BOSTON, June 4, 2026 /PRNewswire/ — Researchers at Northeastern University’s Institute for Intelligent Networked Systems (INSI) have developed GENESIS, an agentic AI framework that automates the full research-and-development life-cycle of Radio Access Network (RAN) software for cellular 5G/6G — from translating a specification clause or research idea directly into validated, over-the-air code on production 5G hardware. The article describing GENESIS (available online: https://arxiv.org/abs/2605.27360) represents the first end-to-end demonstration of an AI-agent system that can synthesize, test, harden, optimize, and secure a cellular network stack without continuous human intervention, closing an engineering loop that has historically taken months per iteration.
Cellular R&D is structurally slow. Bringing a single new feature from a 3GPP specification clause to a working radio implementation typically requires months of manual engineering, including reading dense standards documents, writing and debugging protocol code, integrating across multi-vendor stacks for conformance and interoperability, and validating on production hardware. GENESIS addresses this bottleneck directly. Given a high-level intent — a specification clause, a telemetry anomaly, or a research hypothesis — the framework autonomously plans, codes, tests, and iterates across a three-tier validation continuum: from software simulation, through channel emulation, to live over-the-air transmission on the Open6G testbeds at Northeastern. Every artifact produced — code changes, test results, logs — feeds back into a persistent knowledge base, compounding the system’s capabilities over time.
In head-to-head experiments on a representative 5G feature implementation task, GENESIS achieved a 100% success rate across multiple independent runs. An off-the-shelf state-of-the-art coding agent used as a baseline — given access to the same tools and testbeds — produced no working implementation on any attempt. The GENESIS paper details three end-to-end case studies: implementing a 3GPP Key Performance Measurement from specification to over-the-air reporting; synthesizing and hardening a Conditional Handover procedure with a closed-loop optimization application; and autonomously generating and validating new MAC scheduling algorithms for 5G.
“Agentic AI is fundamentally changing what’s possible in wireless R&D. Translating a 3GPP clause into validated, over-the-air code has historically taken months of expert engineering, while GENESIS does it in hours. That collapses a timeline that even today’s 5G roadmaps strain against, and makes the pace of innovation 6G will demand achievable,” said Tommaso Melodia, Director of INSI and William Lincoln Smith Professor with the Department of Electrical and Computer Engineering at Northeastern University. “The breakthrough isn’t any single coding agent. It’s closing the loop. GENESIS reads grounded 3GPP and O-RAN specifications, writes the code, validates it across a continuum of testbeds from simulation to live radios, and feeds every result back into the next iteration. That closed loop is what carries an idea from intent to working implementation in hours, and it is what opens the door to rapid prototyping of features that would otherwise never make it out of a whiteboard or of a specification document.“
GENESIS is built around three composable primitives: agents (AI reasoners with domain-specific expertise), skills (deterministic, parameterized procedures that execute infrastructure operations), and hooks (event-driven safety gates and audit trails that fire around every action). These primitives compose into six autonomous capability pipelines covering the complete RAN R&D life-cycle: SYNTHESIZE, TEST, HARDEN, OPTIMIZE, DISCOVER, and SECURE. A shared knowledge plane called SYNAPSE grounds every agent decision in curated 3GPP and O-RAN specifications and accumulates the results of every run as institutional memory.
The framework is designed to be portable across major agentic AI runtimes and LLM backends, and is validated on the Open6G testbeds — production-grade, multi-site private 5G O-RAN platform at Northeastern University. The full paper, “GENESIS: Harnessing AI Agents for Autonomous 6G RAN Synthesis, Research, and Testing,” is available on arXiv: https://arxiv.org/pdf/2605.27360.
About the Institute for Intelligent Networked Systems (INSI)
INSI at Northeastern University is a leading research and innovation hub focused on AI-native wireless systems, Open RAN, and next-generation 5G/6G networks. INSI is home to Open6G, a government-industry-university cooperative R&D center pushing the boundaries of open, programmable, and AI-powered wireless systems. Open6G works with partners to design, prototype, and validate advanced network technologies on large-scale experimental platforms, accelerating the development of secure, energy-efficient, and scalable wireless solutions.
Contact: [email protected]
SOURCE Institute for Intelligent Networked Systems
