MagmaML: Towards Automated Management for Low-resource 5G Cellular Network Deployments
summary
Project integrates Magma with campus testbeds RENEW and POWDER, builds labeled fault‑injection datasets (high user load, malfunctioning UE, interfering node, bad eNodeB/gNodeB configs) and adds a Magma microservice that combines rule‑based inference and trained ML models to detect and suggest fixes. Physical‑layer stack uses Agora and achieves 64x16 real‑time MIMO on a 36‑core Intel server meeting 4 ms 5G latency. Initial focus is offline dataset collection and model prototyping for automated fault discovery and recovery in low‑resource deployments.
AI-written summary · about
chapters (10)
- 0:00Introduction to MagmaML
- 1:51Challenges in cellular network management
- 3:17City-scale wireless testbeds
- 5:02Rice RENEW and POWDER testbeds
- 7:09Open source RAN software limitations
- 8:54Agora and massive MIMO processing
- 10:19MagmaML project goals
- 13:27MagmaML system architecture
- 14:10Summary and future work
- 14:55Q&A session
related talks
20mBuilding Flexible, Low-Cost Wireless Access Networks With Magma
18mMagma 101: An introduction to Magma
36mPanel: Magma End Users
04mLightning Talk: Enabling Enterprise Cellular Mesh Solutions
24mMission Possible: Forging a New Paradigm in 5G Deployment...
01h 33mDay 3 Part 2 Sessions
18mEnabling SLO-Aware 5G Multi-Access Edge Computing with SMEC
22mDefensive 5G