Innovative Application of AI Technology in Communication Network Planning

With the rapid development of information and communication technologies, artificial intelligence (AI) is gradually penetrating various industries, demonstrating great potential and value in network construction and O&M fields. The application of AI technology in communication networks can significantly improve network intelligence and effectively address many challenges in network construction and O&M. ZTE actively explores innovative practices in intelligent technologies such as AI and large models in the communications field. In network planning, ZTE has introduced an innovative solution, the AI robot for neighbor cell planning. This solution, as a crucial manifestation of AI technology in network planning, addresses the weaknesses of traditional neighbor cell planning applications, providing a brand-new solution for network neighbor cell planning and facilitating efficient network deployment.

What Is AI Robot Needed for Neighbor Cell Planning

In the 2G and 3G eras, neighbor cell planning often relied on manual and tool-based methods. In the 4G era, the introduction of automatic neighbor relation (ANR) technology gradually automated this process. However, the ANR technology requires cell sites to operate within the network for a period before new neighbor cells can be added, and it does not support planning before the cell site accesses the network. As a result, the timeliness of neighbor cell planning is affected.

Tool-based planning involves adding multiple nearby cells to the neighbor cell list in the forward/backward coverage direction of the cell, determined by factors such as the cell’s longitude, latitude, azimuth, distance, and angle. However, existing neighbor cell planning methods based on distance and angle have problems such as missing neighbor cell configurations, low accuracy in neighbor cell planning, and an excessive number of neighbor cells, which fail to guarantee the accuracy of neighbor cell planning. Therefore, there is a need for a more timely and accurate neighbor cell planning application to adapt to the increasingly complex communication network environment.

The AI robot for neighbor cell planning innovatively incorporates factors such as the number of neighbor cell layers, distance multiplier, and overlapping coverage area. It can use machine learning to analyze massive neighbor cell data in the existing network, self-learn from the experience of adding neighbor cells in the network, and self-predict new neighbor cells before accessing the network. This implementation enables intelligent and automatic neighbor cell planning, enhancing the timeliness and accuracy of the process.

Self-Analyzing, Self-Learning, Self-Planning Neighbor Cell Planning Solution

As a typical application of AI technology in communication network deployment, ZTE’s AI robot for neighbor cell planning comprises several modules, including basic data extraction, data processing, feature calculation, model training, neighbor cell planning for new sites, and neighbor cell script making. This solution can effectively address the pain points and challenges encountered in each procedure of neighbor cell planning.

• Introducing Hierarchical Structure by Using Geometric Knowledge

The AI robot for neighbor cell planning innovatively introduces the concept of Delaunay triangle network. Delaunay Triangles, a common geometric concept, are used to divide a given set of points into disjoint triangles, ensuring no other points lie within the outer circle of each triangle. Utilizing the engineering parameters of cell sites, the robot treats each cell site as a vertex and uses latitude and longitude data to build a Delaunay triangle network. By traversing the adjacent relationships between triangle vertices, the robot can establish the hierarchical relationships between cell sites.

This method, based on the Delaunay triangle, considers the locations of cell sites as a point set. It avoids the creation of simple one-to-one neighbor cell relationships and instead establishes relationships through a hierarchical structure. This approach makes neighbor cell planning more effective, reduces radio interference and handover times, and thus enhances overall network performance and user experience.

• Learning Rules of Adding Neighbor Cells to Existing Network Through Massive Data Analysis

AI technologies have the capability of processing vast amounts of data, including both structured data (such as tabular data in the database) and unstructured data (such as text, images, audio, and video). Compared with traditional methods, AI technologies, especially deep learning, can efficiently handle large datasets through parallel computing and distributed processing. They can discern patterns within these datasets and provide decision-making support more efficiently.

The AI robot for neighbor cell planning uses the deep learning technology to extract valuable information from vast amounts of existing network data, including engineering parameters and neighbor cell configuration data. It summarizes the rules of neighbor cell addition, which are usually difficult to discover or understand using traditional statistical methods. This enables the robot to automatically learn and adapt to new data and environment changes. Once the model is established, it can continuously learn new neighbor cell configurations within the existing network and optimize them, thereby steadily enhancing performance and accuracy. This learning mechanism also allows the robot to be applied to different operator networks and to train AI models applicable to various operator networks.

• Planning New Neighbor Cell Sites Through Intelligent AI Prediction

During the learning process, the AI model will attempt to generalize existing knowledge, meaning it applies the rules learned from training data to new, unprocessed data. This capability allows the model to make reasonable predictions in novel situations, rather than simply memorizing the training data. The neural network is a common model in deep learning, employing a network composed of multiple layers of neurons to model complex data relationships.

Using the multi-layer neural network model, the AI robot for neighbor cell planning reads data from both newly constructed sites and existing networks to make intelligent decisions in complex environments. Once the model completes training and is deployed in the actual network, the robot can predict the best neighbor cell planning solution based on the geographical location and surrounding environment of the new site, aiming to optimize network coverage and performance. This intelligent neighbor cell planning method, based on the deep learning model, significantly improves the efficiency and accuracy of network neighbor cell planning, providing robust technical support and optimization solutions for telecom operators.

Application of AI Robot for Neighbor Cell Planning

With its strong adaptability and accurate planning capability, the AI robot for neighbor cell planning has been widely applied in many overseas projects, providing robust support for intelligent network deployment.

In a large-scale dual-network convergence project in a Southeast Asian country, various network scenarios and site statuses are involved, making neighbor cell planning complex and demanding high computer performance. The AI robot for neighbor cell planning is used to learn the neighbor cell data of the project within the existing network. Leveraging the frequency planning and special rules for neighbor cell addition in the existing network, the time required for neighbor cell planning and configuration is reduced from 600 minutes each time to just ten minutes. This significant reduction greatly improves network convergence efficiency for the operator.

In an overseas large-scale swap project, neighbor cell planning and optimization scenarios are highly complex, involving intra-frequency, intra-layer, inter-frequency, and inter-system neighbor cell planning and optimization. Traditionally, it takes two weeks to optimize neighbor cells across the entire network each time. However, the AI robot for neighbor cell planning can automatically output key information, such as neighbor cell pair relationships, number of neighbor cell layers, and forward/backward relative azimuth, within a few minutes. The planning can be completed in just a few minutes, and network-wide neighbor cell optimization can be completed within one day, greatly improving the efficiency of network swap and O&M.

Looking ahead, the AI robot for neighbor cell planning will continue to evolve and grow, comprehensively enhancing the intelligence of network neighbor cell planning. This advancement aims to deliver more intelligent and high-quality services and experiences, driving the network towards an intelligent future.