A revolution in the automotive industry is upon us. Fifteen years ago, the very thought of a commercial showing a self-driving car was science fiction. While companies like Waymo have performed wide-scale experiments with hundreds of cars in field trials, proving that autonomous vehicles are possible, the assumption was that it would take 15 to 20 years to become a mature technology. As time passes, people are starting to realize that this revolution will happen sooner rather than later. The investments that companies are making in autonomous vehicles are gigantici. Legislation is also under way to create the legal and regulatory framework for autonomous driving.
Still, the biggest barrier to commercially available autonomous vehicles is trust, or more accurately the mistrust humans have in machine-driven vehicles. According to the National Safety Council (NSC), there are over 100 fatal accidents per day in the U.Sii. This hardly ever makes headlines. However, a single Tesla or Uber accident makes the front page, and everyone gets nervous. People ignore the small print in the National Highway Traffic Safety Administration (NHTSA) accident report on Tesla that states that the first year that Tesla AutoPilot (auto-steering) was introduced, the number of accidents with Tesla vehicles dropped by 40%! This was even before the additional improvements Tesla did in this system, after this feature became commercially available.
The U.S. Department of Transportation estimates that autonomous vehicles could reduce traffic fatalities by 94% by eliminating accidents due to human error.
To truly gain people’s trust in machine-driven cars, autonomous car systems have to be extremely reliable. It starts with the brain of the system, which is the SoC/GPU running Artificial Intelligence (AI) programs to accurately analyze the 360-degree view around the car and define the right driving policy. It extends to the eyes and ears of the car, which are the cameras and other sensors. These cameras and sensors need to work reliably in every extreme condition whether at night, in bad weather, or at extremely high temperatures.
The next big challenge is to build the nervous system that connects everything. This is the In-Vehicle Network (IVN) that carries data reliably between the different nodes of the network and provides the redundancy that delivers a system with no failures.
In addition, with the Autonomous-driving Level moving from Level 1-2 to Level 3-4 – with Level 5 on the horizon– the speed of the IVN and its complexity are growing in parallel, as shown in Figure 1.
Figure 1. The Path Towards Full Autonomy
While Level 1-2 vehicles with Advanced-Driver Assistance Systems (ADAS) include one to three low-resolution cameras together with a basic controller that makes minimal decisions, mainly for braking and velocity, the system for a Level 4-5 vehicle is more like a data center on wheels. Level 4-5 systems will have anywhere from 10 to more than 20 cameras and sensors, plus a few centralized computing systems that will be responsible for everything including vision analytics, driving policy, high-bandwidth telematics,advanced storage systems. All these nodes need to be connected with a reliable network that can provide the speed, quality-of-service (QoS) and security levels that are expected in autonomous vehicles.
Over the last 30 years, many protocols and solutions have been developed for the IVN based on the requirements of Level 1-2 cars. A few new protocols, which are mostly proprietary, were introduced in the last few years with the goal of providing a solution for future Level 3-4-5 cars. Once the market identified the requirements and complexity of the network for future cars, many OEMs and Tier-1s recognized that Ethernet had the highest potential to meet all the needs for these networks.