Scholarlink

This paper offers a comprehensive exploration of aerial computing, with a primary focus on the integration of tiny machine learning (TinyML) and federated reinforcement learning (FRL) for low-altitude unmanned aerial vehicles (LAUs) trajectory optimization. Aerial computing, an emerging paradigm in computing infrastructure, combines space-air-ground integrated networks (SAGINs), edge computing, and advanced aerial platforms. TinyML, at the intersection of machine learning and embedded systems, addresses challenges related to resource-constrained LAUs. FRL extends reinforcement learning principles to encourage collaboration among LAUs while ensuring agent privacy. This study highlights the application of TinyML and FRL in trajectory planning and data offloading strategies for LAUs. Furthermore, a case study presents the successful integration of TinyML and FRL in LAUs for efficient data collection. Comparative analysis indicates a significant reduction in energy consumption when employing TinyML and FRL.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Potential of TinyML and Federated Reinforcement Learning-Based Trajectory Optimization in Space-Air-Ground Integrated Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shahnila Rahim,
  • Salman Khalil

摘要

This paper offers a comprehensive exploration of aerial computing, with a primary focus on the integration of tiny machine learning (TinyML) and federated reinforcement learning (FRL) for low-altitude unmanned aerial vehicles (LAUs) trajectory optimization. Aerial computing, an emerging paradigm in computing infrastructure, combines space-air-ground integrated networks (SAGINs), edge computing, and advanced aerial platforms. TinyML, at the intersection of machine learning and embedded systems, addresses challenges related to resource-constrained LAUs. FRL extends reinforcement learning principles to encourage collaboration among LAUs while ensuring agent privacy. This study highlights the application of TinyML and FRL in trajectory planning and data offloading strategies for LAUs. Furthermore, a case study presents the successful integration of TinyML and FRL in LAUs for efficient data collection. Comparative analysis indicates a significant reduction in energy consumption when employing TinyML and FRL.