Bertrand Boisseau
on 16 November 2023
Implementing edge computing for V2X use cases in automotive
Vehicles are becoming more and more like mobile data centres. On average, a modern vehicle contains over 60 sensors that monitor various aspects of the vehicle, generating an immense amount of data that is processed on the go. This transformation is creating an unprecedented set of challenges for OEMs.
Edge computing is a new paradigm that is changing how data is processed in these types of environments. It involves decentralising data processing and analysis, bringing computational capabilities closer to where the data is generated. Clouds optimised for edge processing can address the problems posed by the vast volumes of data generated by autonomous systems, ensuring consistent performance regardless of the vehicle’s position.
This blog will dive into edge computing and study how clouds that are optimised to process data at the edge can address automotive challenges.
What is edge computing?
At its core, edge computing is a distributed computing model that processes data at or near the source of data generation. This means that rather than sending all the data to a remote cloud server for processing, analysis and decision-making, the computations are done as close as possible to where the sensors or operations are located. This provides three main benefits:
- Reduced latency – Providing compute in proximity to the data source ensures that the latency is reduced to a minimum, which guarantees faster response times and enables real-time processing.
- Data pre-processing – Only the required, pre-processed data is transmitted to the centralised cloud servers. Fewer data transfers mean lower bandwidth requirements, enabling massive cost savings. And you benefit from better privacy as the stored data is already curated and protected.
- Improved scalability and network resilience – The distributed approach makes it easier to scale and enhances network resiliency. You can deploy processing capacity wherever it is needed and with the necessary redundancy based on local requirements. For example, if you need to use additional sensors in a specific context, you can deploy localised clouds for high-performance processing near the sensors.
Relevance of edge computing in automotive
These benefits are uniquely suited to addressing the new challenges facing the automotive industry, especially when it comes to software and data processing. Vehicles are moving objects by definition. Their location is constantly changing, which leads to difficulties in processing the large amounts of data they generate. This is especially true for autonomous vehicles, which depend on real-time data to operate. Let’s explore the different challenges the industry faces and how they relate to edge computing.
Autonomous driving and vehicle-to-everything (V2X)
With Autonomous Driving (AD) and emerging regulations come new “vehicle-to-everything” (V2X) use cases. Authorities in multiple countries are considering making V2X communications mandatory in future vehicles. For instance, the United States National Highway Traffic Safety Administration (NHTSA) is considering specific V2X technologies for collision avoidance systems.
V2X involves all the possible interactions between vehicles and the surrounding environment, like the local infrastructure (V2I), the surrounding vehicles (V2V), and so on. For features related to accident information sharing for example, it is important for the vehicles to communicate with the road infrastructure itself, even in areas with very limited network coverage. These use cases are very challenging today with a traditional cloud infrastructure.
The real-time requirement is high and critical for the safety of the passengers. The communication between multiple vehicles and an infrastructure point is subject to intermittent connectivity problems. And V2X often involves exchanging of sensitive information, so ensuring that this data remains secure and private is crucial. Edge computing delivers the resilience, security and minimal latency necessary to satisfy these requirements.
Far-edge computing for sensor fusion
AD vehicles are known for generating enormous amounts of data collected through numerous sensors. The volume of data is nearly impossible to transmit to a central cloud for processing. Not only would the duration of the upload be too long before obtaining a processed response, the cost of cloud processing and data transfer would also be too high.
This is why all OEMs pre-process the data from within the vehicle first, before sending only relevant information to the cloud. That way, most of the urgent decisions can be done onboard, using pre-trained algorithms and ensuring that the response is taken below a certain time threshold.
That being said, the raw processing power and the low-latency requirements are still difficult to meet due to the quantity of data generated for sensor fusion. Sensor fusion involves gathering data from different sensors within a vehicle, such as cameras, radars, lidars, ultrasonic sensors, and other detection mechanisms. By combining the information from these diverse sensors, the system can obtain a more comprehensive view of the vehicle’s surroundings. Pre-processing such huge volumes of combined data requires an edge-computing appliance which is as close as possible to the sensors, but far from the central cloud infrastructure – what we call “far-edge cloud”.
Near-edge computing within factories
When it comes to vehicle manufacturing, quality checks often require the analysis of large volumes of precise 3D data. This processing demands substantial 3D capabilities. Traditional cloud-based solutions pose networking challenges as well as security concerns.
Sending proprietary data off-site for analysis, which could include confidential details on a company’s manufacturing approach, should be avoided as much as possible. In this case, the vehicle manufacturer would benefit from edge clouds that reside within the factories and are directly integrated with the central cloud infrastructure (“near-edge cloud”). The data processing is done in parallel within the facilities and protected by the factory premises.
MicroCloud, the Canonical solution for edge computing
Edge computing architectures deploy cloud capabilities to the edges, enabling computing and storing features in various distributed locations. There is a growing demand for simple edge cloud solutions due to the rising volume of data generated at the edge. The deployment of cloud capabilities at the edge comes with challenges such as orchestration, security and maintenance.
To solve these challenges, Canonical recently announced MicroCloud. MicroCloud is a low-touch cloud solution designed for scalable clusters and edge deployments. Delivering extensive automation, this solution enables you to deploy your edge cloud with a single command, and significantly simplifies ongoing maintenance.
MicroCloud offers several distinct advantages over traditional cloud-based processing, by enhancing and simplifying the deployment and operations of clouds in remote locations. This new Canonical product enables the deployment of a lightweight but scalable cloud which is perfectly suited to edge use cases that require security and efficiency.
Pushing further with 5G and AI
It’s possible to combine edge computing with emerging technologies in order to enable even more capabilities. For example, edge computing architectures are invaluable for 5G network slicing.
5G slices can be configured in order to guarantee low-latency in some areas and high-speed communication in others. The networking architecture can then be tailored and optimised according to the local needs of the overall infrastructure. This will be instrumental for the rise of remote-controlled autonomous vehicles.
AI is also deeply intertwined with edge computing, as algorithms power the decision-making processes within autonomous vehicles. OEMs need to ensure that the trained AI models receive the right data required to ensure safety and efficiency of the vehicles and their occupants. The full stack needs to be defined accordingly so that the models can use the power and latency of edge clouds in the best possible way.
Indeed, in order to take full advantage of the capabilities offered by edge clouds, there needs to be an optimised software (and hardware) stack. The distributed nature of edge computing architectures makes them a natural deployment paradigm for AI systems.
Conclusion
As the automotive industry continues to evolve towards software, connectivity and autonomy, new challenges arise. Edge computing offers solutions that enhance performance and user experience while addressing security risks and data privacy concerns.
As we move into 2024, expect to see strong investments and integrations of edge computing technologies in automotive factories, V2X applications and autonomous driving features.
Make sure you use optimised data-driven solutions and learn all that there is to know about MicroCloud, Canonical’s low-touch private cloud optimised for edge use cases. MicroCloud is ideal to drive innovation in automotive with safer, smarter, and connected vehicles, factories and infrastructures.
