Autonomous Aerial Robots in NearEarth Environments

1
Autonomous Aerial Robots inNear-Earth
Environments

• William E. Green
• Drexel Autonomous Systems Lab (DASL)
• Mechanical Engineering, Drexel University

Koerner Fellows Presentation 06/09/04
2
Unmanned Air Vehicles Advocacy
I hope Ive demonstrated personally my commitment
to UAVs, and I am committed You can count on
the fact that my commitment will
continue. Gen. John Jumper, Air Force Chief of
Staff, Senate Confirmation Hearings, September
2001
It shall be a goal of the Armed Forces to
achieve the fielding of unmanned, remotely
controlled technology such that by 2010,
one-third of the aircraft in the operational deep
strike force aircraft fleet are unmanned. 2001
Defense Authorization Conference Bill - H.R. 4205
Unmanned aerospace
vehicles
(UAVs)
are the hallmark
of our future Dr. James
Roche, Secretary of the Air Force,
October 10, 2001
I do NOT believe that the
current DOD inventory of airborne
intelligence platforms or programmed procurement
of additional assets is
adequate to satisfy all
of the CINCs' requirements. Marine Gen. Peter
Pace,
Vice Chairman, Joint Chiefs of Staff, then
Commander of US Southern Command,
September 6, 2000
3
Classes of UAVs
Improving UAV reliability (autonomy) is the
single most immediate and long reaching need to
ensure their success Office of the Secretary
of Defense UAV Roadmap 2002-2027
4
Near-Earth Environments
Characteristics
Labor Intensive

• Rugged

• Target Localization

• Rich with obstacles

• Search-and-Rescue

• Poor GPS

• Recon and Surveillance

• Degraded communications

• Inspection, Assessment

5
What Type of Aircraft?
Specifications

• Fly through small openings

• Carry a payload

• Fly slowly and safely

4 mph
Market East Station
banquet table
Green, W.E., Oh, P.Y., An Aerial Robot Prototype
for Situational Awareness in Closed Quarters
IEEE 2003
International Conference of Intelligent Robots
and Systems
6
What Type of Sensors?
Collision Avoidance

• Turn away when optic flow high

Speed Control

• Constant optic flow

Hover, Gust Stabilization

• Zero out optic flow

7
From Insects to MAVs

• Implementation
• Mount sensors on nose and belly
• Incoming collisions and altitude
• Rapidly expanding region on right
• 1D Theory
• OF (V/D) sin q - w
• Computationally expensive?

Green, W.E., Oh, P.Y., et al, Flying Insect
Inspired Vision for Autonomous Aerial Robot
Maneuvers in Near-Earth Environments IEEE 2004
International Conference of Robotics and
Automation
8
Previous Research
Green, W.E., Oh, P.Y., et al, Autonomous Landing
for Indoor Flying Robots Using Optic Flow
2003 ASME
International Mechanical Engineering Congress and
Exposition
9
A Closer Look

• Focus of Expansion (FOE)
• Optic flow vectors radiate
• Pure translation yields zero OF
• Obstacles in line with optical axis
• Small appear to get larger
• Large will not be detected!!

Platform
Cant Fly in Wind!!
Cant Hover!!
Low Endurance
Large Inertia
Difficult to Control
10
Autonomous Flight in Caves Tunnels
What configuration?
What type of sensors?
Wingspan 44 inches Weight 10 ounces Max Speed
15 MPH Endurance 20 minutes
11
Near-Earth Platform

• Designed to Hover!
• High thrust-to-weight ratio (T/W1.52)
• Aircraft wt. balanced by motor thrust
• Allows rapid transition thru stall regime
• Large control surface area
• Capable of flying in tight spaces
• Failsafe collision avoidance maneuver

12
Future Roadmap
June-July
August-Sept
Oct-Nov
Dec-Jan
Backpackable CF Prototype
Autonomous Hover-To-Cruise Transition
Autonomous Cruise-To-Hover Transition
Aircraft Model
Develop Aircraft Simulator
Begin Autonomous Flight Tests In Caves
Autonomous Hovering
13
Contributions

• Slow flying fixed-wing test bed vehicle
• Non-GPS local processing based sensor suite
• Demonstrated autonomous tasks in near-Earth
  environment
• Collision avoidance
• Landing
• Altitude hold

14
Conclusions
Localization
Collision Avoidance

• IMUs initially too heavy

• Must detect obstacle

• Now, lighter and more compact

• Weve proven this is possible

• Maps can then be built (SLAM)

• Then comes path planning

15
Thank You!
Special Thanks to Dr. Koerner and His Family
16
(No Transcript)
17
Gimbal Lock
18
Optic Flow Microsensors
Sensor architecture Imaging fabric mesh
includes photo sensing and analog
preprocessing Other circuitry performs feature
detection and digitization
2-D imager
Contrast Enhance
Feature Detection
Digitization
Low BW signal
MOP µcontroller Biomimetic motion detection
algorithms and I/O.
µC
1-D imager
Photo- receptors
Edge Detectors
Vision Chip Morphology
Vision chip outputs edge locations.
Microcontroller need only track edge movements.
Winner-Take-All (WTA)
To mux and vision chip output
19
Centering Response
Hypothesis

• Bees fly through openings by balancing image
  speed to left and right
• Trained bees to fly through tunnel
• Vertical stripe pattern on side walls
• Grating on one wall could be moved at any desired
  speed and direction
• Results confirm that the bees balance the speeds
  of the retinal images in the two eyes and not the
  contrast frequencies

Experiment
Srinivasan et al., Journal of Experimental Biology
20
Additional Issues

• Questions
• Is local computing necessary?
• Range of wireless cameras in NEE?
• Can small obstacles (wire) be detected?
• Soon enough to be avoided?
• Alternative to following GPS waypoints?

21
2005 Inaugural Competition

• Target Identification
• Identify survivors
• Deploy beacon to pinpoint location
• Determine video tx range

• Collision Avoidance
• Demonstrate collision avoidance
• line following
• Gust stabilization

Manual Control
Autonomous Control