Wixor

AI-driven proactive TDD policy optimization system for 5G/xG networks

Wixor

AI-Powered Dynamic TDD Policy Adaptation for 5G/xG Networks

Up to
96%
QoE Improvement
Within
88%
of the Oracle
Less than
10%
System Overhead

Awards & Recognition

🏆 CoNEXT 2024 Best Paper Runner-Up 📢 Presented at HPE Tech Con 2024

Demo Videos

Demo videos showcasing Wixor's performance will be added here.

Problem & Motivation

In the era of 5G and beyond, dynamic Time Division Duplex (TDD) has become essential for supporting applications that demand high bandwidth and low latency. Traditional 5G deployments use static, downlink-biased TDD configurations that fail to serve emerging customer-facing applications requiring heavy uplink bandwidth or tight latency deadlines. These include AR/VR streaming, live video ingest, industrial automation, real-time perception analytics, drone telemetry, remote assistance, and high-fidelity sensor workloads.

Who can Benefit from Wixor?

  • Mobile Network Operators aiming to deliver superior quality-of-experience (QoE) for consumers and enterprises while improving spectrum efficiency.
  • Enterprises deploying private 5G seeking deterministic performance for automation, robotics, analytics, and safety-critical workloads.
  • Application and cloud providers that depend on reliable, low-latency UL connectivity to support next-generation digital services.
  • Government and smart-city operators deploying real-time video, mobility intelligence, and IoT services.

Wixor is built as a 3GPP-compliant, operator-friendly software module that seamlessly integrates with public 5G networks, enterprise private 5G installations, and neutral-host environments.

Key Innovations

Proactive RL-based optimization: Forecasts demand in advance, unlike reactive solutions
Cross-layer intelligence: Uses BS-level features for scalability across dense environments
Joint UL/DL and pattern optimization: Optimizes both slot distribution and arrangement
Transport-aware policy smoothing: Prevents TCP destabilization during TDD transitions
Deployment-ready: Fully implemented 2.3k-line C/C++ system integrated with srsRAN

Results

Wixor’s effectiveness has been demonstrated through extensive evaluation totaling more than six hours of real workloads, including edge perception, video analytics, mobility scenarios, and live video transmissions.

Application-Level Performance

Overall QoE improvement for application workloads and 6+ hours of channel traces. Wixor outperforms baselines for most metrics. The up (↑) and down (↓) arrows indicate the direction of QoE improvement. The gray shaded region highlights the best performing scheme. Default, SFair, Reactive, and DRP are the baselines.

Application Metric Default SFair Reactive DRP Wixor
Edge Video Analytics (EVA) Response Latency (ms) ↓ 93.3 77.2 44.9 57.4 38.1
Perceptive Accuracy (%) ↑ 34.7 40.1 54.8 47.6 68.2
Edge-assisted Autonomous Vehicle Perception (EAVP) Response Latency (ms) ↓ 67.8 60.3 51.7 56.8 46.3
Mean IoU ↑ 0.68 0.71 0.77 0.74 0.78
Live Video Ingest (LVI) Ingest Delay (ms) ↓ 284.9 255.3 246.4 233.8 191.5
Sending Bitrate (Mbps) ↑ 3.8 5.5 5.8 6.1 6.2

Network-Level Improvements

  • • Wixor's fairness and spectral efficiency are as high as the default TDD policy (less than 2% difference).
  • • The gap between Wixor and the optimal TDD policy is small: Wixor is within 82.2% and 88.0% of the Oracle for per-packet latency and BS throughput, respectively.
  • • Wixor's gains are larger for higher mobility scenarios.

Collaborators

This work is a collaboration between:

USC
HPE

References

Ahmad Hassan, Shivang Aggarwal, Mohamed Ibrahim, Puneet Sharma, and Feng Qian. 
2024. Wixor: Dynamic TDD Policy Adaptation for 5G/xG Networks. 
Proc. ACM Netw. 2, CoNEXT4, Article 38 (December 2024), 24 pages. 
https://doi.org/10.1145/3696395