Back to Search Document details
40th Meeting: Geneva, CH, October 2025 2025-09-27 01:04
AHG18: Request on inclusion ultra-low latency and packet loss resilience category into CfP
Authors: Sergey Ikonin Huawei Xiang Ma Huawei Elena Alshina Huawei S. Wenger Tencent
Abstract
This contribution suggests to include ultra-low latency and packet loss resilience as specific category for future Call-for-Proposals.
Meeting Notes
JVET-AN0204 AHG18: Request on inclusion ultra-low latency and packet loss resilience category into CfP [S. Ikonin, X. Ma, E. Alshina (Huawei), S. Wenger (Tencent)]

See notes in section 7.3.1.

JVET-AN0204 AHG18: Request on inclusion ultra-low latency and packet loss resilience category into CfP [S. Ikonin, X. Ma, E. Alshina (Huawei), S. Wenger (Tencent)]

See section 4.16.3.

JVET-AN0204 AHG18: Request on inclusion ultra-low latency and packet loss resilience category into CfP [S. Ikonin, X. Ma, E. Alshina (Huawei), S. Wenger (Tencent)]

This contribution suggests to include ultra-low latency and packet loss resilience as a specific category for the future Call for Proposals

Considering the requirements mentioned in sections 6 “End-to-end delay” and 9 “Packet loss resilience” of Annex B of the April 2025 JVET and Q6/21 meeting report (TD 142/WP3), also issued as MPEG WG 2/N 448 “Draft of use cases and requirements for potential next-generation video coding standard beyond VVC capability”, and progress of JVET AHG18 work in part of test conditions design, simulation platform preparation, and evidence generation, the proponents suggested to include ultra-low latency and packet loss resilience as a specific category of a Call for Proposals document.

Based on contributions JVET-AL0051 and JVET-AN0079, a marked-up version of JVET-AM2026 “Joint Call for Evidence on video compression with capability beyond VVC” covering ultra-low latency and packet loss resilience was provided, adding the following:

  1. Test cases related to ultra-low latency and packet loss resilience

The subset of categories listed in section 2 are considered as test cases. Namely:

  • Gaming LB HD
  • SDR LB HD
  • SDR RA UHD/4k coded in LDB configuration

Submitters are strongly encouraged (but not required) to submit results for all test cases. However, submitters are required to provide results for all sequences in a given test case.

Test cases

Two main scenarios are considered for evaluation.

  • Unicast scenario. Feedback channel is available with feedback time well above 10ms . The encoder is allowed to adjust the reference picture list based on the feedback.
  • Broadcast scenario. No feedback channel except maybe for statistics; feedback channel delay in the seconds range.
  • Clean channel performance verification. To verify BD-rate overhead of assessed methods the clean channel performance should be reported.

During the unicast and broadcast test the reconstruction quality is evaluated against different latency restrictions, e.g.: 16, 24, 32 and 50ms.

Network channel model

Network channel model emulates the transmission and obtains packet losses and delays based on pre-recorded packet trace from NS-3 network simulator. The trace files are available in AHG18 simulation software and contribution JVET-AN0079.

For every sequence and bitrate the corresponding bandwidth point is specified to approximately match the network channel capacity to the average bitrate of the coded sequence. The bandwidth point consists of packet trace file and start-line are specified, same way as for example in JVET-AN0079.

Reconstruction performance is evaluated in different last mile connection types:

  • Wi-Fi connection. The Wi-Fi connection emulating office or home usage scenario
  • 5G connection. The 5G connection emulating office or home usage scenario using.

Reconstruction performance is evaluated in regular channel model and channel model with transmission priorities.

  • Regular transmission model. Packets of every frame are transmitted one-by-one (without any priority mechanism) and retransmitted during their lifetime. Packet lifetime is a common mechanism used in Wi-Fi and 5G to control retransmission logic. The sending buffer of the router keeps each specific packet in the buffer until lifetime is over. During this time router keeps retransmitting the packet in case of transmission fail. The packet lifetime is set equal to default value of 512 ms (as in Wi-Fi and 5G).
  • Transmission priority model. Modern transmission protocols have various priority mechanisms (JVET-AM0188, JVET-AM0201). Such mechanisms may be very useful for ULL video transmission scenarios. An approximated transmission priority model is used. The model operates by traffic ID (from 0 to 7) which may correspond to different layers, NAL unit types, data partitions, or any other categories depending on information importance.

Anchors

  • Single layer VTM. Regular VTM with regular LDB configuration is used.

Reported Metrics

Proposals on other metrics suitable for error resilient decoding are highly encouraged.

Software

The simulation software of AGH18 located at: https://vcgit.hhi.fraunhofer.de/jvet-ahg-ull/VVCSoftware_VTM. is used for latency constrained transmission emulation and anchors generation.

It was commented that there has been active work in this area, but JVET had not yet looked into all of the results of the CfE results in that area. The artefacts experienced in such scenarios are more difficult to evaluate, and it was mentioned that using traditional methods with short test sequence viewing may not be appropriate when such phenomena as frame freezes can occur.

ARCHIVE :
JVET-AM2026-v2-markup-ull.docx JVET-AN0204-v1.docx
Decisions
See notes in section 7.3.1.
See section 4.16.3.
It was commented that there has been active work in this area, but JVET had not yet looked into all of the results of the CfE results in that area. The artefacts experienced in such scenarios are more difficult to evaluate, and it was mentioned that using traditional methods with short test sequence viewing may not be appropriate when such phenomena as frame freezes can occur.
Citation