EdgeRIC: Empowering Real-time Intelligent Optimization and Control in NextG Cellular...
Talkat 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) · 2024-04-16..04-18·
summary
A real-time RIC called EdgeRIC is implemented co‑located with the DU (built on SRS RAN) and synchronized to the TTI with 1 ms TTI sync, using a lightweight ZeroMQ messaging RTE2, an in-memory Redis DB, and a GNU Radio emulator for offline training. Deployed as micro apps, RL agents trained on the emulator run in-loop and yield practical gains: up to ~25% system throughput improvement and AI policies that beat model-based strategies by 5–25% and reduce video rebuffering.
AI-written summary · about
chapters (12)
- 0:00EdgeRIC vision
- 2:01Cloud RAN limits
- 5:31EdgeRIC for real-time edge
- 6:34EdgeRIC design
- 7:37EdgeRIC performance gains
- 9:43Technical architecture and synchronization
- 10:46Micro apps & reinforcement learning
- 11:50Edge-cloud collaboration
- 12:53Demo: RL scheduling
- 13:14Max throughput scenario
- 15:42Video QoE scenario
- 17:07Conclusion & Q&A
related talks
11mA Comparison of Communication Mechanisms in Vehicular Edge Computing
18mEnabling SLO-Aware 5G Multi-Access Edge Computing with SMEC
17mTraffic Engineering for AI Training Networks
01h 25mDay 2 Part 1 Sessions
17mTowards Energy Efficient 5G vRAN Servers
21mScaling RoCE Networks for AI Training
12mP4EC: Enabling Terabit Edge Computing in Enterprise 4G LTE
19mProgramming the Electromagnetic Spectrum