The goal of the Clemson University Unmanned Aerial Vehicle (UAV) Laboratory is to promote the collaborative integration of recent developments in small-scale (<50 pound payload) aircraft, control system theory, and video and image processing to build new aerial platforms. Research activities center on developing, simulating, and testing new systems built with a mechatronic approach wherein the issues typically regarded as separate, e.g., flight control, path planning, position sensing, image processing, are combined and solved together. The scope of the mechatronic approach is shown in the system diagram in Figure 1. The expected result of this approach is a more efficient and effective use of system resources and enhanced task capabilities. The UAV platforms are targeted for applications that require automated launch and travel to focus area, precision hover and survey, and automated return. Example applications include crop monitoring, forest monitoring, security monitoring, and land topology monitoring.

System Block Diagram

Facilities and Equipment

The Unmanned Vehicle Laboratory is housed in the Robotics and Mechatronics Laboratory in the Department of Electrical Engineering at Clemson University. The laboratory is a general purpose facility to support control and robotic experiments and is well equipped for this purpose and includes: electric actuators, encoders, tachometers, laser displacement transducers, torque meters, signal conditioners, oscilloscopes, multimeters, function generators, scaling amplifiers, and Techron 2KW linear amplifiers, high speed vision systems for visual servoing, and QNX real-time workstations for data acquisition and control. Aircraft include the following:

	Description
	SR100 UAV Helicopter System Gasoline power plant 7 kg / 18 lbs Payload Capacity WAAS differential Safety/Manual Aircraft Controller & Transmitter 802.11-based Telemetry System Stable hover
	SR20 UAV Electric Helicopter System Electric Propulsion Motor 10lbs Payload Capacity WAAS differential included Ready-to-Fly Autonomous Safety/Manual Aircraft Controller & Transmitter 802.11-based Telemetry System Stable hover More about our SR20s ...
	Draganflyer X-Pro Four rotor electric, radio controlled, electronically stabilized flying platform. Full pitch, roll, yaw, and altitude control using conventional helicopter inputs. Dimensions are 55.5" from rotor tip to rotor tip and 7" high. 4.8 volt 7800mAh Li-Poly rechargeable battery Professional Quality Pan and Tilt CCD Color Videocamera with 900Mhz Transmitter and Reciever 1lb payload capacity
	Gohbee Stinger 50 Heli .50 radio controlled helicopter 600mm Carbon Fiber Blades Main Rotor Diameter: 1348mm (53.1") Fully Equipped Weight: 2850g (6.25lbs) Futaba controller and servos OS Engine
	The UAV Trailer Before Modifications.
Need a picture of the field here	Directions to field: Take US-76 toward Anderson (~3miles) Turn right at W Queen St (Signs for Garrison Arena) (~.8 miles) Continue straight onto Fants Grove Rd/SC-S-4-56 (Pass Garrison Arena) (1.1 miles) Turn left at Fants Grove Rd/SC-S-4-1098 (~.4 miles) Google Map Satelite Image

Research Activity

The potential for multi-bladed UAVs in applications as diverse as fire fighting, emergency response, military and civilian surveillance, crop monitoring, and geographical registration has been well established. Many research groups have provided convincing demonstrations of the utility of UAVs in these applications; however, there is still a large chasm between the anticipated “tool of the future” and currently available systems.

Flight Control

The challenges facing further development lie along several fronts but one of the most fundamental issues is to ensure that the craft can move to or hold a desired position and orientation. Specifically, as shown in Figure 2 the aircraft must be able to move from a current location to a new desired position (denoted by triple x,y,z) and achieve a new orientation (denoted by roll, pitch, and yaw angles).
This is one function of the low-level control block in Figure 1. It is at the low-level control that the peculiarities of the multi-bladed UAV system such as nonlinearities and the fundamental fact that the system is under-actuated must be addressed. An under-actuated system is especially challenging to control since it has fewer control inputs than degrees of freedom, i.e. it has degrees of freedom that cannot be directly actuated. In order to achieve high overall performance the low-level control problem must be solved. The control problem will be approached using Lyapunov-based control design techniques adopted primarily from the field of robotics.

