By Ghyslain Pelletier, InterDigital
From user devices to base stations to the network core, AI will be written into all aspects of 6G. Combining that with Integrated Sensing and Communications should give 6G capabilities not found in previous wireless generations.
The next generation of mobile networks will be unlike anything to come before it. While every generational leap in wireless technology has brought new capabilities, 6G will be the first designed from the start with support for Artificial Intelligence (AI) as a core feature. In 4G and 5G, AI was primarily used to improve network performance. With 6G, AI will enable it to make a broader range of decisions and to self-optimize. AI will play a pivotal role in 6G, enhancing many aspects of connectivity with intelligence-driven capabilities.
In Release 20, the 3GPP standards community is already exploring how AI and ML can optimize the air interface, building on previous work on AI/ML applications in the core network, radio-access network (RAN), and air interface for 5G-Advanced. This new work includes using one-sided models to improve mobility management. The work also includes two-sided model deployments where AI models are running both on the device and the network to make tasks such as channel state information (CSI) compression more efficient.
Figure 1. In 6G, AI will be everywhere from the radio to the network core.
The transformative impact of AI on future 6G networks will be nothing short of pervasive, with the potential to improve performance at any layer of the wireless system. Figure 1 shows how AI and cloud computing will affect the radio through to the network core.
At the application level, AI will enable entirely new categories of end-user services that will include personalized immersive experiences to ultra-reliable, real-time applications.
Within the core network, AI-driven automation will streamline complex orchestration tasks, self-heal disruptions, and dynamically adapt to traffic demands with minimal human intervention.
Figure 2. AI will enter user devices, operate at the network edge, and in the telecom network itself.
On the radio access side, machine learning will unlock new levels of efficiency in network planning. ML will learn user behaviors and mobility patterns to deliver more consistent Quality of Service and increased end-user Quality of Experience. At the base station level, adaptive algorithms will optimize the performance of the radio to mitigate interference, reduce latency, and squeeze more performance from the finite spectrum (Figure 2).
Building networks that think for themselves
A 6G network will go beyond the traditional communications system; it will sense, process, and intelligently respond in milliseconds. This requires a system design that uses AI at any layer of the system, from the device in the user’s hand, through the edge of the network (Figure 3), and into the operator’s core network. These AI models mustn’t be static. They will need to be trained, improved, or replaced as conditions evolve, and decisions must often be made close to where the data is generated to reduce delay.
Figure 3. Operating in base station and user device radios, AI will improve connections and optimize resources.
This kind of intelligence can transform networks, and AI can improve radio operation. It can enhance beamforming, allowing antennas to focus on signals more effectively. It can predict when a user is likely to move from one cell to another by preparing the handover in advance to prevent service interruptions. It can dynamically manage interference, so users enjoy better reliability and speed.
AI can reduce the overhead in reference signals used to estimate the radio channel for Demodulation Reference Signals (DM-RS), which can increase capacity without sacrificing reliability. The result: faster speeds, more stable connections, and better battery life for end users. It can even increase efficiency and reliability of feedback information by combining data compression, error protection, and modulation into a single process that can adapt to both channel and service conditions. Advanced use cases, such as joint source–channel coding and modulation (JSCCM), could eventually be explored in later phases of 6G to further improve efficiency and reliability in challenging radio environments.
This kind of intelligence can also transform the way people use their mobile devices. AI can enhance the end-user experience with new mobile AI services deployed in devices such as smartphones and smart glasses, which can capture and communicate a user’s context together with a personalized, possibly task-driven request and provide a response in real-time in a secure manner that respects the end-user’s privacy.
Radio design for new applications
This transformation is not just about optimizing existing services, but about enabling a completely new generation of applications. XR, immersive media, real-time global collaboration, and Mobile AI-services will heavily depend on the network’s uplink capacity, as users send far more data from their devices to the network than in today’s download-dominated traffic patterns.
