Digital Twin-Based Network Management for Better QoE in Multicast Short Video Streaming

Xinyu Huang,Shisheng Hu,Haojun Yang,Xinghan Wang,Yingying Pei,Xuemin Shen

2024-01-23

Abstract:Multicast short video streaming can enhance bandwidth utilization by enabling simultaneous video transmission to multiple users over shared wireless channels. The existing network management schemes mainly rely on the sequential buffering principle and general quality of experience (QoE) model, which may deteriorate QoE when users' swipe behaviors exhibit distinct spatiotemporal variation. In this paper, we propose a digital twin (DT)-based network management scheme to enhance QoE. Firstly, user status emulated by the DT is utilized to estimate the transmission capabilities and watching probability distributions of sub-multicast groups (SMGs) for an adaptive segment buffering. The SMGs' buffers are aligned to the unique virtual buffers managed by the DT for a fine-grained buffer update. Then, a multicast QoE model consisting of rebuffering time, video quality, and quality variation is developed, by considering the mutual influence of segment buffering among SMGs. Finally, a joint optimization problem of segment version selection and slot division is formulated to maximize QoE. To efficiently solve the problem, a data-model-driven algorithm is proposed by integrating a convex optimization method and a deep reinforcement learning algorithm. Simulation results based on the real-world dataset demonstrate that the proposed DT-based network management scheme outperforms benchmark schemes in terms of QoE improvement.

Networking and Internet Architecture,Image and Video Processing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the user's Quality of Experience (QoE) in multicast short - video stream transmission. Specifically, the existing network management schemes mainly rely on the sequential buffering principle and the general QoE model, and these methods may deteriorate QoE when the user's sliding behavior shows significant spatio - temporal variations. To improve this situation, the paper proposes a network management scheme based on Digital Twin (DT) to enhance QoE.

### Main Problems and Solutions

1. **Buffering Problems Caused by Non - sequential Viewing Behaviors**
- **Problem**: Due to the user's random sliding behavior, the user's viewing sequence is usually discontinuous. The existing sequential buffering principle may lead to the situation that the video segments to be viewed are not buffered in time, thus causing playback delay.
- **Solution**: Through DT to simulate the user's future network conditions and sliding behaviors, abstract the viewing probability distribution of video segments, and make adaptive video - segment buffering decisions accordingly.

2. **Complexity of the QoE Model**
- **Problem**: The QoE model is usually composed of multiple factors, such as rebuffering time, video quality and quality change. The segment - buffering priorities among different Sub - Multicast Groups (SMGs) will affect these factors, thus affecting the QoE of the entire Multicast Group (MG).
- **Solution**: Establish a QoE model that takes into account the mutual influence of multicast - segment buffering. This model is composed of the weighted sum of rebuffering time, video quality and quality change, and the weight factors combine the buffering order and the user's sensitivity.

3. **Complexity of Multivariable Optimization Problems**
- **Problem**: Network management is usually a multivariable decision - making problem, such as segment - version selection and resource scheduling. The complexity of these problems makes it difficult to solve directly using model - based or data - driven algorithms.
- **Solution**: Propose a data - model - driven algorithm, which combines convex optimization methods and deep reinforcement learning algorithms to decouple the joint optimization problem and reduce the action dimension, thus efficiently solving the optimization problem.

### Main Contributions of the Paper

1. **Propose a New DT - Assisted Buffering Management Scheme**
- **User - State Simulation**: Store historical state information such as the user's geographical location, channel conditions, preferences and sliding timestamps for state simulation.
- **Adaptive - Segment - Buffering Decisions**: Based on the abstracted information, DT can make adaptive segment - buffering decisions and construct and manage the virtual buffer of each SMG to achieve fine - grained buffer updates.

2. **Establish a Multicast QoE Model**
- **Model Construction**: The QoE model is composed of the weighted sum of rebuffering time, video quality and quality change, where the weight factors combine the buffering order and the user's sensitivity.
- **Model Elements**: The rebuffering - time estimation depends on the multicast - transmission delay and the parallel - transmission and transcoding processes; video quality and quality change depend on the segment version and the Structural Similarity Index Measurement (SSIM).

3. **Propose a Data - Model - Driven Algorithm**
- **Algorithm Design**: Combine convex optimization methods and deep reinforcement learning algorithms to decouple the joint optimization problem and reduce the action dimension, and efficiently solve the multivariable optimization problem.
- **Performance Verification**: Extensive simulation results based on the real - world short - video - stream data set show that the proposed DT - assisted network management scheme is superior to the existing network management schemes in terms of QoE improvement.

Through the above schemes, the paper aims to effectively improve the user - experience quality in multicast short - video - stream transmission.