Autonomous Cargo Transport System for an Unmanned Aerial Vehicle, using

Visual Servoing

Noah Kuntz and Paul Oh

Drexel Autonomous Systems Laboratory

Drexel University, Philadelphia, PA

Motivation

Helicopter cargo transport requires dangerous sling-load attachment maneuvers

Cargo must often be delivered to high risk areas, endangering the crew

HOWEVER

Helicopter cargo transport using allows delivery of payload to otherwise unreachable areas

Pictures source: http://www.mccoy.army.mil/ReadingRoom/Triad/06112004/Sling-load%20Sinai.htm

UAVs CAN FIX THIS!

Potential Cargo

Medicine Specialized parts or tools for in-field repair UGVs for bomb disposal or surveillance

Such as the Bombot, a low cost compact bomb disposal robot manufactured by the West Virginia High Technology Consortium (WVHTC) Foundation

Left picture source: http://robotgossip.blogspot.com/2006/01/bombot-to-be-built-in-west-virginia.html

Helicopter Cargo Carrying Tests

Test cargo was a small remote control UGV, for potential UGV/UAV teaming missions

Computer controlled takeoff, flight, and landing Demonstrated suitability of the SR-100

unmanned helicopter for light cargo transport

SR-100 platform proves capable

Cargo Carrying Methods

Fixed Cargo Bay CONs – Requires landing, limited cargo size,

decrease in maneuverability PROs – Cargo is protected and stable

Sling Load CONs – Oscillation danger, difficult attachment PROs – Common, allows diverse cargo

Actuated Hook CONs – Limits weight of cargo PROs – Can provide active damping, allows

autonomous attachment

Actuated Hook Wins for Unmanned Heli

Concept of Operations

SR-100 is capable of Autonomous

takeoff.

Autonomous hovering and GPS

waypoint navigation is

integral to the SR-100’s control

package.

Tracking is performed with visual servoing using onboard

camera and computer.

When criteria are met for proximity to the target, the hook is servoed

through the target loop.

1 Takeoff 2 GPS Waypoint Navigation 3 Track Cargo 4 Hook Cargo

Concept of Operations

5 Increase Altitude

6 GPS Waypoint Navigation

Unhook Cargo7

The cargo will be set on the ground

and the hook retracted.

The cargo will then be lifted off the

ground.

GPS navigation will occur again.

Technical Requirements

Accurate tracking in all lighting conditions

Reliable cargo pickup

Weight within capability of the helicopter

Research Path

Establish load carrying ability of unmanned helicopter platform

Set up hardware-in-the-loop simulation environment for testing and evaluation

Develop the cargo pickup system in test environment

Refine system and retest

Flight test the system, for verification and validation

Challenges

Overall “Mobile Manipulation” problem

Tracking target under variant lighting

Tracking while helicopter wanders

Servoing the hook fast enough

Systems Integrated Sensor Test Rig (SISTR)

6DOF capable with velocity control

Environmental simulation including lighting control

Allows recreation of flight conditions for testing and evaluation

Sponsored by the National Science Foundation

SISTR Flight Data Playback

Recreate helicopter motion under controlled condition

Encoder data validates the gantry velocity controller

SISTR replicates flight movements

Mechanism Notional

Gantry Arm

Control Computer

Manipulator PTU

Target

Fiducials

Camera

IR Filter

Camera PTU

Manipulator

Batteries

Mechanism

2DOF stepper motor camera PTU for high speed and precision

2DOF hook PTU for high torque, low cost, and light weight

Vision

Structured lighting approach used for initial testing

Target uses krypton bulbs as fiducials, with high IR emission

IR band-pass filter removes non-infrared light

Threshholding operation isolates fiducials which are tracked using image-based pose regulation

Simple tracking for low computation / high speed

Controller

Control Computer Mini-ITX single board computer Solid state drive for vibration resistance

Testing Procedure

Gantry replays recorded helicopter velocities Target is placed in each of nine positions within 20 cm

(GPS accuracy) from ideal

Testing

Results

Near-miss conditions could be eliminated Success rate of ~83% should be possible with minor

improvements Closed loop pickup detection will improve

Contributions + Future Work

Objectives Met Accurate tracking in all lighting conditions

Tracking demonstrated under most difficult condition

Consistent cargo pickup 61% - work in progress

Weight within capability of the helicopter ~ 15 lbs, within 20 lb limit

Results will be confirmed with flight tests

Acknowledgements

National Science Foundation US Army Telemedicine Advanced

Technology Research Center (TATRC) Piasecki Aircraft Inc

For more info please see: http://www.pages.drexel.edu/~nk752/