An Intelligent Drone-Based Framework for Securing Petroleum Zones Using Lightweight Learning

Abstract

The increasing complexity and risk associated with petroleum operations in remote desert environments, suchastheAlgerianSahara, demandadvancedtechnologicalsolutionstoenhance safety and operational efficiency. Traditional monitoring and inspection methods rely heavily on human intervention, exposing personnel to hazardous conditions and limiting the ability to perform continuous and real-time risk assessment. This thesis proposes the design and implementation of an intelligent autonomous drone based system for real-time risk assessment using Agentic Artificial Intelligence (AI). The system integrates an unmanned aerial vehicle (UAV)equippedwithacameraandanonboardprocessing unit based on Raspberry Pi, enabling edge computing and real-time data analysis. The proposed approach leverages computer vision techniques to detect potential risks, in cluding the presence of humans, animals, and vehicles, either through direct visual identifi cation or indirect environmental cues such as traces and movement patterns. Furthermore, the system incorporates agentic AI capabilities to enable autonomous navigation, adaptive decision making, and dynamic mission execution without continuous human supervision. According to our knowledge of current research, this work introduces a novel approach for petroleum operations in desert environments, where intelligent aerial systems are utilized not only for monitoring but also for proactive and autonomous risk interpretation. The implemented system demonstrates the feasibility of combining autonomous drones with intelligent decision-making to improve industrial safety. Experimental results validate the ef fectiveness of the proposed solution in detecting risks and enhancing monitoring processes in harsh and inaccessible environments. This work contributes to the development of scalable and intelligent risk assessment systems for petroleum industries operating in extreme conditions.ﻳﺸﻬﺪ ﺗﻨﻔﻴﺬ ﺍﻟﻌﻤﻠﻴﺎﺕ ﺍﻟﺒﺘﺮﻭﻟﻴﺔ ﻓﻲ ﺍﻟﺒﻴﺌﺎﺕ ﺍﻟﺼﺤﺮﺍﻭﻳﺔ ﺍﻟﻨﺎﺋﻴﺔ، ﻣﺜﻞ ﺍﻟﺼﺤﺮﺍء ﺍﻟﺠﺰﺍﺋﺮﻳﺔ، ﺗﺰﺍﻳﺪًﺍ ﻣﻠﺤﻮﻇ ً ﺎ ﻓﻲ ﺍﻟﺘﻌﻘﻴﺪ ﻭﺍﻟﻤﺨﺎﻃﺮ، ﻣﺎ ﻳﺴﺘﻠﺰﻡ ﺍﻋﺘﻤﺎﺩ ﺣﻠﻮﻝ ﺗﻜﻨﻮﻟﻮﺟﻴﺔ ﻣﺘﻘﺪﻣﺔ ﻟﺘﻌﺰﻳﺰ ﺍﻟﺴﻼﻣﺔ ﻭﺍﻟﻜﻔﺎءﺓ ﺍﻟﺘﺸﻐﻴﻠﻴﺔ. ﻭﺗﻌﺘﻤﺪ ﻃﺮﻕ ﺍﻟﻤﺮﺍﻗﺒﺔ ﻭﺍﻟﺘﻔﺘﻴﺶ ﺍﻟﺘﻘﻠﻴﺪﻳﺔ ﺑﺸﻜﻞ ﻛﺒﻴﺮ ﻋﻠﻰ ﺍﻟﺘﺪﺧﻞ ﺍﻟﺒﺸﺮﻱ، ﻣﻤﺎ ﻳﻌﺮﺽ ﺍﻟﻌﺎﻣﻠﻴﻦ ﻟﻈﺮﻭﻑ ﺧﻄﺮﺓ ﻭﻳﺤﺪ ﻣﻦ ﺍﻟﻘﺪﺭﺓ ﻋﻠﻰ ﺇﺟﺮﺍء ﺗﻘﻴﻴﻢ ﻣﺴﺘﻤﺮ ﻭﻓﻮﺭﻱ ﻟﻠﻤﺨﺎﻃﺮ. ﺗﻬﺪﻑ ﻫﺬﻩ ﺍﻟﺪﺭﺍﺳﺔ ﺇﻟﻰ ﺗﺼﻤﻴﻢ ﻭﺗﻨﻔﻴﺬ ﻧﻈﺎﻡ ﺫﻛﻲ ﻗﺎﺋﻢ ﻋﻠﻰ ﺍﻟﻄﺎﺋﺮﺍﺕ ﺑﺪﻭﻥ ﻃﻴﺎﺭ ﺍﻟﺬﺍﺗﻴﺔ ﻹﺟﺮﺍء ﺗﻘﻴﻴﻢ ﺍﻟﻤﺨﺎﻃﺮ ﻓﻲ ﺍﻟﻮﻗﺖ ﺍﻟﺤﻘﻴﻘﻲ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺍﻟﺬﻛﺎء ﺍﻻﺻﻄﻨﺎﻋﻲ ﺍﻟﻮﻛﻴﻞ Agentic AI ﻳﺪﻣﺞ ﺍﻟﻨﻈﺎﻡ ﻃﺎﺋﺮﺓ ً ﺑﺪﻭﻥ ﻃﻴﺎﺭ ﻣﺰﻭﺩﺓ ﺑﻜﺎﻣﻴﺮﺍ ﻭﻭﺣﺪﺓ ﻣﻌﺎﻟﺠﺔ ﻣﺪﻣﺠﺔ ﺗﻌﺘﻤﺪ ﻋﻠﻰ Raspberry Pi، ﻣﺎ ﻳﻤﻜ ّﻦ ﻣﻦ ﺇﺟﺮﺍء ﺍﻟﺤﻮﺳﺒﺔ ﺍﻟﻄﺮﻓﻴﺔ Edge Computing ﻭﺗﺤﻠﻴﻞ ﺍﻟﺒﻴﺎﻧﺎﺕ ﺑﺸﻜﻞ ﻓﻮﺭﻱ. ﺗﻌﺘﻤﺪ ﺍﻟﻤﻨﻬﺠﻴﺔ ﺍﻟﻤﻘﺘﺮﺣﺔ ﻋﻠﻰ ﺗﻘﻨﻴﺎﺕ ﺍﻟﺮﺅﻳﺔ ﺍﻟﺤﺎﺳﻮﺑﻴﺔ ﻻﻛﺘﺸﺎﻑ ﺍﻟﻤﺨﺎﻃﺮ ﺍﻟﻤﺤﺘﻤﻠﺔ، ﺑﻤﺎ ﻓﻲ ﺫﻟﻚ ﻭﺟﻮﺩ ﺍﻟﺒﺸﺮ ﻭﺍﻟﺤﻴﻮﺍﻧﺎﺕ ﻭﺍﻟﻤﺮﻛﺒﺎﺕ، ﺳﻮﺍءً ﻋﺒﺮ ﺍﻟﺘﻌﺮﻑ ﺍﻟﺒﺼﺮﻱ ﺍﻟﻤﺒﺎﺷﺮ ﺃﻭ ﻣﻦ ﺧﻼﻝ ﻣﺆﺷﺮﺍﺕ ﺑﻴﺌﻴﺔ ﻏﻴﺮ ﻣﺒﺎﺷﺮﺓ ﻣﺜﻞ ﺍﻷﺛﺮ ﻭﺃﻧﻤﺎﻁ ﺍﻟﺤﺮﻛﺔ. ﻛﻤﺎ ﻳﺘﻀﻤﻦ ﺍﻟﻨﻈﺎﻡ ﻗﺪﺭﺍﺕ ﺍﻟﺬﻛﺎء ﺍﻻﺻﻄﻨﺎﻋﻲ ﺍﻟﻮﻛﻴﻞ ﻟﺘﻤﻜﻴﻦ ﺍﻟﻤﻼﺣﺔ ﺍﻟﺬﺍﺗﻴﺔ، ﻭﺍﺗﺨﺎﺫ ﺍﻟﻘﺮﺍﺭﺍﺕ ﺍﻟﺘﻜﻴﻔﻴﺔ، ﻭﺗﻨﻔﻴﺬ ﺍﻟﻤﻬﺎﻡ ﺍﻟﺪﻳﻨﺎﻣﻴﻜﻲ ﺩﻭﻥ ﺍﻟﺤﺎﺟﺔ ﺇﻟﻰ ﺇﺷﺮﺍﻑ ﺑﺸﺮﻱ ﻣﺴﺘﻤﺮ. ﻭﻳُﻌﺪ ﻫﺬﺍ ﺍﻟﻌﻤﻞ، ﺑﺤﺴﺐ ﻣﻌﺮﻓﺘﻨﺎ ﺑﺎﻷﺑﺤﺎﺙ ﺍﻟﺤﺎﻟﻴﺔ، ﻣﻘﺎﺭﺑﺔ ﺟﺪﻳﺪﺓ ﻟﻠﻌﻤﻠﻴﺎﺕ ﺍﻟﺒﺘﺮﻭﻟﻴﺔ ﻓﻲ ﺍﻟﺒﻴﺌﺎﺕ ﺍﻟﺼﺤﺮﺍﻭﻳﺔ، ﺣﻴﺚ ﺗ ُﻮﻇ ﱠﻒ ﺍﻷﻧﻈﻤﺔ ﺍﻟﺠﻮﻳﺔ ﺍﻟﺬﻛﻴﺔ ﻟﻴﺲ ﻓﻘﻂ ﻟﻠﻤﺮﺍﻗﺒﺔ، ﺑﻞ ﻟﺘﻔﺴﻴﺮ ﺍﻟﻤﺨﺎﻃﺮ ﺑﺸﻜﻞ ﺍﺳﺘﺒﺎﻗﻲ ﻭﺫﺍﺗﻲ. ﻭﺗﻮﺿﺢ ﺍﻟﻨﺘﺎﺋﺞ ﺍﻟﺘﺠﺮﻳﺒﻴﺔ ﻟﻠﻨﻈﺎﻡ ﺍﻟﻤُﻨﻔ ﱠﺬ ﺟﺪﻭﻯ ﺩﻣﺞ ﺍﻟﻄﺎﺋﺮﺍﺕ ﺍﻟﺬﺍﺗﻴﺔ ﻣﻊ ﺃﻧﻈﻤﺔ ﺍﺗﺨﺎﺫ ﺍﻟﻘﺮﺍﺭ ﺍﻟﺬﻛﻲ ﻟﺘﺤﺴﻴﻦ ﺍﻟﺴﻼﻣﺔ ﺍﻟﺼﻨﺎﻋﻴﺔ. ﻛﻤﺎ ﺗﺆﻛﺪ ﻫﺬﻩ ﺍﻟﻨﺘﺎﺋﺞ ﻓﻌﺎﻟﻴﺔ ﺍﻟﺤﻞ ﺍﻟﻤﻘﺘﺮﺡ ﻓﻲ ﺍﻛﺘﺸﺎﻑ ﺍﻟﻤﺨﺎﻃﺮ ﻭﺗﻌﺰﻳﺰ ﻋﻤﻠﻴﺎﺕ ﺍﻟﻤﺮﺍﻗﺒﺔ ﻓﻲ ﺍﻟﺒﻴﺌﺎﺕ ﺍﻟﻘﺎﺳﻴﺔ ﻭﺍﻟﺘﻲ ﻳﺼﻌﺐ ﺍﻟﻮﺻﻮﻝ ﺇﻟﻴﻬﺎ. ﻭﻳُﺴﻬﻢ ﻫﺬﺍ ﺍﻟﻌﻤﻞ ﻓﻲ ﺗﻄﻮﻳﺮ ﺃﻧﻈﻤﺔ ﺗﻘﻴﻴﻢ ﺍﻟﻤﺨﺎﻃﺮ ﺍﻟﺬﻛﻴﺔ ﻭﺍﻟﻘﺎﺑﻠﺔ ﻟﻠﺘﻮﺳﻊ ﻟﻠﺼﻨﺎﻋﺎﺕ ﺍﻟﺒﺘﺮﻭﻟﻴﺔ ﺍﻟﻌﺎﻣﻠﺔ ﻓﻲ ﻇﺮﻭﻑ ﻗﺎﺳﻴﺔ.