Why Edge Computing is key for the automotive industry.
What is edge computing and what are the challenges in making it happen efficiently?
WHAT IS EDGE?
In the IoT jargon, ‘edge’ refers to computing located close to the data source, e.g. a sensor or device. Edge computing is mostly used to store, process and send data to the cloud. However, the computing capabilities and power supply of such edge devices are often limited. At the moment, there are new upcoming technologies enabling edge computing to process more data, run analytics and store efficiently.
The challenge in edge computing is to bring Machine Learning and Artificial Intelligence (AI) from the cloud to the devices at the edge, or to even bring this to the actual sensors at the very edge. The specific challenge is to process data accurately and efficiently in environments with (far) less computing power and storage capacity. The devices or sensors — i.e. ‘the edge’ — can range from simple sensors (for instance temperature or speed measurements) to highly data-intensive sensors such as cameras or LIDAR.
Edge computing is not a new technology, but It is now starting to realise its true potential in the real-time data transfer from device to cloud and in real-time data processing at the device. As the number and frequencies of sensor signals increases, smarter algorithms will be required to efficiently process this explosion of sensor data. In use cases where low latency cannot be tolerated or where 100% connectivity should be guaranteed, bringing computing to the edge or the remote sensor is the solution.
Edge Computing in the automotive industry is required to cope with the exponential growth of data in (partly) autonomous vehicles. As cars generate significantly more data every day, it is becoming a big challenge to process all that sensor data efficiently in the car and to transfer parts of that data to the cloud. In addition to that, safety related functions need to be available all the time and cannot rely for their functioning on wireless connectivity. For such needs, intelligent efficient edge computing comes to rescue.
It is important to understand that 5G will not be available everywhere as it requires a roll-out of 4X more antennas. This is too capital intensive to warrant 100% coverage. In addition, like any mobile network today, 5G will not have a 100% uptime either. Many applications in the car are safety related or real-time and can therefore not fully rely on a network. Hence these applications will need to operate autonomously inside the car.
Consider this example: when an Autonomous Vehicle on a highway requires to break in an emergency or in a sudden dangerous situation. The application must identify the hazard and react by applying the brakes, and all within milliseconds. The application “emergency braking” cannot afford the 100ms transmission over a cellular network would take. As a matter of fact, the computing by in-car chips should be done in 10’s of ms. In case of long, combined delays of computational and transmission latencies, lives will be at risk.
Safety critical applications running in cars, can only be reliable when the data is processed accurately near the device and that’s why edge computing will play a vital role, as accurate and fast decisions must be made quickly and close to the source. Hence either from a safety point of view or from an efficiency point of view, computing (and decision making) will be done more and more at the edge. The challenge is how to do this quick and accurately given the constrained computing capacities. This is what we at Teraki will deep dive further into in our next blogs.