LightTR: A Lightweight Framework for Federated Trajectory Recovery Ziqiao Liu, Hao Miao, Yan Zhao, Chenxi Liu, Kai Zheng, Huan Li Abstract—With the proliferation of GPS-equipped edge devices, huge trajectory data is generated and accumulated in various domains, motivating a variety of urban applications. Due to the limited acquisition capabilities of edge devices, a lot of trajectories are recorded at a low sampling rate, which may lead to the effectiveness drop of urban applications. We aim to recover a high-sampled trajectory based on the low-sampled trajectory in free space, i.e., without road network information, to enhance the usability of trajectory data and support urban applications more effectively. Recent proposals targeting trajectory recovery often assume that trajectories are available at a central location, which fail to handle the decentralized trajectories and hurt privacy. To bridge the gap between decentralized training and trajectory recovery, we propose a lightweight framework, LightTR, for federated trajectory recovery based on a client-server architecture, while keeping the data decentralized and private in each client/platform center (e.g., each data center of a company). Specifically, considering the limited processing capabilities of edge devices, LightTR encompasses a light local trajectory embedding module that offers improved computational efficiency without compromising its feature extraction capabilities. LightTR also features a meta-knowledge enhanced local-global training scheme to reduce communication costs between the server and clients and thus further offer efficiency improvement. Extensive experiments demonstrate the effectiveness and efficiency of the proposed framework. Index Terms—Trajectory Recovery; Lightweight; Federated Learning; LightTR is a lightweight framework for federated trajectory recovery that addresses the challenges of scalability and communication cost in decentralized trajectory processing. The framework is designed to recover high-sampled trajectories from low-sampled trajectories in free space, without relying on road network information. LightTR is based on a client-server architecture and includes two major modules: a local trajectory preprocessing and light embedding module, and a meta-knowledge enhanced local-global training module. The local trajectory preprocessing and light embedding module is designed to capture effective spatio-temporal correlations of trajectories with a customized lightweight trajectory embedding ST-operator. The meta-knowledge enhanced local-global training module is designed to reduce communication cost and speed up model convergence by means of knowledge distillation. The framework is evaluated on two real-world datasets, Tdrive and Geolife, and shows significant improvements in performance and efficiency compared to existing methods. The results demonstrate that LightTR achieves the best results among all the baselines on the two datasets with different settings of keep ratio. LightTR performs better than the best among the baselines by up to 14.7% and 13.19% with keep ratio=12.5% for Recall and Precision, respectively, while obtaining MAE and RMSE reduction by at most 22.3% and 29.2%, on Geolife. The framework is also shown to be more efficient in terms of running time and computationalLightTR: A Lightweight Framework for Federated Trajectory Recovery Ziqiao Liu, Hao Miao, Yan Zhao, Chenxi Liu, Kai Zheng, Huan Li Abstract—With the proliferation of GPS-equipped edge devices, huge trajectory data is generated and accumulated in various domains, motivating a variety of urban applications. Due to the limited acquisition capabilities of edge devices, a lot of trajectories are recorded at a low sampling rate, which may lead to the effectiveness drop of urban applications. We aim to recover a high-sampled trajectory based on the low-sampled trajectory in free space, i.e., without road network information, to enhance the usability of trajectory data and support urban applications more effectively. Recent proposals targeting trajectory recovery often assume that trajectories are available at a central location, which fail to handle the decentralized trajectories and hurt privacy. To bridge the gap between decentralized training and trajectory recovery, we propose a lightweight framework, LightTR, for federated trajectory recovery based on a client-server architecture, while keeping the data decentralized and private in each client/platform center (e.g., each data center of a company). Specifically, considering the limited processing capabilities of edge devices, LightTR encompasses a light local trajectory embedding module that offers improved computational efficiency without compromising its feature extraction capabilities. LightTR also features a meta-knowledge enhanced local-global training scheme to reduce communication costs between the server and clients and thus further offer efficiency improvement. Extensive experiments demonstrate the effectiveness and efficiency of the proposed framework. Index Terms—Trajectory Recovery; Lightweight; Federated Learning; LightTR is a lightweight framework for federated trajectory recovery that addresses the challenges of scalability and communication cost in decentralized trajectory processing. The framework is designed to recover high-sampled trajectories from low-sampled trajectories in free space, without relying on road network information. LightTR is based on a client-server architecture and includes two major modules: a local trajectory preprocessing and light embedding module, and a meta-knowledge enhanced local-global training module. The local trajectory preprocessing and light embedding module is designed to capture effective spatio-temporal correlations of trajectories with a customized lightweight trajectory embedding ST-operator. The meta-knowledge enhanced local-global training module is designed to reduce communication cost and speed up model convergence by means of knowledge distillation. The framework is evaluated on two real-world datasets, Tdrive and Geolife, and shows significant improvements in performance and efficiency compared to existing methods. The results demonstrate that LightTR achieves the best results among all the baselines on the two datasets with different settings of keep ratio. LightTR performs better than the best among the baselines by up to 14.7% and 13.19% with keep ratio=12.5% for Recall and Precision, respectively, while obtaining MAE and RMSE reduction by at most 22.3% and 29.2%, on Geolife. The framework is also shown to be more efficient in terms of running time and computational