無線電地圖可提供資訊並應用於各領域，因此建立精準度高的地圖是首要目標。然而在實際環境中存在著各種干擾，使得準確預測成為一項具有挑戰性的任務。機器學習是一個的強大工具，可應用於地圖重建。不過隨著物聯網的普及，大眾逐漸開始意識資訊安全的問題，為了解決機器學習涉及的隱私問題，人們提出了聯邦學習演算法。除了隱私問題，在物聯網發展同時，大量數據生成以及運算耗能增加，則是引起環保議題上的關注。因此本論文以階層式聯邦學習算法，搭配二次形式或多層感知器重建無線電地圖。模擬結果顯示，我們提出的算法在保有用戶隱私的同時，成功降低其複雜度，且地圖的預測準確度達到相同的效能水準。;A radio map provides valuable information and has applications in various fields. Creating highly accurate maps is a primary objective, but practical environments introduce various interferences, posing a challenging task for accurate prediction. Machine learning is a powerful tool for reconstructing map models. However, the rapid development of the Internet of Things has brought attention to information privacy issues. To address these concerns, federated learning (FL) algorithms have been proposed. In addition to privacy concerns, the generation of significant amounts of data and the increasing energy consumption during development have raised environmental concerns. Therefore, this paper employs hierarchical federated learning (HFL) algorithms with a quadratic form or multilayer perceptron to reconstruct the radio map. The simulation results demonstrate that our proposed algorithm effectively reduces client-side complexity while maintaining privacy. Additionally, the accuracy of the map also achieves the same level of performance.
