Intelligent Spatiotemporal Freshness Framework for Multi-UAV Target Detection and Tracking

In the rapidly evolving field of Unmanned Aerial Vehicles (UAVs), these miniaturized platforms are increasingly being designed for intelligent aerial data acquisition, enabling dynamic target detection and tracking to facilitate the deployment of new use cases. This paper presents a novel integrated sensing and communications framework for UAV networks to enhance latency, reliability, and resource allocation efficiency. A novel multi-dimensional freshness metric, the Age of Valid Sensing (AVS), is introduced as a measure of actionable intelligence to quantify and prioritize the sensing data, accurately capturing the quality and relevance of information in dynamic environments, leading to improved UAV coordination and efficient resource allocation. The effectiveness of AVS is strengthened by the presence of Frechet distance, which performs the behavioral analysis of moving targets to enable spatiotemporal clustering based on their trajectory similarity for efficient sensing. Intelligence is being added to each UAV through a robust Multi-Agent Reinforcement Learning (MARL) framework to regularly update their target sensing and communications information, dynamically balancing data freshness and the likelihood of successful information gathering. This approach allows for the efficient integration and processing of sensing data from multiple geographically dispersed targets, significantly improving real-time tracking and decision-making capabilities in complex environments. Our simulation results demonstrate the superior performance of the proposed framework in achieving lower latency, higher detection accuracy, and improved resource efficiency compared to existing methods.