Obstacle Avoidance System Design of An Unmanned Vehicle with Stereo Vision
Date Issued
2009
Date
2009
Author(s)
Shan, Yu-May
Abstract
The main theme of this thesis is to develop an outdoor navigation vehicle system. This system consists of three parts which are the main station, reference station, and an unmanned vehicle. The GPS receiver, electronic compass, stereo camera, and six ultrasonic sensors are integrated into the unmanned vehicle. The obstacle avoidance is performed by the vehicle which can detect the shape and position of the obstacles within a detectable range in the direction of its motion. The position of the unmanned vehicle is provided by GPS in an obstacle-free situation with the help of Triple Difference Carrier Phase method KGPS. The Heading angle is obtained through the output of an electronic compass. In addition, the main station can monitor the performance and behavior of the vehicle via the data exchange mechanism constructed by several sub-systems and the wireless network. With stereo vision, the geographic information ahead of the vehicle is obtained by the principle of plane induced parallax. The profiles of left and right vision are acquired by Canny Edge Detection algorithm. Then, an intersection line of the obstacle and the level ground is found by the way of image processing. Finally, the position and distance of the obstacle is calculated by the homography matrix method. The method of potential field is used to perform obstacles avoidance. The distance between the vehicle and the destination is used to compute the attractive force and those between the vehicle and obstacles are used to find the repulsive force. Gaussian distribution-like potential energy is derived for each obstacle. Then, the unmanned vehicle is required to move on a smooth curve with lowest potential energy by controlling its angular velocity and the motion velocity. For real time obstacle avoidance, the input and output circuits of the six ultrasonic sensors are designed by VHDL language and implemented by FPGA development kit. The information of distance is obtained through the ultrasonic sensors which are managed by the design controller. Finally, the real time moving obstacle avoidance is performed by the angle and velocity that are inferred from the rule-based system. Integrating the above-distributed schemes unmanned vehicle can avoid collisions with obstacles and arrive at the destination in different scenario. Experimental results show that the proposed navigation and control methodology are practicable.
Subjects
Stereo camera
Stereo vision
Plane induced parallax
Potential field model method
Navigation obstacle control
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-98-R96543005-1.pdf
Size
23.53 KB
Format
Adobe PDF
Checksum
(MD5):6275ce36098cde0a27802de96e40c339