|
Work item:
|
F.MAC-A
|
|
Subject/title:
|
Architecture of multimodal agent collaboration in video surveillance systems
|
|
Status:
|
Under study
|
|
Approval process:
|
TAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2029 (Medium priority)
|
|
Liaison:
|
SG20, SG13, SG11, SG17, ISO/IEC JTC1/SC42
|
|
Supporting members:
|
Zhejiang Dahua Technology, Zhejiang University, China Mobile, China Telecom, BUPT, SGCC, Zhejiang Lab
|
|
Summary:
|
Video surveillance systems are evolving from traditional single-stream video processing toward multimodal sensing and intelligent decision-making. However, significant challenges remain in the practical implementation of multi-agent collaborative architectures. Firstly, system heterogeneity is prominent, with a lack of interoperability among devices, algorithm platforms, and data protocols from different manufactures. This makes it difficult to achieve low-latency, high-consistency fusion analysis of multi-source data such as video, audio, and radar. Secondly, there is a deficiency in inter-agent collaboration mechanisms. Current systems predominantly rely on centralized scheduling, which struggles to support autonomous coordination and resource optimization among distributed agents in dynamic environments, leading to delayed event response and imbalanced computational loads. These fragmented issues severely constrain the overall system efficiency and hinder the evolution of large-scale intelligent surveillance networks.
By establishing a unified standard for multimodal multi-agent collaborative architectures, the industry can systematically overcome existing bottlenecks. The Recommendation will first define interfaces to enable plug-and-play integration of heterogeneous devices and algorithm modules, facilitating seamless interoperability among video, audio, and third-party analytical tools. Secondly, the Recommendation can define a universal framework for agent communication, task allocation, and knowledge sharing, supporting flexible collaboration between edge and cloud resources. This will enhance real-time decision accuracy and resource utilization in complex scenarios. Furthermore, a standardized architecture will accelerate the deployment of intelligent technologies, driving the evolution of systems toward adaptive and scalable next-generation agent networks. Ultimately, this will enable video surveillance systems to transition from "perceptual stacking" to "cognitive collaboration".
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2026-07-28 14:41:04
|
|
Last update:
2026-08-05 14:37:50
|
|
