Part II
Professor Anthony Finn
Director, Defence & Systems Institute
University of South Australia
Reduce human resources
• Replace dull, dirty, dangerous work
• Act as a force multiplier (reach & access)
• Reduce costs, risks and liabilities
Improve performance
• Perform tasks not previously possible
• Perform existing tasks better
• Greater precision, higher accuracy
• Rapid response to developing situations
Increase survivability
• Reduce susceptibility to communication failures
• Reduces susceptibility of information interception
• Reduce human and other machine interaction
REDUCING HUMAN INTERACTION IS KEY, BUT NEED TO
MAINTAIN REQUISITE LEVEL OF PERFORMANCE
Military Benefits of UAVs
IFR, 2009
Challenges & Issues Energy & Power
Navigation & Mapping
Sensing & Perception
Learning & Behaviour
Planning & Cognition
Human-Robot Interaction
Cooperation/Collaboration
Technology/Software re-use
Trustworthiness & -ability
Evaluation, Metrics, T&E
Operational Concepts/Uses
Legal, Policy, Ethical …
Can they
do the job?
How will we
use them?
What technology
trade-offs are
involved?
What systems
& infrastructure
required?
What impact will technology have on work function?
The Big Ones (1): Value of UAVs
Time
TRL 1
TRL 2
TRL 3
TRL 4
TRL 5
TRL 6
TRL 7
TRL 8
TRL 9
ABSI R&D
CTDs
Industry uptake
Future Capability Acquisition
(SEA4000, AIR6000, LAND400, …)
Legal
Framework
Capability
Exploitation
Technology
Development
Conceive
FMOC
Experiment
Analyse
Implement
Strategic Context * White Paper
* ADF Joint Vision
* Future Warfighting Concept
* Allied Concepts
* Technology Futures
* Political Futures
* Social/Demographic Futures
Experimentation * War games
* Tech Demonstrators
* Operations Research
* Operations Analysis
* Sea Trials
* Exercises
Post-Experimentation * 2nd Order Analysis
* Operations Research
* Headmark Planning Conference
* FMOC Annual Report
* Headmark Annual Report
* Capability Impact Statement
Implementation * Capability Implications
* Plan Green
* Plan Blue
* DCPG
* Maritime Cap Dev Plan
* Doctrine
Technical Feasibility & Cost-Benefit for
Unknown Capabilities (desirable vs. possible)
The Big Ones (2): Not About the UAV
When you have worked out what
you want the UAV to do
(a) Do you just buy a UAV and use
it ‘as is’ or
(b) Do you change your system to
accommodate the UAV or
(c) Do you change both
Organisational fit/business model
The Big Ones (3): Certification
Airworthiness: the design of the aircraft must be
approved; the aircraft must be manufactured in
accordance with this design; and, the aircraft must
be maintained in accordance with appropriate
maintenance and configuration control procedures.
Flight rules: the responsibilities and authority of the
‘pilot’ must be defined, as must operating rules for
different classes of airspace, weather conditions, etc
and any equipment that may be required onboard the
aircraft.
Operator qualifications: the licensing and training
regimes for any pilots or crew need to be defined,
together with any periodic activities required to
maintain the currency of these qualifications.
