Nakkar, Ruaa (2026) Laser-Guided Control for Multicopter Uavs. Masters thesis, King Fahd University of Petroleum and Minerals.

PDF Ruaa Nakkar- M.S. Thesis.pdf Restricted to Repository staff only until 22 July 2027. Download (14MB)

Arabic Abstract

ﺗﻘﺪﻡ ﻫﺬﻩ ﺍﻟﺮﺳﺎﻟﺔ ﻣﻨﻬﺠﻴﺔ ﻗﺎﺋﻤﺔ ﻋﻠﻰ ﺍﻟﺮﺅﻳﺔ ﺍﻟﺤﺎﺳﻮﺑﻴﺔ ﻟﺘﺤﺴﻴﻦ ﺩﻗﺔ ﺗﺘﺒﻊ ﺍﻟﻤﺮﻛﺒﺔ ﻟﻠﻤﺴﺎﺭ ﺍﻟﻤﺤﺪﺩ. ﻭﺗﻌﺘﻤﺪ ﺍﻟﻤﻨﻬﺠﻴﺔ ﻋﻠﻰ ﺭﺻﺪ ﺑﻘﻌﺔ ﻟﻴﺰﺭ ﻣﺴﻘﻄﺔ ﻭﺧﻂ ﻣﺮﺟﻌﻲ ﺑﺎﺳﺘﺨﺪﺍﻡ ﻣﻌﺎﻟﺠﺔ ﺍﻟﺼﻮﺭ ﻭﺍﻟﺘﺠﺰﺋﺔ ﺍﻟﺪﻟﺎﻟﻴﺔ، ﺛﻢ ﺗﻘﺪﻳﺮ ﺍﻧﺤﺮﺍﻑ ﺍﻟﻤﺮﻛﺒﺔ ﻋﻦ ﺍﻟﻤﺴﺎﺭ ﺍﻟﻤﻄﻠﻮﺏ. ﻭﻳُﺴﺘﺨﺪﻡ ﺍﻟﺨﻄﺄ ﺍﻟﺒﺼﺮﻱ ﺍﻟﻨﺎﺗﺞ ﺇﺷﺎﺭﺓ ً ﺇﺿﺎﻓﻴﺔ ً ﻟﻠﺘﻐﺬﻳﺔ ﺍﻟﺮﺍﺟﻌﺔ ﺿﻤﻦ ﻣﺘﺤﻜﻢ ﺍﻟﺤﺮﻛﺔ. ﻃﻮﺭ ﻧﻈﺎﻡ ﺍﻟﺮﺅﻳﺔ ﺃﻭﻟﺎ، ﻭﻗﻴِّﻢ ﻣﻦ ﺣﻴﺚ ﺩﻗﺔ ﺍﻟﺘﺠﺰﺋﺔ ﻭﺯﻣﻦ ﻣﻌﺎﻟﺠﺔ ﺍﻟﺼﻮﺭ، ﺛﻢ ﺷُﻐﻞ ﻋﻠﻰ ﻣﺘﻦ ﺍﻟﻤﺮﻛﺒﺔ ﻟﺘﻘﺪﻳﺮ ﺍﻟﺨﻄﺄ ﺍﻟﺒﺼﺮﻱ ﺁﻧﻴًﺎ. ﻭﺑﻌﺪ ﺫﻟﻚ، ﺩ ُﻣﺠﺖ ﺍﻟﺈﺷﺎﺭﺓ ﺍﻟﻨﺎﺗﺠﺔ ﻓﻲ ﺍﻟﻤﺘﺤﻜﻢ، ﻭﺍﺧﺘُﺒﺮ ﺍﻟﻨﻈﺎﻡ ﺗﺠﺮﻳﺒﻴًﺎ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺇﻋﺪﺍﺩﺍﺕ ﻣﺨﺘﻠﻔﺔ ﻟﻠﻤﺘﺤﻜﻢ ﻭﺍﺗﺠﺎﻫﺎﺕ ﻣﺘﻌﺪﺩﺓ ﻟﻠﺤﺮﻛﺔ ﻭﻣﺴﺘﻮﻳﺎﺕ ﻣﺨﺘﻠﻔﺔ ﻟﺘﺄﺛﻴﺮ ﺍﻟﺈﺷﺎﺭﺓ ﺍﻟﺒﺼﺮﻳﺔ. ﺣﻘﻖ ﻧﻤﻮﺫﺝ ﺍﻟﺸﺒﻜﺔ ﺍﻟﻌﺼﺒﻴﺔ ﻣﻦ ﻧﻮﻉ U-Net ، ﻣﺘﻮﺳﻄﺎ ﻗﺪﺭﻩ %94.5 ﻓﻲ ﻣﻘﻴﺎﺱ ﺍﻟﺘﻘﺎﻃﻊ ﺇﻟﻰ ﺍﻟﺎﺗﺤﺎﺩ ﻭﻣﺘﻮﺳﻄﺎ ﻗﺪﺭﻩ %97.2 ﻓﻲ ﻣﻌﺎﻣﻞ ﺩﺍﻳﺲ. ﻭﺑﻠﻎ ﺯﻣﻦ ﻣﻌﺎﻟﺠﺔ ﺍﻟﺈﻃﺎﺭ ﺍﻟﻮﺍﺣﺪ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﺈﻋﺪﺍﺩ ﺍﻟﻤﺨﺘﺎﺭ 340.23 ﻣﻠﻲ ﺛﺎﻧﻴﺔ. ﻭﺃﻇﻬﺮﺕ ﺗﺠﺎﺭﺏ ﺍﻟﺤﻠﻘﺔ ﺍﻟﻤﻐﻠﻘﺔ ﺃﻥ ﺃﺛﺮ ﺍﻟﺘﻐﺬﻳﺔ ﺍﻟﺮﺍﺟﻌﺔ ﺍﻟﺒﺼﺮﻳﺔ ﻳﺨﺘﻠﻒ ﺎﺧﺘ ﺎﻑ ﺇﻋﺪﺍﺩ ﺍﻟﻤﺘﺤﻜﻢ ﻭﺍﺗﺠﺎﻩ ﺍﻟﺤﺮﻛﺔ. ﻭﻋﻨﺪ ﺍﺳﺘﺨﺪﺍﻡ ﺃﻓﻀﻞ ﺇﻋﺪﺍﺩ ﻋﺎﻡ، ﺍﻧﺨﻔﺾ ﺍﻟﺠﺬﺭ ﺍﻟﺘﺮﺑﻴﻌﻲ ﻟﻤﺘﻮﺳﻂ ﻣﺮﺑﻌﺎﺕ ﺧﻄﺄ ﺍﻟﺘﺘﺒﻊ، ﺑﻌﺪ ﺟﻤﻊ ﻧﺘﺎﺋﺞ ﺍﻟﺎﺗﺠﺎﻫﻴﻦ، ﺑﻨﺴﺒﺔ %43.2. ﻛﻤﺎ ﺃﻇﻬﺮﺕ ﺩﺭﺍﺳﺔ ﺗﺄﺛﻴﺮ ﺍﻟﺈﺷﺎﺭﺓ ﺍﻟﺒﺼﺮﻳﺔ ﺃﻥ ﺯﻳﺎﺩﺓ ﻣﺴﺎﻫﻤﺘﻬﺎ ﻟﺎ ﺗﺤﻘﻖ ﺑﺎﻟﻀﺮﻭﺭﺓ ﺗﺤﺴﻨًﺎ ﺇﺿﺎﻓﻴًﺎ، ﻭﺃﻥ ﺍﻟﺈﻋﺪﺍﺩ ﺍﻟﻘﻴﺎﺳﻲ ﺣﻘﻖ ﺃﻓﻀﻞ ﺃﺩﺍﺀ ﺇﺟﻤﺎﻟﻲ ﺗﺸﻴﺮ ﺍﻟﻨﺘﺎﺋﺞ ﺇﻟﻰ ﺃﻥ ﺩﻣﺞ ﺍﻟﻤﻌﻠﻮﻣﺎﺕ ﺍﻟﺒﺼﺮﻳﺔ ﻓﻲ ﺣﻠﻘﺔ ﺍﻟﺘﺤﻜﻢ ﻳﻤﻜﻦ ﺃﻥ ﻳﺤﺴﻦ ﺩﻗﺔ ﺍﻟﺘﺘﺒﻊ ﻋﻨﺪ ﺿﺒ ﻂﺍﻟﻨﻈﺎﻡ ﺑﺼﻮﺭﺓ ﻣﻨﺎﺳﺒﺔ. ﺇﻟﺎ ﺃﻥ ﻣﻘﺪﺍﺭ ﺍﻟﺘﺤﺴﻦ ﻳﺘﺄﺛﺮ ﺑﺈﻋﺪﺍﺩﺍﺕ ﺍﻟﻤﺘﺤﻜﻢ، ﻭﺍﺗﺠﺎﻩ ﺍﻟﺤﺮﻛﺔ، ﻭﻣﺴﺎﻫﻤﺔ ﺍﻟﺈﺷﺎﺭﺓ ﺍﻟﺒﺼﺮﻳﺔ، ﻭﺯﻣﻦ ﺍﻟﻤﻌﺎﻟﺠﺔ، ﻭﺩﻗﺔ ﺍﻟﻤﻌﺎﻳﺮﺓ. ﻭﺗﺒﺮﺯ ﻫﺬﻩ ﺍﻟﻨﺘﺎﺋﺞ ﺃﻫﻤﻴﺔ ﺗﻘﻠﻴﻞ ﺯﻣﻦ ﺍﻟﻤﻌﺎﻟﺠﺔ ﻋﻠﻰ ﻣﺘﻦ ﺍﻟﻤﺮﻛﺒﺔ .ﻭﺍﺗﺒﺎﻉ ﻣﻨﻬﺞ ﻣﻨﻈﻢ ﻟﺎﺧﺘﻴﺎﺭ ﻌﺎ ﺎﺕ ﺍﻟﺘﺤﻜﻢ، ﺑﻤﺎ ﻳﺤﻘﻖ ﺗﺼﺤﻴﺤًﺎ ﺃﺩﻕ ﻭﺃﻛﺜﺮ ﻣﻮﺛﻮﻗﻴﺔ ﻟﻠﺤﺮﻛﺔ

