Can Open-RAN Networks Break the Mobile Monopoly

The mobile industry operates like a well-oiled machine, thanks to 3GPP, a central body that acts as a referee and negotiator. Companies collaborate here, agreeing to share their inventions at a fair price (FRAND) to avoid lawsuits. Here comes, O-RAN Alliance, seeking to improve collaboration between different equipment providers by creating even more open standards. Imagine having the freedom to build a mobile network just the way you want, without being tied to one vendor. That's what Open RAN offers – a new way of doing things in the telecommunications industry. Traditional RAN can feel like a closed system, limiting your choices and putting pressure on your budget. But Open RAN is changing that, opening up new possibilities for flexibility and choice. It's like building with Lego blocks, but for mobile networks. Let's take a closer look at Open RAN and how it's changing the game for operators like you.

O-RAN Alliance Architecture The O-RAN Alliance is driving a transformation in mobile networks by breaking them down into smaller parts – RU, DU, and CU components, along with different hardware platforms and software. This shift creates a more diverse ecosystem for building RAN infrastructure, moving away from the old vendor-specific solutions. With Open RAN, operators can pick and choose hardware and software from different suppliers, giving them more flexibility and scalability. It works with all mobile technology generations, from 2G to 5G and beyond. In addition to meeting 3GPP standards, O-RAN is pushing RAN capabilities towards openness and intelligence, offering key features like:

• Open standardized interfaces: These connect RUs and BUs, as well as the EMS, network management system, control and data planes, and the CU and DU.
• Decoupling software from hardware: This lets software run on different hardware platforms.
• Open hardware: Using platforms with general-purpose processors and accelerators.
• Open software: Ensuring commercial viability for high-performance real-time systems.

• Transport Bandwidth: Allows operators to tradeoff between flexibility and latency based on their chosen split option.
• Architectural Function Split: Based on vendor-neutral hardware and software, with open RAN interfaces defined by the O-RAN Alliance.
• Interoperability: Ensuring different vendor systems work together seamlessly with easily implementable and rigorously tested interfaces.

O-RAN defines additional interfaces (O1, O2, A1, E2) for managing and controlling the network, compared to 3GPP. These interfaces handle configuration, cloud services, policy, and real-time control. For lower layer splits, the existing eCPRI interface connects the DU to the RU, reducing data traffic and using Ethernet for efficiency.

Additionally, the O-RAN Alliance distinguishes the 7-2x split between Category A and Category B type O-RUs, depending on whether they support the precoding function. The Category B type O-RU supports multi-antenna systems for massive MIMO, while Category A O-RU supports RRHs with different Tx/Rx configurations. The O-RAN Alliance also introduces a new concept: split RICs based on latency. Non-RT RIC handles slower tasks like fault management, while near-RT RIC handles faster jobs like radio resource allocation. This division allows for complex AI/ML tasks to be performed in central locations and then executed in real-time, improving efficiency and compatibility with legacy systems.

Wireless networks share time and frequency resources among multiple users. As the number of user equipment (UE) increases in a cell, quality of service (QoS) can decline due to congestion. Admission control (AC) in LTE and 5G helps manage this by ensuring guaranteed QoS. AC algorithms decide whether incoming calls can be accepted based on preset rules, crucial for new and handoff calls. They help meet diverse QoS needs based on service type, user preference, and network load. Operators use static thresholds at the radio resource control (RRC) layer to limit users per cell. However, static AC struggles with varying traffic, device types, and QoS profiles, especially for 5G services like enhanced mobile broadband (eMBB) and ultra-reliable low-latency communications (URLLC). Dynamic methods, like those from AI/ML, improve AC efficiency, with deep learning-based AC outperforming traditional methods, especially in cloud and edge networking.

Integrating AI/ML in the O-RAN framework involves:

We evaluate an AI/ML-driven AC algorithm within O-RAN, integrating it as an xApp in the NS-3 simulation framework. This involves implementing the AC scheme in O-RAN, training the model, and integrating it with NS-3. Performance results are shown through extensive simulations.

The AC algorithm operates in the near-RT RIC as an xApp, dynamically controlling UE capacity at the cell level. It sends E2 messages to the CU to adjust capacities based on ML-driven QoS predictions. Key functions like QoS prediction and capacity recommendation run in the near-RT RIC, while real-time tasks like RRC connection handling occur in the CU.

The ML model, trained to predict PDCP delay, uses features like SINR and RSRP. Data is aggregated per cell and split into training and test sets. Various ML models, including ridge regression, KNN, random forest, and neural networks, are trained and validated. The best model, typically a gradient boosting or neural network, is deployed in a live network environment for real-time AC decisions.

The proposed AC algorithm integrates into NS-3 using model files like LteUEPhy, LteEnbRrc, and LteUePdcp. Measurements are collected every 1 ms and aggregated per cell.

Open RAN represents a transformative shift in the telecommunications industry, promising greater flexibility, innovation, and cost efficiency compared to traditional RAN systems. By disaggregating the network architecture into modular components and embracing principles of openness and intelligence, Open RAN allows operators to build and manage networks with unprecedented versatility. The O-RAN Alliance's efforts in standardizing interfaces and promoting interoperability among diverse vendor equipment further enhance the robustness and scalability of these networks. As 5G and beyond networks demand higher performance and adaptability, the integration of AI/ML-driven solutions within the O-RAN framework provides a powerful tool for optimizing network operations. Advanced admission control algorithms and dynamic network management capabilities ensure that the evolving needs of users and applications are met effectively. The future of mobile connectivity lies in the ability to innovate and adapt rapidly.