Supervisor: A/Prof. Tomonari Furukawa
PhD Student: Lin Chi Mak (Localisation & Control: Mechanical, Electrical & Software Design)
Undergraduate Student: Mark Whitty (Control: Electrical & Software Design)

Contact
For more information, please contact:
Lin Chi (Edward) Mak

Background

Micro Aerial Vehicles (MAVs), defined as flying vehicles which have no dimension longer than 0.3 m (defined in MAV08), have wide indoor robotics applications such as Urban Search And Rescue (USAR) and surveillance. Due to the high maneuverability of rotary-wing MAVs, they have distinct advantages over fixed-wing and flapping-wing MAVs. The development of autonomous indoor rotary-wing MAVs requires the corresponding development of robust pose and position estimation systems. The key challenges in such systems are the unavailability of GPS data, the limited sizes and payloads of the MAVs.

This project is aimed at developing a localisation and control system for MAVs with a ground vehicle or a group of MAVs. The required on-board devices must be suitable for MAVs in terms of weight, size and power consumption. Also, the system should work in any unknown environment without any infrastructure.

In this project, a commercially available 4 channels radio-controlled coaxial helicopter (Fig.1) within 0.3m is chosen and modified. Thanks to the weighted flybar attached to the top blades, it is easier to be controlled comparing to conventional helicopter. To track its 6 Degrees-of-Freedom (DoF) motion, 2 blue blade-mounted LEDs form an ellipse in the captured image (Fig.2) (See Video 3). Since the width of the ellipse is independent on the orientation of the helicopter, it gives the distance (x) between the helicopter and the camera. The center of the ellipse gives the y- and z-coordinates of the helicopter. An additional red LED is located at the tail for predicting the yaw.

The developed system is the first 6 DoF visual localisation system for helicopters using only 1 off-board camera and 3 LEDs. At least 4 LEDs and 1 off-board camera or 3 LEDs and 2 off-board camera are required in other existing systems. Moreover, the strength of the developed system is the camera can be mounted at any positions in the Line-of-Sight of the LEDs.

Based on the proposed localisation technique, a control system for the helicopter has been developing. Fig. 3 shows the block diagram of the control system with visual feedback. The computer sends the analog signals to the remote controller through D/A converter to change the control actions of the helicopter. Due to the complexity and non-linearity of the co-axial helicopter's aerodynamics, it is very difficult to obtain the helicopter aerodynamics model and the optimal controller. To overcome this problem, a lift force measurement system with different control inputs has been developed. The current approach for the design of 4 PD controllers is to use Ziegler Nichols tuning method. Video 4 shows how to tune the PD values based on the response of the helicopter. In the video, the helicopter was controlled by a Proportional control. The period of the oscillation was recorded for tuning the components in the PID controller.

The visual tracking system has been developed successfully. Fig.4 shows the recognized ellipse (red dash line) on the flying helicopter (See Video 1). Fig.5 and Fig.6 depict the real-time positioning ability of the system and the 3D positioning results (See Video 2) and the accuracies of the developed system at different ranges. The 3D RMS positioning error of the system are 2-3.5% at the range varied from 1 to 8m from the camera. The pose estimation RMS error is 4 degree at the distance of 2 m from the camera.

The computer-controlled system has been developed and tested successfully. Video 7 shows the successful 3D position and yaw control by PD controllers. Using the oscillating period in Video 4, the ratio between P and D in the height controller is obtained. The PD controller is good enough as the steady state error is small. In the current control system, the MAV is able to hover around a point with 0.2 m (RMS) 3D positioning error.

This project is still on-going. The progress will be updated.

Videos
1. Recognition of Ellipse - Play Video, Save Video
2. Localisation of a Flying MAV Version 1- Play Video, Save Video
3. Manually-Controlled MAV Version 2 Flight Test- Play Video, Save Video
4. Computer-Controlled MAV Version 2 Flight Test for PID tuning- Play Video, Save Video
5. PD-Altitude-Only-Controlled MAV Version 2 Flight Test- Play Video, Save Video
6. Height and Yaw Control of a MAV- Play Video, Save Video
7. 3D Position and Yaw Control of a MAV- Play Video, Save Video
8. Autonomous MAV for USAR- Play Video, Save Video

P.S. The red dash line represents the estimated best fit ellipse for the detected cyan colour.

Publication

1. Lin Chi Mak, Mark Whitty, Hang Xu, Kai Zhan and Tomonari Furukawa, "Visual Localization and Control of an Indoor Rotary-Wing Micro Aerial Vehicle", IEEE/RSJ 2008 International Conference on Intelligent Robots and Systems (IROS2008) Workshop on Visual guidance systems for small autonomous aerial vehicles, Nice, France, Sep. 22-26, 2008, submitted.

2. Lin Chi Mak, Mark Whitty and Tomonari Furukawa, "A localisation system for an indoor rotary-wing MAV using blade mounted LEDs", Sensor Review, Emerald, Vol. 28, Issue 2, pp. 125-131, 2008.

3. Lin Chi Mak and Tomonari Furukawa, "A 6 DoF Visual Tracking System for a Miniature Helicopter", 2nd International Conference on Sensing Technology (ICST ‘07), Palmerston North, New Zealand, Nov. 26-28, 2007, 6 pages.