Improved LSTM and Kalman Filter Fusion Method for Asymmetric Network Link Delay Compensation

Delay issues are prevalent in modern communication, control, and signal processing systems, posing significant challenges to real-time performance, stability, and security. Traditional delay compensation methods are typically model-based, which often perform poorly in complex nonlinear systems, while purely data-driven methods may lack sufficient robustness. This research proposes an innovative delay compensation method combining Long Short-Term Memory (LSTM) networks and Kalman filtering, specifically designed for asymmetric network link delay characteristics. The method develops an adaptive fusion mechanism and feedforward compensation strategy. LSTM captures nonlinear dynamic characteristics of the system by learning from historical data and accurately predicts future delays, while Kalman filtering performs online correction using LSTM's prediction results, thereby enhancing system robustness and anti-interference capability. Theoretical analysis and simulation experiments demonstrate that the proposed method significantly outperforms traditional methods in position tracking accuracy, system stability, and control energy consumption, with RMSE reduced by approximately 40%, state fluctuation decreased by 47%, and control energy lowered by 19%, providing an efficient and reliable solution for delay compensation in networked control systems.