OpenAirInterface (OAI) Framework

Updated 21 January 2026

OpenAirInterface (OAI) is an open-source framework that implements full-stack 3GPP wireless protocols for LTE, 5G NR, and O-RAN, supporting all modern network domains.
It facilitates reproducible software-defined radio testbeds and rapid prototyping, integrating with virtualization, orchestration, and AI/ML-driven management.
OAI’s modular architecture and community-driven development empower research in cutting-edge wireless innovations, from advanced MIMO to experimental protocol splits.

OpenAirInterface (OAI) is a comprehensive open-source software framework implementing full-stack 3GPP wireless protocols for LTE, 5G NR, and supporting key O-RAN interfaces and innovations. OAI enables reproducible software-defined radio systems, emulated and over-the-air testbeds, and rapid prototyping of network and air-interface features for both academic research and pre-commercial deployments. It encompasses all major domains of modern cellular networks: Radio Access Network (RAN), Core Network (CN), User Equipment (UE), Operations, Administration and Maintenance (OAM), and exposes programmable interfaces for integration with system virtualization, orchestration, and AI/ML applications. The project is governed by the OpenAirInterface Software Alliance and is developed collaboratively by leading academic, industry, and community partners worldwide (Kaltenberger et al., 2024).

1. Historical Evolution, License, and Governance

OAI originated in the early 2000s at EURECOM as an SDR-based LTE RAN prototype running on x86 commodity servers and custom RF front-ends (Kaltenberger et al., 2024). By 2010, it demonstrated the first LTE Release 8 softmodem with interoperability against COTS smartphones and compliant EPCs using USRP radios. As the user and contributor base expanded, the non-profit OpenAirInterface Software Alliance (OSA) was established in 2014 to provide transparent governance, a coordinated roadmap, integration/test labs, and to manage the OAI Public License v1.1 (an Apache 2.0–like license with FRAND-patent clauses aligned to 3GPP’s SEP regime). The OSA board includes major academic, operator, silicon, and manufacturing stakeholders.

Subsequent milestones included the first 5G NR R15/NSA system (2020, CU/DU split), open-source New Radio and Core (2021), O-RAN 7.2 fronthaul, F1/E1 and E2/O1 interfaces (2023–24), and full-featured software-defined UE stacks. OAI maintains a CI/CD pipeline blending human review, unit/integration tests, container image builds, and regular OTA/emulated validation for each merge (Kaltenberger et al., 2024).

2. Core Architectural Components

OAI implements the 3GPP protocol stack for LTE and NR, including O-RAN NVF/CNF-compliant splits. Architectural modules are:

2.1 Radio Access Network (OAI-RAN)

Can be deployed monolithically (gNB) or using split CU/DU (with F1/E1) and O-RAN standardized functional splits (7.2, 8).
PHY supports FR1 (up to 100 MHz, 15/30 kHz SCS), FR2 (up to 200 MHz, 120 kHz SCS), with systematic support for TDD/FDD, 4×4 downlink/2×2 uplink MIMO, and advanced coding (LDPC/turbo/polar).
MAC/RLC/PDCP/SDAP/RRC together support HARQ, AM/UM RLC, QoS flows, and complete full-state machine operations.
O-RAN interfaces: supports O-RU fronthaul over eCPRI/split 8 (and 7.2), E2 for near-RT RIC, O1 for SMO/OAM integration.

2.2 Core Network (OAI-5GC)

Service-Based Architecture (Rel-15+) with stateless (cloud-native) implementations of AMF, SMF, UPF (optionally eBPF/XDP), AUSF, UDM, UDR, NRF, NSSF, PCF, NWDAF, and LMF for positioning (Kaltenberger et al., 2024).
APIs for registration, session management, handover, slicing, multi-UPF, UL classifiers, and real-time analytics via NWDAF.

2.3 Operations, Administration and Maintenance (OAM)

Includes an automated CI/CD lab, testing with real UEs and RUs/USRP, hooks for commercial system validation, and integration with orchestration systems such as ONAP, OSM, and Colosseum (Kaltenberger et al., 2024).
Continuous AI/ML driven test processes and performance monitoring.

2.4 Software-Defined User Equipment

OAI supports full SDR UE stacks, featuring modular RRC to PHY implementation, and interoperable with commercial gNBs by timing parameter tuning.
Supports multi-slice PDU session, real-time channel state reporting, NTN modes (GEO/LEO), and positioning enhancements (Kaltenberger et al., 2024).

3. Feature Implementation: Protocols and O-RAN Integration

OAI is at the forefront of open RAN innovation with sysadmin/developer-controllable protocol splits and hardware acceleration (Kaltenberger et al., 2024, Villa et al., 2023):

PHY acceleration is supported both inline (e.g., via NVIDIA Aerial SDK on A100 GPUs) and in a look-aside model (e.g., AMD Xilinx T1/T2 with Intel bbdev).
OAI’s FAPI (Small Cell Forum standard) is leveraged to decouple L2 (MAC/RLC) and L1 (PHY), enabling deployment of GPU-accelerated lower PHY and CPU upper layers. The FAPI messages synchronize slot-level grants, feedback, and configuration (Villa et al., 2024, Villa et al., 2023).
OAI’s E2 agent implements the O-RAN E2AP, exposing KPM/RC service models for RIC xApp integration (Villa et al., 2024, Kaltenberger et al., 2024).
OAI supports NTNs (Rel-17), positioning with LMF/NRPPa (UL-TDoA, PRS), and 5G NR Sidelink; the platform includes robust support for slicing, beam management, and experimental SMO hooks (Kaltenberger et al., 2024, Elkadi et al., 2023).