Some General Trends for UAVs Before 1970’s
Special Projects
1970’s Technical Research
1980’s Scientific Applications
1990’s Military Applications
2000’s Commercial Applications
2010’s Growth of
New Industry
Economic
Viability
Time
Proof of
Concept
Build
Confidence
Increased
Acceptance
Routine
Operational
Use
Commercial
Products
Emerge
New
Sampling
Strategies
1980 2010 2000 1990
Single to multi-vehicle cooperatives (heterogeneous mix)
Segregated use to integrated operations (incr. autonomy)
Single to general/multi-purpose use (novel applications)
System Trends: HW vs. SW
Engineers Programmers
System
Factor 4.3 x 106
Hardware
Factor 1,000
Processors & Sub-Systems
IC Devices Chip
Architecture
Memory Bus Bandwidth Architecture
Software
Factor 43,000
Compiler(s)
Mathematical
Functions
Algorithmic Implementation
Operating System(s)
Users
Gilder’s Law O(2.9)/10yrs
Moore’s Law O(1.5)/10yrs
Brooks Law O(1.3)/10yrs
Chess Playing, Voice & Facial
Recognition O(2.9)/10yrs
Moore’s Law O(1.5)/10yrs
FFT/DFT & PDE Solver
O(N2) -O(NlogN)
Fastest Supercomputers
O(2.8)/10yrs
Linear Programming O(3.4)/10yrs BW 25%/year
Latency 5%/year
SW acquisitions 46% over-budget by avg 47%
Even “successful” projects have only 68% of
specified features (Niddifer, 2011)
Google cars 200-300MLOC (Frost/Sullivan 2009)
65M potential defects, 20% of which are high severity
90-95% of errors usually found prior to delivery
Still leaves 650,000 high severity defects
And needs 50-100 million test cases
Testing & Stability Implications
UAVs, Networks & Payloads EO/IR/Thermal
Multi/Hyper-spectral
Foliage Penetrating Radar
SAR and mmWave Radar
Communications Relay
Acoustic Detection & ID
Acoustic Countermeasures
Meteorology (Tomography)
Router/Internet in the Sky
Electronic Warfare/Systems
Radar Target/Repeaters
Meteorological Sensors
LIDAR & 3-D Imaging
Polarimetric Sensors
Biomimetic Cameras
Chemical/Biological Sensors
Large Files
eg. Images
& Video
Multi-Hop
Multiple
Paths &
Networks
Extended
Terrestrial
Footprint
Relay
Trials Data DSTO 2009
Electro-
optic vs
thermal
vs IR …
Hyper/multi-spectral
3-D Terrain Reconstruction
Trials Data
DSTO 2005 Van Den Hengel
et al, 2005
LIDAR & EO
Trials Data
DSTO 2006
3-D Terrain Mapping
Gibbins, Finn &
Swierkowski, 2006
Feature Analysis
Feature estimation aimed at determining terrain types (e.g. ground, vegetation,
buildings) and ground vehicle traversability
Gradient estimation based on local surface
fitting to the raw 3D scatter-point data
(bright regions indicate steep terrain)
Local Curvature Estimation based on local
surface fitting to the raw 3D scatter-point data
(bright/dark regions indicate potentially
undulating surfaces)
UAV Image Enhancement
Low-resolution UAV images High-resolution reconstruction
Buffer (9 consecutive frames)
Register Reconstruct
Low-resolution
High-
resolution
Trials Data
DSTO 2005 Gibbins & Swierkowski, 2006
Self noise level – 110dB
• Engine/Prop NB
• Engine/Prop BB
• Air Flow Noise
• Mechanical Noise
• Electrical noise
• Eddies
Aerosonde (k-twin) – 110dB
DA-42 Twin Star – 137dB Noise from target
• Engine/prop NB & BB
• Propagation loss
Acoustic Detection (360⁰ FOV)
Air & Ground Targets
All Weather
Tomography
Used in physics, medicine & remote sensing
Radon Transform
3D Wind & Temp Profiles
Modelling Biological Motion
Brinkworth & O'Carroll (2009)
EMD
Full bio-inspired vision model accurate across a range of
images for 3 decades of motion
Practical Outcomes 1
Traditional cameras
see either the dark parts…
…or the light parts of an
image
Visual processing in retina
allows both to be
seen…
…and remove redundancy
before sending it to the visual
cortex
Time domain processing on a per-pixel basis –
algorithm extremely parallel.
Practical Outcomes 2
Biomimetic
Original Post-processed
Brinkworth & O'Carroll (2008)
Good edge detection
Hard target detection
Small target detection
TYPICAL PERFORMANCE METRICS
Utilization of Operator • % of time the operator is busy servicing interactions
• Can be used to predict performance according to an Yerkes-Dodson curve
(Nehme et al., 2009). Mean/Max number of serviceable units/interactions
Utilization of Autonomous Machine • Average time of autonomous machines waiting for interaction
• Average number of autonomous machines waiting to be serviced
• Probability that autonomous machine must wait > X for interaction
Modelling HMI
Questions
Polarisation Compass
& Micro UAV Avionics
Polarisation sensitive facets of insect eyes
Rayleigh scattering causes sky polarisation
Mag compass unreliable in many situations
Sky Polarisation Pattern
Mars Steel Sheet Reinforced Concrete
Micro Avionics
All UAV/UAS
functionality
on a single
board incl.
antenna