To support this, the 6G radio will instantly allocate resources to where they are needed most, adjust coding based on immediate conditions, use AI-enhanced video codecs to deliver rich, high-quality experiences, and facilitate the success rate of machine-based processing for Mobile AI applications with less data.
InterDigital’s research explores a technique that partitions the AI/ML model into two distinct components — Feature Coding for Machines and split-squeeze computing — that run concurrently on the local device and the remote server to better balance AI processing requirements. Split-squeeze computing can optimize AI applications deployed at scale by balancing acceptable latency with reasonable power consumption across a variety of devices. By processing only the most important data before transmission, devices can save both bandwidth and battery life, which is a significant advantage for XR headsets, drones, and other devices.
Communicating and sensing at the same time
Figure 4. 6G is expected to combine integrated communication with sensing, which could help detect objects and people.
One of 6G’s most remarkable capabilities will be Integrated Sensing and Communication (ISAC), which will enable radio signals to sense the environment and coexist with or use the same radio signals used for data transmission. As Figure 4 shows, networks could detect, identify, or track objects by detecting motion and even micro-gestures. In effect, the network becomes a massive, distributed sensor. AI plays a vital role here, instantly interpreting the sensing data to power applications such as predictive maintenance, where machinery can be monitored without additional sensors, or improve safety systems in industrial environments and thereby prevent accidents.
Sensing motion
The combination of AI and 6G will also have a huge impact on mobility, particularly in the realm of autonomous vehicles. Self-driving cars, drones, and delivery robots will benefit from the combination of data from various modalities and/or sources. This level of integration will leverage the ultra-low latencies of 6G to enable real-time route adjustments to avoid congestion, as well as split-second responses to hazards, such as emergency braking to prevent collisions.
With 6G’s ultra-low latency, high reliability, and vast capacity, these decision-making processes can be made safe, scalable, and energy-efficient, paving the way for smarter cities and more efficient systems.
AI on the device
Another defining shift with 6G will be the move from applications that primarily download content to those that create, stream, and collaborate in real life. This uplink-heavy traffic requires more intelligence to reside on the device itself, something that can be achieved using agentic AI. Placing AI agents directly on devices allows for personalized, context-aware services without sending all data to the cloud, protecting privacy while reducing network load and power consumption.
By enabling devices to transmit features that are most relevant to a given task, such as the meaning of a message rather than raw data, communication becomes more efficient. When this is combined with AI, it allows users to use responsive, high-quality services without overwhelming the network.
Implications for engineers
For engineers, the arrival of AI-native 6G impacts the design process at various levels. Proprietary and standardized AI-based techniques must be embedded directly into the communication system. Doing so enables operators and network vendors to include their own products and services that can make intelligent decisions. End-user products will use AI to further tailor and differentiate products, thus extending the baseline 6G radio to commercial and end-user needs. Hardware must be AI-aware, designed with the power and efficiency to handle complex on-device workloads. AI models will need to be managed throughout their lifecycle, trained, and updated to match changing network concerns and user needs. Network resources such as spectrum, processing power, and storage will need to be allocated in real time based on AI predictions.
These changes will result in networks that are not only faster and more reliable than those in use today. They will be self-optimizing, adaptive, and capable of supporting entirely new categories of services.
The shift to AI-native networks will require close cooperation between research bodies, industry leaders, and global standard organizations. Efforts within 3GPP are already establishing standardized data-collection interfaces necessary for AI model creation, defining the necessary mechanisms to exchange and manage AI models to adapt to changing conditions while maintaining the desired performance, protocols for distributed AI processing, and embedded security and privacy protections from the outset.
AI as the brain of 6G
By the time 6G reaches commercial deployment in the early 2030s, AI will act as the brain and nervous system in key parts of the network, constantly learning from the environment, making real-time decisions, and optimizing the network with little human intervention. For engineers, this represents an opportunity to design networks that can “think for themselves” and adapt instantly.
The convergence of AI and 6G is set to bring mobile connectivity to the next level, enabling safer transportation, smarter industries, and richer experiences. The journey toward that future is already underway. It’s time to create intelligent networks that will make this a reality.