Sample Collection

Interaction of the aerial vehicle with a ground based target creates the additional kinematic and dynamic complexities shown in Figure 3. Of particular significance is that in order to interact with the object the robot must be able to apply and regulate the end-effector force, F, in Figure 3. This further exacerbates the under-actuated control problem described above.

Scalable Aircraft

The ideal UAV to serve the expected applications will possess the following characteristics:
• Scalablility
• Agility
• Flight duration
• Accuracy
• Precision
• Cost
• Reliability
The vision of the Unmanned Aerial Vehicle Laboratory is to design an aerial platform that can be easily tailored to best match specific needs. The ideal situation is a technology platform that is both scalable, Figure 5, and modular. The system should be scalable in the sense that the same basic structure can be used to build vehicles with different payload capacities, i.e., the major components can be made smaller to build a craft that can carry less weight. The system should be modular in the sense that a particular advancement of technology can be rapidly incorporated into the system.

Terrain Reconstruction

One of the potential applications of the UAV is to reconstruct the 3D terrain below. Instead of the common lidar approach, we are developing techniques to perform this reconstruction using video cameras. This alternative has several distinct advantages: potentially higher accuracy and denser sampling, automatic registration of the 3D geometry with images of the scene, and lower power and cost. Our approach involves tracking feature points throughout the scene and using nonlinear estimation techniques to compute the velocity of the camera and the 3D coordinates of the points in a provably convergent manner. We have successfully applied the method to indoor scenes in our laboratory, and we are in the process of extending its applicability to larger outdoor settings.

Image and Video Processing

With a camera on-board the vehicle, algorithms can be developed to exploit the rich information content of the available image data to give the vehicle additional knowledge about its surroundings. By tracking features points on the ground, the so-called “time-to-impact” can be computed to yield crucial information about the height of the craft above the ground, a value that is difficult to determine from GPS sensors. By performing texture segmentation and classification of the images, important areas such as forests, marshes, the coast, and the sea can be determined and measured. Motion segmentation and model-based tracking can yield man-made structures such as buildings and bridges in order to perform obstacle avoidance.

Projects

The projects in the UAV lab include:

• Closed Radio-link Loop - To close the loop on a controller, a robotic system must read in the sensor data to determine the next feedback outputs. This requires sensors, a processing unit, and an interface between the two. On a UAV weight is a huge concern when trying to implement a controller. It is possible to use a lightweight processor for implementing a controller onboard, however many controllers in modern control are too complicated for a simple micro controller to handle. This project investigates the idea of using a radio link to send and recieve the data required for a full state controller. By placing an IMU on the UAV and remoting all information back to a ground computer, all calculations for the control algorithm will be made and sent back to the helicopter via it's transmitter.
  
• PID control on trainer - After building the trainer, a simple PID controller will be implemented on the helicopter as certain bugs are worked out. The trainer setup allows for rapid prototyping of control ideas while offering a variety of contraints on the helicopter, including z, yaw (and small variations of x, y, roll, and pitch)
  
• Output state feedback control - An inertia measurement unit (IMU) can measure the angular rate and acceleration of the IMU. While GPS, magnetometers and inclinometers can measure the location and orientation directly, they have many short comings. GPS and magnetometers are not very percise and slow to update, inclinometers may malfunction due to vibration and they still cannot measure yaw. So to get the actual position and orientation from the IMU, certain intergrations must be executed. However many IMU's like the MIDG II INS automatically do this. Once the position and orientation of the IMU are aquired, a controller will be implemented using only these states.
  
• Balloon - Current electric helicopter can only last about 15 minutes in the air. This kind of life span can severly limit the number of applications for such a helicopter. In an attempt to extend the flight time of a helicopter, a helium filled balloon will be attached to the UAV. Like a blimp, the balloon will relieve some of the weight and add a dampening effect to the helicopter dynamics. Unlike a blimp, the UAV will still have 4 degrees of freedom.

  Here are some movies of the Ballon project
  Setting up the DraganFlyer
  The DraganFlyer-Balloon in flight
  The Dragan's eye view

• Genentic algorithms - Genetic algorithms will be investigated in an attempt to find control parameters. A MATLAB model of a control system will be used while genetic algorithsm try to find the best control paramters.

E-Mail

CRB Home Page