Live streaming is evolving from simple passive distribution to highly interactive scenarios (video conferencing and cloud gaming), where end-to-end latency becomes the main constraint on quality of experience. This thesis analyzes recent techniques for low-latency streaming across the entire chain: encoding, packaging, and transport. On the encoding side, we discuss the tools introduced by the HEVC/VVC standards to increase parallelism and granularity, and solutions for stabilizing bitrate, in particular Gradual Decoder Refresh (GDR), which mitigates peaks caused by intra-coded frames. The packaging phase is analyzed using CMAF and Chunked Transfer Encoding, highlighting the impact on adaptation mechanisms (ABR) and the need for different bandwidth estimates. For distribution, LL-DASH and LL-HLS are explored in depth, which exploit partial segments and preloading, as well as advances in transport with QUIC and WebRTC. Finally, emerging approaches such as HESP and MoQ are examined, which aim to combine interactivity, low latency, and scalability. The analysis shows that low latency does not depend on a single mechanism: it is the result of coordinated choices between encoding, packaging, and protocols, with explicit trade-offs between quality and stability.
Lo streaming live sta evolvendo da semplice distribuzione passiva a scenari fortemente interattivi (videoconferenza e cloud gaming), in cui la latenza end-to-end diventa il principale vincolo di qualità dell’esperienza. Questa tesi analizza le tecniche recenti per lo low-latency streaming lungo l’intera catena: codifica, packaging e trasporto. Dal lato della codifica, vengono discussi gli strumenti introdotti dagli standard HEVC/VVC per aumentare parallelismo e granularità e le soluzioni per stabilizzare il bitrate; in particolare il Gradual Decoder Refresh (GDR) che attenua i picchi dovuti ai frame codificati con tecnica intra. La fase di packaging è analizzata tramite CMAF e Chunked Transfer Encoding, evidenziando l’impatto sui meccanismi di adattamento (ABR) e la necessità di stime della banda diverse. Per la distribuzione, si approfondiscono LL-DASH e LL-HLS, che sfruttano segmenti parziali e preload, nonché i progressi del trasporto con QUIC e WebRTC. Infine, vengono esaminati approcci emergenti come HESP e MoQ, che mirano a coniugare interattività, bassa latenza e scalabilità. Dall’analisi emerge che la bassa latenza non dipende da un singolo meccanismo: è il risultato di scelte coordinate tra codifica, packaging e protocolli, con trade-off espliciti su qualità e stabilità.
Tecniche recenti per lo streaming a bassa latenza
GRANATI, ALBERTO
2024/2025
Abstract
Live streaming is evolving from simple passive distribution to highly interactive scenarios (video conferencing and cloud gaming), where end-to-end latency becomes the main constraint on quality of experience. This thesis analyzes recent techniques for low-latency streaming across the entire chain: encoding, packaging, and transport. On the encoding side, we discuss the tools introduced by the HEVC/VVC standards to increase parallelism and granularity, and solutions for stabilizing bitrate, in particular Gradual Decoder Refresh (GDR), which mitigates peaks caused by intra-coded frames. The packaging phase is analyzed using CMAF and Chunked Transfer Encoding, highlighting the impact on adaptation mechanisms (ABR) and the need for different bandwidth estimates. For distribution, LL-DASH and LL-HLS are explored in depth, which exploit partial segments and preloading, as well as advances in transport with QUIC and WebRTC. Finally, emerging approaches such as HESP and MoQ are examined, which aim to combine interactivity, low latency, and scalability. The analysis shows that low latency does not depend on a single mechanism: it is the result of coordinated choices between encoding, packaging, and protocols, with explicit trade-offs between quality and stability.| File | Dimensione | Formato | |
|---|---|---|---|
|
Granati_Alberto.pdf
accesso aperto
Dimensione
977.05 kB
Formato
Adobe PDF
|
977.05 kB | 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/20.500.12608/92198