1. Exploring the IOCT’s
IOCT s
large fleet of helicopter robots
Mario A Gongora & Benjamin N. Passow
A. N
2. About us
Dr. Mario A. Gongora
D M i AG
• Senior Lecturer
Benjamin N. Passow
• PhD student
Project: Using a fleet of helicopters to:
• research aspects of emergent societal behaviours
• sonic signatures to identify/control the fleet
• Creative way to achieve control?
• creativity in behaviour and artistic coordination?
3. The Fleet
• About 8 x
Twister Bell 47 helicopters
• Helicopter properties:
• 34 cm rotor span
• 210 grams net weight
• Max 100 grams payload
• Remote controlled
• 4 actuators (2 servos, 2 motors)
• flight duration: 10 minutes
g
4. How do they work
Degrees of freedom:
• 3 translational
• 3 rotational
Controlled by:
y
• Lift (overall rotors speed)
• Heading (difference rotors speed)
g( p)
• Pitch (rotor blade angle)
• Roll (rotor blade angle)
5. Difficulties
• Helicopters are unstable
and nonlinear systems
y
• Difficult to achieve stability:
• lik standing on a ball
like di b ll
• …so what do we need to do to:?
• Making them autonomous
• Letting them dance … without crashing!
6. Autonomy
• Autonomy = helicopter can control itself without
the need for remote control
• Embedded system developed
• H li t can now hover
Helicopter h
autonomously using (lightweight):
• Digital
Di it l compass
• Distances to ground
• Classical control methods
• Computation done on-board
7. HeRo in detail
Stabilising flybar
Dual Rotors
Digital Compass
(CMPS03)
Processing unit
(Microchip
dsPIC30F) 2 strong
DC motors
Battery pack 2 servos
(LiPo 7.4V)
3 Sonar Sensors
(SRF08)
8. What we have
• Current prototypes:
• Capable of hovering
at a predefined height
• Future work:
• Stable flight manoeuvres
• Controlled flight of a whole swarm
• Coordination among the swarm
• Application in research & performance art
10. Restrictions & Limitations
• DANGEROUS! No flying around people!
• Limited payload
• Remote: 100 gram max
• Autonomous: 60 gram max
• Limited flight duration: ~10 minutes
• Indoor use only
• Due to sensitivity to wind
11. Optimising the Controller
• Enhance stability by
evolving controller’s parameters
• Using Evolutionary Computing (GA)
• Optimising 5 parameters of PID controller
p gp
• Implemented on host computer
• Evaluation on real system
• GA can run fully automatic once started
12. GA Setup
• P ibl solutions evaluated on real helicopter
Possible l ti l td l h li t
• Helicopter bound to turn-table
• Controller to react to artificial perturbation to both sides
• Fitness inverse proportional to amount of error to set point
14. GA Results
• Found better solution than hand-tuning
• Noise and uncertainties
in real system:
• Significant variability
re-evaluating individuals
l ti i di id l
• Keeping GA running
• GA finds more
“consistent” solutions
• Less variability
y Fig. 1. GA (black) and hand tuned (gray) PID
controllers response to heading perturbed by 90◦ at
and more robustness t=0 and -90◦ at t=92. Mean of 12 individual tests for
each controller
15. Creative Approach
• We need to enhance the stability, achieve in-air
synchronisation as well as obstacle avoidance
• Could be done adding many sensors
• Helicopter would become too heavy
p y
• Instead we use a novel creative approach
• Using the intrinsic sound signature of the helicopter
• Not a single additional sensor is needed
16. Control using Sound
The helicopter s intrinsic sound signature is recorded and
helicopter’s
analysed by the HaRT robot
17. HaRT Robot
• HaRT - Humans and Robots Together
• for human-robot interaction
• (it doesn’t attempt to look humanoid)
• Is controlled by super-computer
• Recording and analysing
Helicopter sound signatures
• HaRT has microphones
to record sound signatures
• Super-computer analyses these
18. Sound = Information
• Motors and rotors generate vibrations
• Vibration = Sound
• Sound acquired by HaRT
• Sound = Information
• Where is the sound coming from? (Localisation)
• The power of the motors reflects on the sound
• The difference between motor speeds also affects the
sound
• Servos generate sound too
19. Enhance control
• Coordination information is fed back to helicopter
using a Bluetooth link
• Controller to incorporate this new information
• “Too far left” – Fly to the right
• “Other h li
h helicopter close to the right” – Fly a bit left
l h ih l bi l f
• Etc.
• Benefits:
• Enable in-air synchronisation
• Enhance stability
20. Aesthetic Applications
• HaRT to translate sonic signatures into musical calls
• Swarm of helicopters in formation flight and dancing
• Performance art using helicopter’s with coloured
trails
• Helicopter reacting to music / dancing (inverse
control of performance)
• Dancing helicopters with lights attached in darkness
i h li i h li h h di d k
• Many more…