Sequence-to-Sequence Predictive Control for Telesurgery Under Network Fluctuations in the Internet of Medical Things

The Internet of Medical Things (IoMT) is transforming healthcare by enabling remote surgical procedures that deliver critical care to patients in underserved and geographically isolated regions. Among these, telesurgery plays a pivotal role, while its reliability remains vulnerable to network fluctuations such as latency, jitter, and packet loss, which can degrade teleoperation performance and increase surgical risk. To address this challenge, we propose a Teleoperation Dynamic Predictive Control (TDPC) strategy to predict future positions of surgical instruments and mitigate uncertainties introduced by unstable IoMT networks. The strategy integrates a deep time series forecasting model with adaptive delay compensation, packet loss monitoring, and Kalman based correction to enhance robustness under adverse conditions. A high-fidelity teleoperation system was built within the Robot Operating System 2 to benchmark TDPC against state-of-the-art approaches. Furthermore, we conducted comprehensive evaluations under dynamic delay, stochastic packet loss, and different predictive models using a real-time surgical communication protocol. Finally, practical efficiency of our strategy in teleoperation was demonstrated through a phantom experiment. Experimental results show that TDPC significantly outperforms state-of-the-art models in delay compensation and demonstrates exceptional robustness to packet loss. The strategy improves both operational efficiency and procedural safety during teleoperation tasks, as evidenced by reduced task completion time and minimized execution errors. These results collectively indicate that TDPC offers substantial potential for advancing telesurgery within IoMT-driven healthcare systems, enabling safer and more reliable remote surgical interventions.