This project presents a decentralized, autonomous control algorithm for a multi-drone swarm, designed to optimize interior lighting and achieve maximum illumination. We developed a localized, reaction-based navigation framework in which each drone operates as an independent agent equipped with a light source and a 6-zone omnidirectional light sensor grid. Utilizing a gradient descent mechanism, drones dynamically sense local illuminance differences and continuously navigate away from highly lit regions. This behavior drives the swarm toward uniform spatial dispersion, reducing mutual interference and enabling automatic coverage of darker, under-illuminated areas of the environment. The algorithm explicitly incorporates real-world physical effects, including surface reflections. Tested within a 3D simulation space, the system demonstrates high computational efficiency, real-time convergence, and robust adaptation to irregular room geometries. To evaluate performance, we compared the proposed method against Grid Search Optimization (GSO). Experimental results show that our approach achieves significantly lower runtime while maintaining effective illumination optimization, demonstrating improved scalability and suitability for real-time multi-agent deployment.