Recent advancements in teleoperation impose increasingly stringent constraints on visual quality and latency, encapsulated in the Glass-to-Glass (G2G) paradigm. Because these requirements are inherently conflicting, state-of-the-art approaches have proposed a Model Predictive Control (MPC) mechanism combined with temporal frame extrapolation at the receiver to enforce a strict upper bound on latency. However, the effectiveness of this architecture heavily relies on the accuracy of channel capacity estimation, which currently employs a simple moving average that is slow to respond to rapid network fluctuations. This thesis integrates the hybrid Weighted Harmonic-Exponential Averaging (WHEA) estimation algorithm into the MPC framework, experimentally evaluating its performance on real-world traffic traces. The objective of this work is to demonstrate how a more accurate and responsive throughput estimation prevents transmission buffer saturation, significantly reducing both the occurrence and magnitude of Glass-to-Glass delay constraint violations (D_G2G), thereby laying the foundation to minimize the need for concealment and extrapolation at the receiver.
Le recenti evoluzioni nell'ambito della teleoperazione impongono vincoli sempre più stringenti sulla qualità visiva e sulla latenza, sintetizzabili nel paradigma Glass-to-Glass (G2G). Trattandosi di requisiti intrinsecamente contrastanti, lo stato dell'arte ha proposto un meccanismo di controllo predittivo (Model Predictive Control, MPC) abbinato all'estrapolazione temporale dei frame al ricevitore per garantire un limite massimo alla latenza. Tuttavia, l'efficacia di tale architettura dipende fortemente dall'accuratezza nella stima della capacità di canale che, allo stato attuale, impiega una semplice media mobile, lenta nel reagire alle rapide fluttuazioni di rete. Questo lavoro di tesi integra nell'architettura MPC l'algoritmo di stima ibrido Weighted Harmonic-Exponential Averaging (WHEA), valutandone sperimentalmente le prestazioni su tracce di traffico reale. Lo scopo del lavoro è dimostrare come una stima di throughput più precisa e reattiva consenta di prevenire la saturazione dei buffer di trasmissione, riducendo significativamente l'occorrenza e l'entità delle violazioni del vincolo di ritardo Glass-to-Glass (D_G2G) e ponendo così le basi per minimizzare il fabbisogno di occultamento e di estrapolazione al ricevitore.
Integrazione e testing dell'algoritmo WHEA per la predizione della capacità di canale in architetture Glass-to-Glass
ROCCATELLO, NICOLÒ
2025/2026
Abstract
Recent advancements in teleoperation impose increasingly stringent constraints on visual quality and latency, encapsulated in the Glass-to-Glass (G2G) paradigm. Because these requirements are inherently conflicting, state-of-the-art approaches have proposed a Model Predictive Control (MPC) mechanism combined with temporal frame extrapolation at the receiver to enforce a strict upper bound on latency. However, the effectiveness of this architecture heavily relies on the accuracy of channel capacity estimation, which currently employs a simple moving average that is slow to respond to rapid network fluctuations. This thesis integrates the hybrid Weighted Harmonic-Exponential Averaging (WHEA) estimation algorithm into the MPC framework, experimentally evaluating its performance on real-world traffic traces. The objective of this work is to demonstrate how a more accurate and responsive throughput estimation prevents transmission buffer saturation, significantly reducing both the occurrence and magnitude of Glass-to-Glass delay constraint violations (D_G2G), thereby laying the foundation to minimize the need for concealment and extrapolation at the receiver.| File | Dimensione | Formato | |
|---|---|---|---|
|
Roccatello_Nicolò.pdf
accesso aperto
Dimensione
2.29 MB
Formato
Adobe PDF
|
2.29 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/114288