Access Official Technical PDF

Share Summary:LinkedInEmail

Edge computing for IoV—the Internet of Vehicles—offers a way to process vehicle data closer to where it is generated and used. Intelligent vehicles combine cameras, radar, LiDAR, sonar and satellite navigation with onboard computing and communication systems. Their data can support safer driving and better-informed transport operations, but collecting, transmitting and processing it creates substantial demands on vehicle hardware and communications networks.

Placing storage and computing resources at roadside units (RSUs) or wireless base stations can help address some of these demands. Rather than relying exclusively on a vehicle’s onboard systems or a distant cloud data center, an edge information system (EIS) can acquire, aggregate and process data near the road. Its practical value depends on assigning each task to the right part of the system.

Why intelligent vehicles need computing at the edge

Vehicle sensors produce large volumes of data, and tasks such as image-based perception, sensor fusion and localization can require significant computing capacity. Increasing onboard processing may help, but available compute, storage and power are constrained. Power-hungry processors can also create thermal demands that affect vehicle efficiency.

Cloud computing provides access to substantial storage and processing resources, making it useful for activities such as training machine-learning models on large datasets and retaining historical information. However, sending data to a remote cloud can introduce communication overhead and delay. Network conditions, available bandwidth and traffic congestion can also affect the time needed to exchange information. For applications that depend on timely results, a remote service may not be the appropriate place for every processing task.

Edge computing places servers closer to vehicles, typically at base stations or RSUs. These servers can communicate with vehicles over vehicle-to-infrastructure (V2I) links and use local data to provide nearby computation or content. This can reduce the need to send all sensor data to the cloud and support localized applications, while keeping onboard processing available for the most time-sensitive vehicle functions.

How an edge information system fits together

An EIS combines edge caching, mobile edge computing and edge AI. Caching stores selected content—such as maps or road information—near where it is likely to be requested. Edge computing provides processing resources for tasks that are computationally demanding or sensitive to delay. Edge AI supports the use of machine-learning models on edge servers and devices.

In a vehicle-as-a-client model, vehicles access edge resources. An edge server might collect information from vehicles passing through its coverage area and process it for local uses, including monitoring traffic or updating map data. In a vehicle-as-a-server model, vehicles can provide services or share information with passengers, third parties or other vehicles. This can support cooperative applications, although the usefulness of vehicle-provided resources depends on connectivity and mobility.

The system is not a replacement for onboard computing or cloud services. A practical division of work is:

  • Onboard systems: Handle highly time-sensitive vehicle control and can preprocess sensor data before communication.
  • Edge servers: Support latency-sensitive, computation-intensive applications and aggregate local information, such as data relevant to a local high-definition map.
  • Cloud platforms: Support non-real-time aggregation, storage of historical data and training of models using large datasets.

Edge computing for IoV applications

Perception and cooperative awareness

Perception systems interpret sensor inputs to identify objects, lanes and other features of the driving environment. Individual sensors have limitations, and combining data from multiple sensors can increase onboard processing demands. Edge resources may help by handling selected computation-intensive subtasks, including parts of image processing or sensor fusion. When vehicles and infrastructure share information, cooperative perception can also extend awareness beyond the view available to a single vehicle. Such systems rely on communication and coordination, so they do not remove the need for onboard perception and decision-making.

High-definition mapping

High-definition maps represent road features in detail and can support precise localization. Creating and maintaining them requires repeated data acquisition, processing and updates as road conditions change. Map files can also be large, making distribution and storage challenging. Edge servers can help aggregate mapping data from vehicles, process local updates and distribute relevant map content in the area where it is needed. Caching can further reduce repeated retrieval of content that is requested by multiple vehicles.

Localization and SLAM

Simultaneous localization and mapping (SLAM) involves estimating a vehicle’s location while building a map of its surroundings. It can help vehicles adapt to changing environments, but the associated computations can be demanding. Edge servers may take on selected subtasks, and cooperation between vehicles may contribute additional observations. The system still has to account for communication delay and the need for timely results, particularly when processing supports driving functions.

Design considerations for integrators and operators

Deploying edge services for connected vehicles involves more than selecting a server. Base stations can offer wider coverage and mobility management, while RSUs are closer to the road and may support lower-latency communication. The choice of location and hardware should reflect the application, coverage needs and available communication links.

Resource allocation is another central challenge. Edge servers have finite computing and storage capacity, while vehicle connections and content requests vary over time and location. Caching policies must account for which information is useful in a given area and when it becomes outdated. Computing services must decide which workloads to offload without making applications overly dependent on network availability.

For system integrators and operations leaders, this points to a task-by-task design approach: identify the data needed, the acceptable processing delay, where the data is most useful and what must continue to run onboard if a connection is unavailable. Local processing can be valuable, but its role should be defined alongside cloud and vehicle capabilities.

Conclusion

Edge computing for IoV provides an intermediate layer between vehicle hardware and remote cloud platforms. By combining local storage, computing and AI, an EIS can support applications such as perception, high-definition mapping and localization while reducing unnecessary data movement. The practical implication is not to shift all vehicle processing to the edge, but to distribute work according to latency, power, data locality and resource constraints. A well-designed architecture uses onboard systems for immediate decisions, edge resources for local and time-sensitive workloads, and the cloud for broader aggregation and model training.

Access Official Technical PDF

About The Author