1. Computational Mechanics and Robotics The University of New South Wales
Micro Aerial Vehicles for Search, Tracking and Reconnaissance (MAVSTAR)
Computational Mechanics and Robotics
ARC Centre of Excellence in Autonomous Systems
The University of New South Wales
Australia
Tomonari Furukawa

2. Content Conceptual solution Overall System Architecture MAV
Propulsion and lift system
Guidance, Navigation and Control
Flight termination system and Safety
UGV
Electronic
Mechanical BS
Man/Machine Interface
Path Planning
Conclusion & Future Work
MAVSTAR, UNSW, AU

3. Conceptual Solution 4 Remote control MAVs 4 Autonomous UGVs BS
Strategy:
1 MAV takes off from the IP
2 MAVs are carried by UGV
1 MAV stays in IP
MAV tasks:
Search and locate for obstacles, mines, guard and hostage
Search, locate and defuse mines and locate hostage
UGV tasks:
Relay MAVs’ signals
Search and locate obstacles, guard and hostage
Deactivate mines

4. Overall System Architecture
Communication between MAV and BS:
Direct (LoS, >1km)
Relay (NLoS)
Developed GUI:
Sensor monitoring
Human-in-the-loop control
Kill Switch
Video: 1.2, 1.5 and 5.8 GHz
Data: 2.4 GHz
Radio Frequency Interference (RFI)
MAVSTAR, UNSW, AU

5. MAV Platform Flybar Fully custom-made Low cost (US$500) Carbon blades
4 plies
45deg/0deg/0deg/45deg
IMU
Carbon frames
3 plies
0deg/90deg/0deg
GPS
Camera
Ultrasonic sensor
Battery
Video Transmitter

6. MAV: Propulsion and lift system
Coaxial helicopter
2 brushless motors for height and yaw control
2 servos for pitch and roll control
Advantages:
Fit within 30cm sphere
High lift force (455g) for all sensors
12-14 minutes flight time
Stable in indoor and outdoor environments (Wind speed up to 20km/h)
MAVSTAR, UNSW, AU

7. Indoor and Outdoor Performance
Indoor flight
Outdoor flight
Indoor flight
Outdoor flight

8. Guidance, Navigation and Control
Sensors
Control Sensor: 2 axis accelerometer, 1 axis gyro
Navigation: GPS, compass, ultrasonic
Mission and obstacle avoidance: CCD camera (Range: >1km)
Manual remote control
Base on real-time video image
Have all sensors for autonomous control later
Minimize the on-board computation
MAVSTAR, UNSW, AU

9. Camera view during flight
Sensing Capability
Camera view during flight

10. Flight termination system and Safety
Dangerous to people
Kill switch in BS is activated
Kill command is sent to MAV
Lost communication
Hover in first 2s
After 2s, land using Ultrasonic Range Finder
MAVSTAR, UNSW, AU

11. UGV: Mechanical Custom-made frame 4WD Off-road capability
Able to climb steps
14km/h
1 hour run time
Directional Microphone
Hostage Detection
MAVSTAR, UNSW, AU

12. UGV: Mechanical Custom-made frame 4WD Off-road capability
Able to climb steps
14km/h
1 hour run time
Directional Microphone
Hostage Detection
Launching Mechanism
Power Saving
MAVSTAR, UNSW, AU

13. UGV: Electronic Navigation Sensors: GPS, Compass, Accelerometers
Collision Avoidance: CCD camera, Ultrasonic Range Finder
Mission Sensors (for Mine and Hostage):
CCD camera, Directional Microphone
Autonomous Control for 4 UGVs
1 Crew member
Monitor and override control if necessary
Update waypoints if necessary
Repeater for MAVs’ video and data signals
For MAVs in NLoS condition
MAVSTAR, UNSW, AU

14. BS: Man/Machine Interface
UGV
UGV
UGV
UGV
MAV
MAV
MAV
MAV
Monitor
Server
Controller
Comm
Comm
Comm
Controller
Controller
Controller
Data server
Data server
Data server
Data server
MONITOR
MAV
MONITOR
MAV
MONITOR
UGV
Data flow
Distributed Server-Client Model for sharing
information
Scalability for multiple/heterogeneous MAV/UGV coordination
Rich computational resources
(rapid prototyping/testing purpose)
Multiple operators
MAVSTAR, UNSW, AU

15. BS: Path Planning Location of the guard (SAT MAV)
The system suggests the path to get close to the building with Search-And-Tracking MAV information.
Non-Line of Sight (NLoS) from the guard
Non-Line of Sight estimator utilizes a priori knowledge to find “Safe Zone”.
Path finder
Theoretically, it is guaranteed that the system provides a path from the entry point to the building if there is a path. GV
NLOS
Search space: 500x500 grids
0.53sec for the initial search by Pentium-M 1.1GHz processor
MAVSTAR, UNSW, AU

16. Conclusion Conclusions: 4 MAVs, 4 UGVs and BS
Rotary-wing MAV with coaxial setup fits within 30cm sphere
High lift force, Stable in indoor and outdoor, long flight time
Autonomous control for UGVs and Manual control for MAVs
MAVSTAR, UNSW, AU

17. Future Work – Coordinated Information-theoretic Search and Tracking
Coordinated information-theoretic SAT
Real-time information-theoretic SAT
Coordinated information-theoretic SAT
Real-time information-theoretic SAT

18. Future Work – Continuous Outdoor and Indoor Localization

19. Scientific Research (AFOSR) Air Force Research Lab (AFRL)
Acknowledgements
Air Force Office of
Scientific Research (AFOSR)
Air Force Research Lab (AFRL)
DSTO
Defence Science and
Technology Organisation
MAVSTAR, UNSW, AU