Towards Efficient Underwater Robotic Swarms: Edge-Based Comparative Analysis of Multi-Object Trackers

This article conducts a comparative analysis of State-Of- The-Art (SOTA) multi-object trackers, focusing on their performance when tracking homogeneous swarms of robots and their suitability for edge device deployment. The comparative analysis can serve as a valuable tool for assessing the viability of employing different detectors and Multi-Object Tracking (MOT) models in developing autonomous underwater robotic swarms that can be utilized in various applications such as search and rescue, seabed scanning, environmental monitoring, etc. The evaluation centers around two real-time SOTA multi-object trackers, namely ByteTrack and OCSort, both operating on the tracking-by-detection (TBD) paradigm. Additionally, detectors such as $\mathbf{YOLOv5n}, \mathbf{YOLOv8n}$, and YOLOXn are considered. The study employs two distinct MOT datasets, DanceTrack and BrackishMOT, for assessing accuracy metrics such as ROTA, MOTA, and IDF1 as well as latency. The findings offer practical insights into the optimal combination of detectors and MOT models, crucial for enhancing the efficiency and performance of autonomous underwater robotic swarms. This research serves as a valuable reference for researchers and practitioners in the field, providing a foundation for the development of advanced systems capable of addressing diverse challenges in underwater environments.