The autonomous vehicle of the future — from proof of concept to real-world application

The automotive industry has come a long way in its development. Now, the technology to build self-driving and autonomous vehicles is no longer just science fiction, but is gradually being used in the real world. If you imagine a future world full of space-age flying machines, each of which can freely traverse the earth with carefully planned precision, I admit that this hypothesis is really tempting and exciting. However, the only thing I hope will not happen is that people in the future will be forced to wear shiny silver jumpsuits like in the movies - I don’t know who came up with this idea.

Of course, real-world situations are certainly more complex. Likewise, the technical environment facing developers of future autonomous vehicles can be complex, especially during the proof-of-concept phase. In addition to a unique and demanding development environment, engineers face a variety of challenges with custom on-premise and cloud applications—all of which must communicate with each other in real time, a task that requires a highly autonomous Industrial Internet of Things (IIoT) system.

Drivers, start your engines!

Currently, many automakers are actively involved in autonomous vehicle (AV) projects. As developers gradually enter the proof-of-concept stage, they will more or less encounter some unexpected obstacles in the practice process.

First, a self-driving car’s system must be able to do three main things: sense the environment, process data about the environment, and act based on the information from the environment. Essentially, this is a cycle that needs to be repeated over and over again. But the amount of data generated in this process and the speed at which it must be processed can quickly become overwhelming.

Common challenges in developing autonomous vehicles

We need to break the above down. When we look at an autonomous vehicle, it must have a sensor suite that can observe the environment (which can include simple driver assistance technologies as well as more complex highly automated or fully automated vehicle systems). The environmental sensor suite determines the fidelity of the data and how much data is collected from lidar sensors, radar sensors, actuators, and other input points. We call it sensor fusion or data fusion because it only works when all these components can share data with each other and agree on the accuracy of the conclusions.

Next, we need to think about the scenarios in which the system must use artificial intelligence (AI) to solve the problem. For example: How do I process this information? Should I turn left? Should I go straight? Should I turn right? What is happening in the environment?
In addition, the system needs to analyze different transient factors. For example, is it a person, a bicycle, or a car coming towards you? Then make decisions and response plans based on different situations. Of course, when the car takes action based on the environment, it will in turn change the environment, so the whole cycle will start all over again.

So the real challenge is the high level of interconnectivity: the quality of the system depends on the speed and quality of the data that can be captured and processed. Then, when external interconnections are added (such as connecting to cloud systems or connecting to other systems), they become part of the interconnected solution. The result is a complex distributed system with many components, all packaged very tightly together.

The concept of layered data bus

Massive scalability is a core prerequisite for every highly autonomous system. This is especially true in the field of self-driving cars. This is because there is a huge difference between a system that works under controlled test conditions and a system that is truly ready for market, and even the best team of developers can be blindsided by this complexity difference. If a system wants to reach the market and perform its functions - including all the media scrutiny and new application scenarios that the public demands, it usually adds a whole new layer of mission-critical requirements within the system, and no one has been able to take on this task well so far.

Layered Databus is a concept and term developed by the Industrial Internet Consortium (IIC), an organization that promotes and coordinates priorities and enabling technologies for the Industrial Internet. The goal of layered databus development is to enable development teams to identify different control planes or information planes in the system. In addition to full control over the environment, teams are also able to specify the quality of service (QoS), which determines how data must flow between application software in different application scenarios, including reliability, bandwidth, and latency.

This layered data bus concept allows developers to use the same standards throughout the ecosystem. It also allows developers to set individual conditions and individual rules for different parts of the system to manage data. All of this allows communication between different systems in a standardized way without having to add new protocols, gateways, or other bridges. Of course, the layered data bus also allows teams to set different conditions for data use so that the system is reliable and repeatable.

There is debate about where autonomous vehicles are at in the industry and when we will see Level 4 and Level 5 autonomous vehicles on the road. While the timing often varies depending on who you talk to, one thing developers agree on is that high levels of interconnectivity are a core element necessary to acquire and process data and address system complexity. A layered databus architecture provides standardized communication among these systems and gives developers the tools to bring driverless cars to market efficiently, quickly, and safely.