4. Research Applications and Testbed Deployments

OAI is deployed in a diverse set of research and pre-commercial environments (Kaltenberger et al., 2024):

Campus/private 5G: EURECOM’s Open5GLab, X5G at Northeastern University, and POWDER at Utah deploy OAI in scalable, multi-vendor O-RAN testbeds (Villa et al., 2024).
RF Digital Twin: OAI is used in digital twin frameworks, such as integrating high-fidelity ray tracing from NVIDIA Sionna RT and KPI reporting through FlexRIC/E2 for closed-loop channel measurements and control (Iye et al., 15 Mar 2025).
Slicing and DRL: OAI is used as the RAN substrate for DRL-based slicing xApps, real-time E2-controlled MAC resource scheduling, and performance analytics (Sever et al., 10 Jan 2025).
Positioning: End-to-end 3GPP-compliant UL-TDoA, PRS, and CNN-based fingerprinting for sub-meter indoor/outdoor localization have been demonstrated via OAI’s LMF and O-RAN extensions (Malik et al., 2024, Ahadi et al., 27 Aug 2025, Bouknana et al., 24 Nov 2025).
Virtualization: Containerized C-RANs with OAI as RRH/BBU, supporting clear separation between radio and baseband for research into NFV/edge splits (Trindade et al., 2019).
OAI Extensions: Full stack MIMO/MU-MIMO frameworks, fronthaul compression (Split-7.1 with BVC), hardware-accelerated PHY, and standalone OAI UEs are reproducibly described (Nahum et al., 2020, Bui et al., 29 Sep 2025, Bui et al., 29 Sep 2025, Iye et al., 15 Mar 2025, Bui et al., 29 Sep 2025).

5. Advanced Functionality: MIMO, MU-MIMO, Positioning, and AI/ML

OAI supports 3GPP release–aligned and experimental features with precise control and transparency:

Closed-Loop 2×4 MIMO: Extended CSI reporting at the UE (RI, PMI, CQI), and two-layer PDSCH scheduling yield up to 1.8 Gbps with robust gains over the baseline 2×2 configuration (Bui et al., 29 Sep 2025).
Downlink and Uplink MU-MIMO: OAI supports real-time uplink MU-MIMO scheduling (via SRS channel estimation, RZF combiners) and simultaneous downlink MU-MIMO with orthogonal user PMIs (Uçak et al., 14 Jan 2026, Bui et al., 29 Sep 2025).
Positioning: OAI prototypes 5G NR positioning via RTT, UL-TDoA, and CIR-based inference. Integration of LMF and NRPPa supports per-3GPP UL-TDoA flows, validated against both simulators and O-RAN testbeds (MAE down to 0.35–5.4 m; 90th–tile error ≈2 m), with extensions to CNN-based fingerprinting for severe NLoS (Ahadi et al., 27 Aug 2025, Malik et al., 2024, Bouknana et al., 24 Nov 2025, Mundlamuri et al., 2024).
AI/ML Frameworks: Via FlexRIC (E2 xApp environment), OAI exposes real-time RAN KPIs and full SRS channel estimates, supporting AI-driven resource allocation, scheduler design, and spatial positioning (e.g., channel charting xApps, DRL scheduling agents) (Sever et al., 10 Jan 2025, Bouknana et al., 24 Nov 2025, Kaltenberger et al., 2024).

6. Software Development, Reproducibility, and Open Innovation

OAI is developed with open engineering practices to support transparent, reproducible research and pre-standard investigations:

Build and Deploy: Step-by-step build/deploy flows are documented for all modules. OAI maintains reproducible build branches (e.g., for NB-IoT, 2×4 MIMO) with sample configuration files (Chen et al., 2020, Bui et al., 29 Sep 2025).
Community: Regular roadmap publications, public GitLab for code review/merges, open test case repositories, and OpenRAN Gym tutorials foster collaborative innovation (Kaltenberger et al., 2024).
Interoperability: OAI’s reference stack is interoperable with major open-source and commercial network elements (e.g., Open5GS, Amarisoft UE tester, Keysight RIC Test) (Kaltenberger et al., 2024).
Limitations and Extensions: Limitations (such as no NB-IoT in core, container-level virtualization overhead not directly quantified, PHY bugs) are transparently reported, with code-level remedies and next-step recommendations documented (Chen et al., 2020, Trindade et al., 2019).

7. Impact and Prospective Directions

As an open, fully programmable, end-to-end wireless system stack, OAI catalyzes rigorous, reproducible research in wireless network architectures, protocol engineering, AI/ML-driven RAN, and system-level optimization for LTE, 5G, and pre-standard 6G. The platform’s extensibility and community integration make it an anchor for new algorithmic advances (RIS optimization, slice orchestration, AI-native RIC control, cell-free networking), and for large-scale experimental research in digital-twin, NTN, and advanced wireless scenarios (Kaltenberger et al., 2024, Villa et al., 2024, Iye et al., 15 Mar 2025). Ongoing work aligns OAI with Rel 18/19, AI-native RAN, and full-blown 6G studies, with continuous contributions and validation across academic and industrial partners.