English Abstract

This thesis develops a vision-based method for improving trajectory-tracking accuracy. The method detects a projected laser marker and a reference line using image processing and semantic segmentation, then calculates the deviation from the desired path. The calculated visual error is incorporated into the trajectory controller as an additional feedback signal. The vision system was first implemented and evaluated through segmentation and processing-performance tests. It was then deployed onboard for online operation and experimentally evaluated under different controller settings, directions of motion, and visual-feedback contributions. The trained U-Net achieved a mean foreground intersection-over-union of 94.5% and a mean foreground Dice score of 97.2%. The selected configuration processed each frame in 340.23 ms. During the closed-loop experiments, the effectiveness of visual feedback depended on the controller setting and direction of motion. Under the best overall controller setting, visual feedback reduced the combined root-mean-square tracking error across both directions by 43.2%. The feedback-weight study further showed that changing the contribution of the visual error affected tracking performance, but the standard visual-feedback configuration remained the most effective overall. The results demonstrate that onboard visual feedback can improve trajectory tracking accuracy when properly integrated and tuned. However, its effectiveness is influenced by controller settings, direction of motion, feedback contribution, processing delay, and visual calibration. The findings highlight the importance of faster onboard processing and systematic parameter selection for achieving more reliable trajectory correction.

Item Type:	Thesis (Masters)
Subjects:	Systems Aerospace
Department:	College of Engineering and Physics > Aerospace Engineering
Thesis Advisor:	Hassan Abid,
Thesis Committee Members:	Syed Saad Ali, Mohamed Ismail,
Depositing User:	RUAA NAKKAR
Date Deposited:	22 Jul 2026 05:36
Last Modified:	22 Jul 2026 05:36
URI:	https://eprints.kfupm.edu.sa/id/eprint/144670