Automotive edge computing: Real-time intelligence at the vehicle edge

Vehicles capture mountains of data from LiDAR devices, GPS units, IoT sensors, cameras and so on. Once analyzed, that data powers capabilities like hazard detection, predictive maintenance, vehicle-to-everything (V2X) communication and advanced driver-assistance systems (ADAS).

As soon as a vehicle moves, latency becomes a problem. Sending data to the cloud for analysis slows decision-making and creates safety issues. Bandwidth constraints further limit data transmission, and vehicles often operate in areas without 5G connectivity.

Simply put, edge computing means processing data closer to its source, without sending it to a distant server (as you would if you relied exclusively on cloud computing). Broadly speaking, there are two types of automotive edge computing:

• In-vehicle edge computing for safety-critical tasks like collision avoidance and emergency braking.
• Off-vehicle near-edge computing for in-vehicle infotainment and real-time traffic management.

Automotive software solutions can also benefit from multi-access edge computing (MEC), or mobile edge computing. It’s a network architecture that brings cloud services to the edge of the cellular network, lowering latency to less than 20 ms.

In the automotive industry, MEC can be used for:

• In-vehicle entertainment (IVE)
• Intersection movement assist (IMA)
• Vehicle platooning
• Automated valet parking (AVP)

Edge computing automotive use cases speed up safety-critical tasks, streamline ADAS, reduce costs and improve data privacy and security.

Real-time decision-making for safety-critical systems

Edge computing can bring latency below 10 ms. (For comparison, falling back on cloud computing means making do with latency 50 to 200+ ms.) These milliseconds matter when a system needs to decide quickly (e.g., to avoid a collision).

Reduced latency for ADAS and autonomous driving

Similarly, ADAS and autonomous driving capabilities have to rely on analytics as close to real-time as possible. Edge computing cuts latency, enabling quicker decision-making and reducing safety risks.

According to McKinsey, both end-to-end (E2E) models and in-cabin safety systems are predominantly deployed using edge computing in automotive.

Bandwidth optimization and cost efficiency

It makes more sense to offload non-critical or complex computing tasks to the cloud. That said, costs mount up if you have to transfer every data point to the cloud. Also, bandwidth isn’t infinite.

Edge computing ensures only key data is sent, to optimize bandwidth use. It also reduces data plan costs by minimizing data transfer.

Upgraded data privacy and security

Vehicle manufacturers often face data privacy challenges, leading to consumer concerns about sharing vehicle data with OEMs. Edge computing helps address customer worries by keeping data locally.

Improved reliability in low-connectivity environments

Constant 5G access is often unavailable in rural areas. Even in well-connected locations, tunnels can disrupt connectivity. Edge computing allows vehicles to monitor conditions and run analytics despite interruptions.

While edge computing systems for automotive use cases process data onboard the vehicle, they also interact with roadside nodes and transmit select data points to the cloud using an event-driven architecture (EDA).

In-vehicle edge layer

This is the built-in hardware and software that processes data locally. Think LiDAR sensors, IoT sensors and so on. However, complex edge AI computations require more advanced semiconductor types to balance energy efficiency with computation speed.

Onboard edge nodes 

Some edge computations can be pushed to onboard applications and devices. For example, navigation and infotainment applications enhance the driving experience using data from in-vehicle sensors.

Roadside edge nodes (V2X)

Edge computing also enables vehicles to communicate with external edge nodes, such as:

• Other vehicles for coordination to improve traffic fluidity and safety (vehicle-to-vehicle or V2V) 
• Traffic systems like road sensors and traffic lights (vehicle-to-infrastructure or V2I)
• Pedestrian devices like smartphones and wearables (vehicle-to-pedestrian or V2P)
• Cellular networks for broader communication needs (vehicle-to-network or V2N)

Edge cloud integration 

Edge computing is no silver bullet. It’s energy-intensive and cannot store data long-term. So, cloud services still have their place in edge computing systems, handling additional processing (e.g., big data analytics), OTA updates and data storage.

5G and MEC enablement

Edge computing systems can use cellular (LTE/5G) or Wi-Fi networks to connect to far-edge and multi-access edge (MEC) services. When needed, data travels to the network's edge to leverage cloud computing for more complex computations.