en.Wedoany.com Reported - Researchers at Northeastern University's Institute for Intelligent Networked Systems (INSI) have developed the GENESIS autonomous AI framework, which automates the development lifecycle of cellular 5G/6G Radio Access Network (RAN) software—from specification clauses or research ideas to over-the-air transmission code on commercial 5G hardware. The paper describing GENESIS (available on arXiv: https://arxiv.org/abs/2605.27360) presents the first end-to-end AI agent system capable of synthesizing, testing, hardening, optimizing, and securing a cellular network protocol stack without continuous human intervention, closing the engineering loop that previously required months per iteration.
Cellular R&D progresses slowly by structure. Translating 3GPP specification clauses into a working radio implementation typically requires months of human engineering, including reading dense standards documents, writing and debugging protocol code, integrating across multi-vendor protocol stacks for consistency and interoperability, and validating on commercial hardware. GENESIS directly targets this bottleneck: given a high-level intent—including specification clauses, telemetry anomalies, or research hypotheses—the framework autonomously plans, codes, tests, and iterates across a three-tier validation continuum, from software simulation, through channel simulation, to real-time over-the-air transmission on Northeastern University's Open6G testbed. Every artifact produced, including code changes, test results, and logs, is fed back into a persistent knowledge base, enhancing system capabilities over time.
In direct comparative experiments on representative 5G feature implementation tasks, GENESIS achieved a 100% success rate across multiple independent runs. In contrast, a baseline off-the-shelf state-of-the-art coding agent, given the same tools and testbed access, failed to produce a viable implementation in any attempt. The GENESIS paper details three end-to-end case studies: implementing 3GPP key performance measurements from specification to over-the-air reporting; synthesizing and hardening a conditional handover procedure with closed-loop optimization applications; and autonomously generating and validating a novel 5G MAC scheduling algorithm.
Tommaso Melodia, William Lincoln Smith Professor of Electrical and Computer Engineering at Northeastern University and Director of INSI, stated that autonomous AI is fundamentally changing the possibilities of wireless R&D. Translating 3GPP clauses into validated over-the-air transmission code historically required months of expert engineering, whereas GENESIS accomplishes this in hours—compressing timelines that even today's 5G roadmap struggles with and enabling the pace of innovation required for 6G. Melodia noted that the system's breakthrough lies not in any single coding agent, but in the closed loop itself: GENESIS reads specifications based on 3GPP and O-RAN, writes code, validates on a continuum of testbeds from simulation to real-time radio, and feeds every result back into the next iteration, thus taking an idea from intent to viable implementation in hours and opening the door to rapid prototyping.
GENESIS is built around three composable primitives: agents (AI reasoners with domain expertise), skills (deterministic parameterized programs that perform infrastructure operations), and hooks (event-driven safety gates and audit trails triggered around each action). These primitives are combined into six autonomous capability pipelines covering the complete RAN development lifecycle: SYNTHESIZE, TEST, HARDEN, OPTIMIZE, DISCOVER, and SECURE. A shared knowledge layer called SYNAPSE bases each agent decision on carefully curated 3GPP and O-RAN specifications and accumulates results from each run as institutional memory.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
