing IRAD

MK-84L-JDAM

MK-82 JDAMMK-84 JDAMIDP 2000

08 09 10 11 12

Intelligent Flight Control System (NASA/Boeing)

F-15 ACTIVE

• Gen I, flown 1999, 2003• Gen II, 2002 – 2006

•flight test 4th Q 2005• Gen III, 2006-Present

Boeing/MIT/UIUC/CalTechV&V methods for Adaptive Systems

Phantom Ray

14

Dominator UAS

MRAC

OBLTR

Indirect Adaptive Control Adaptive Dynamic Inversion (ADI) Autonomous Systems| 17

BDS | PhantomWorks

An Adaptive Control Challenge: Very Flexible Aircraft

Output Feedback Control Architectures Needed For New VFA Aircraft DesignsDesire Model Based Flight Control Design

Large Uncertainties Require Robust Adaptive Control, x Ax Bu y Cx= + =

Big HALE

Vulture400 ft span

250 ft span

800 State ModelsFlex Modes < 1 Hz

Large DeflectionsMust Control Shape

Helios

Autonomous Systems| 18

BDS | PhantomWorks

Copyright © 2014 Boeing. All rights reserved.

Boeing Observer-Based Control with Loop Transfer Recovery (OBLTR)Robust Adaptive Flight Control

Technology Transition Timeline

SDB

X-45

1312 14

GBU-57 MOP• Improved Performance, Mission Reliability• Reconfigurable, Adaptive• Retrofits Onto Existing Flight Control Laws• Flight Proven

15 16 17

Pivot UASFF April 2014

Dominator FF Nov 2013 UFP CRAD2015-2016

IPODS CRAD2015-2016

Adaptive Autopilot Technology for Aerial Platforms

AFRL CRAD

Proprietary CRAD FF 2017

JDAM 2kER2015

HAAWC 6DOF2015-2016

T3 6DOF IRAD

Air Launch CRAD 2016

DSTO CRADFF Feb 2015

MECC CRAD2015-2016

Scan Eagle 2016-2017

uGCU

Phantom ICE

Dominator

TIGER CRAD2015-2017

Autonomous Systems|19

BDS | PhantomWorks

OBLTR Adaptive Control: The Algorithm

( )ˆ ˆ,Tad blx uu = −Θ Φ

OBLTR Adaptive Increment

Learning Rate Adjusts How Fast The Adaptive Control Responds

Control

This method is a form of Nonlinear Integral control

( )ˆ 0 0Θ =

Form Error Signal Between Desired Behavior and Aircraft

Closed LoopRef Model

Pilot Column/Wheel InputOr Guid Cmds

Aircraft

ey+-

DeadzoneIf Aircraft Response Within Deadzone of Desired Behavior Baseline Control Is Used. If Error Increases Outside Deadzone Nonlinear Integrator Kicks In To Adapt.

measyP

blu

adu

Cmds Response

Uncertainties

Baseline

Adaptive

Lyapunov - The Greatest Russian Mathematician

( )12

0ˆ ˆ, T

bT

l yx u Re W S−

ΘΘ = −Γ Φ

y

measy

Autonomous Systems| 20

AdaptiveGain Neural

Network

BDS | PhantomWorks

Summary

Autonomous Systems| 21

100+ Years Of Unmanned Aircraft. There Here To Stay– New roles/missions for commercial operation– GAFA all developing capabilities

Intelligent Autonomy– Automotive industry developing autonomous capabilities at a fast pace– Aerospace applications emerging

Adaptive Control– Improved safety if something goes wrong/fails– Automotive and Aerospace applications emerging