Autonomous Nano Drones for Suspicious Activity Detection and Tracking of Doubtful Individuals

Abstract

This paper proposes a novel framework that leverages autonomous nano drones equipped with advanced AI capabilities for real-time detection and tracking of suspicious individuals in dynamic environments. The system integrates lightweight object detection using YOLOv7, temporal behavior analysis via Long Short-Term Memory (LSTM) networks, and swarm-based coordination driven by Particle Swarm Optimization (PSO). Drones collaboratively monitor and analyze human activity, while onboard edge processing ensures low-latency decision-making without reliance on centralized computation. Kalman filters are employed for accurate and continuous target tracking, and a secure mesh communication protocol facilitates real-time alert generation to the control center. Experimental evaluation demonstrates superior performance of the proposed system over traditional surveillance approaches, achieving higher detection accuracy (92.3%), improved activity classification (89.5%), and reduced latency (45 ms). The results affirm the effectiveness and scalability of autonomous nano drone swarms for intelligent surveillance applications in smart cities, critical infrastructure, and defense operations.