A Novel Methodology for Video Transmission Latency Control via Frame Extrapolation: An Experimental Approach

Minimizing Glass-to-Glass (G2G) delay in real-time video systems is crucial for safety-critical applications such as remote operation, robotics, and immersive communication. This thesis introduces a modular simulation framework for latency-aware video transmission, integrating predictive bitrate control and frame extrapolation to ensure high visual quality under strict delay constraints. At the core of the system is a Model Predictive Control (MPC) module guided by a Rate-Distortion-Horizon (RDH) model, which estimates the trade-offs between bitrate, extrapolation horizon, and perceptual quality. We propose and evaluate multiple parametric RDH formulations—polynomial, logarithmic, and logarithmic-quadratic—and assess their performance both analytically and experimentally. Two additional contributions are presented: (1) a personalized RDH fitting strategy based on the initial segment of each video, and (2) a comparative study of RDH models under varying G2G delay budgets. Results show that the MPC controller improves quality on familiar content by dynamically optimizing bitrate and extrapolation horizon. Personalized models offer marginal gains on predictable sequences, while the logarithmic-quadratic RDH variant generalizes best across different scenarios. Overall, the proposed system demonstrates the potential of model-based bitrate control in achieving ultra-low latency video transmission and offers a flexible research platform for future advances in real-time multimedia systems.