Smart Drone-based Window Maintenance: Fusing PIR Detection, YOLOv8 Object Recognition, and Bi-LSTM Prediction

This research presents an intelligent window cleaning drone system designed to adapt to dynamic environmental conditions using a combination of Passive Infrared (PIR) sensing, YOLOv8 object detection, and a Bidirectional Long Short-Term Memory (Bi-LSTM) model. The system leverages PIR sensors for real-time human presence detection, enhancing safety and operational decision-making. YOLOv8 is employed to identify and localize obstacles, such as birds or protruding architectural elements, allowing the drone to navigate complex urban environments with high precision. Furthermore, the Bi-LSTM network analyzes temporal patterns in sensor data to predict environmental changes and optimize flight paths and cleaning routines. By integrating these technologies, the proposed drone system achieves efficient, safe, and adaptive window cleaning in both residential and commercial settings. Experimental results demonstrate the system’s ability to operate autonomously, avoid collisions, and adapt cleaning strategies based on surrounding conditions, making it a promising solution for smart urban infrastructure maintenance.