Browsing by Author "Jardine, Tyron"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- ItemAn autonomous robot to clean heliostats(Stellenbosch : Stellenbosch University, 2020-12) Jardine, Tyron; Smit, W. J.; Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.ENGLISH ABSTRACT: This thesis details the design and testing of a proof-of-concept robot capable of cleaning heliostats in concentrated solar power (CSP) plants. Heliostats accumulate dust, causing a decrease in their reflectivity, decreasing their ability to provide heat to the system. Current heliostat cleaning methods include large trucks which lack automation and use large amounts of water. The challenge was thus to design and build a robot capable of navigating heliostats of various sizes. The robot localises itself on the heliostat using a camera, ultrasonic sensors and wheel encoders. The camera makes use of Hough line detection to detect the edges of the heliostat and orientate the robot when near the middle of the mirror, and the ultrasonic sensors were used to detect when the robot was at the edges. The encoders ensure the robot can keep track of its movement. A pinhole camera model was assumed for the camera because it has been calibrated. This allowed the 2D image co-ordinates of the heliostat edge lines to be transformed to real world 3D co-ordinates and to determine the orientation of the robot relative to the line. The extrinsic parameters of the pinhole camera model were derived both experimentally and theoretically, but the experimental values yielded more accurate line prediction results. It was seen that a higher camera resolution yielded better predictions of the robot’s orientation, but the further away the robot was from the line the less capable it was of reliably determining its orientation. Higher Hough line thresholds also yielded less reliable results. The robot could navigate the heliostat by making use of the ultrasonic sensors and wheel encoders when camera data was not available. The robot experienced problems with navigation at the heliostat corners and if the heliostat had short edges, but solutions to these problems are provided. A proof of- concept robot was thus designed, built, and tested that could navigate the surface of a heliostat using a camera, ultrasonic sensors, and wheel encoders. Artificial intelligence or machine learning can also be implemented with the camera to improve the system, but the robot has the potential to address the challenges faced in the existing cleaning methods of heliostats in large scale concentrated solar power plants.