The main goal of the project was to develop an aerial robot or Unmanned Aerial Vehicle (UAV) to help in the inspection of power lines. A modified model helicopter, in which are embedded camera, sensors and additional hardware, was used for the task.

The main activities executed in the scope of the project were: customization of the model helicopter; embedded software to carry out navigation and automatic pilot tasks; embedded software to perform image acquisition; flight stabilization control strategies; communication link between helicopter and base station; simulator and supervisory software under Linux environment; computer vision techniques to help in the inspection of power lines.

Among those specific goals, my main activity concerned the development of the 3D localization system prototype with low cost sensors and the development of sensor fusion algorithms to improve the estimation provided by this system. Regarding the sensors used in the project, inertial sensors, magnetometers, pressure sensor and GPS were chosen. The algorithms to perform sensor fusion were mainly based on the Extended Kalman Filter (EKF) and the Sigma-Point Kalman Filter (SPKF). Related publications may be found
here.

Below is a real-time demo of the attitude estimated by the system. Estimations provided by the isolated sensors and the sensor fusion are compared.


The entire 3D localization system was tested in experiments where the system was mounted in a truck which was driven across the university campus. The following figure illustrates the system performance when all the available sensors were integrated. The estimated position is shown in orange, the attitude in a RGB frame and the GPS positions are represented by yellow dots. The purple arrows point to locations where the GPS receiver did not reach a valid solution. Y axis [m] represents the North direction, while the X axis [m] represents points to the East.

3DlocalizationAtUnB

.