The Critical Missing Piece of Autonomous Vehicles

Today, billions are being invested into autonomous vehicles with the view that in the not too distant future, robo-taxis, delivery vehicles and self-driving public transport will dominate the roads of major cities around the world. Large companies in the US and China are developing cutting edge software, capable of rapidly analysing data and allowing the vehicle to travel autonomously to the highest degree of safety.

There is one key problem

The vehicle architecture present in today's car was not made for autonomous driving. Asynchronous communication protocols, complex setups, the issue of prioritisation form a barrier, beyond which the industry is limited in rolling out the most efficient and safe technology for autonomous driving.

The Obstacle

The problem that we find is the adoption of disconnected systems and asynchronous communication architecture in the vehicle. This means an array of control systems are set up in a complex arrangement, all directed towards the main onboard computer that is tasked with rapidly trying to make key decisions using a system of prioritisation. To make this task even more difficult, the signals sent often include large chunks of information, a lot of which is duplicated, meaning the share of “useful information” is rather small. 

To no surprise, this requires ever increasing bandwidth capability and the demand for more advanced chips with greater computational power to be able to process information faster and make those critical decisions. In turn, more computational power passing more information across wires requires more electrical energy and more material resources, which ultimately adds to the manufacturing cost of the vehicle and greater energy consumption. 

On top of that, the signals received by the onboard computer, passed from various electronic control units (ECU’s) come at different time intervals, so it’s majorly difficult for the system to accurately understand and pass critical decisions to the actuators. In ADAS level 5, this could turn into a fatal outcome, that’s a key danger of asynchronous communication systems in the vehicle.

Safety and margin of error comparison for Async vs Sync control systems

Synchronous Systems

In a synchronous system, all the ECU’s work in real-time, via a Time Triggered Protocol (TTP). In our case, we call it ‘hard’ real-time, because of the fixed reaction time of 1ms around the network and 10 microseconds in the ASG Core. In this system, all ECU’s are arranged in a single-ring architecture and at any one time, they know exactly what is going on - this is vital!

Information is passed into “the loop” of vehicle communication and in real-time, all nodes know what is happening. To improve the efficiency further, the ASG system transmits only the changes in information as opposed to whole blocks, meaning less strain on computational power and wiring, improving speed and reliability. For manufacturers, this also means no need for high bandwidth and expensive wiring harnesses that add weight and waste energy. 

Additionally, the onboard computer does not need to prioritise and process information asynchronously since all information is passed simultaneously and uses a distributed computing network for vehicle controls. This means that the whole E/E architecture works as a single organism, never conflicting between systems. Without diving into the technical details here, the effect of this is a much more efficient system that allows ADAS level 5 to operate within key safety barriers. 

Supporting the Software Defined Vehicle

When building innovative software, for the challenges of autopilot, connectivity and cybersecurity, it is often limited by the hardware in the vehicle. This factor places constraints on manufacturers to “update” the vehicles to the latest technology (it’s similar to having an old phone and trying to install a new operating system on it).

Therefore, we shouldn’t overlook the key hardware innovation that is absolutely necessary to solve the challenges of today’s global automotive market. Using outdated asynchronous communication protocols like the CAN, LIN and Ethernet will not solve the key challenge of fully autonomous driving reliably. Even the newest Ethernet-based vehicle control platforms cannot totally eliminate the consequences brought about by the delays at the network junctions, negatively affecting the vehicle motion control system stability.

Only a truly synchronous architecture can reliably update the vector of movement for a vehicle and log all events (vectors) in a time-based data logger, important for tracing autonomous vehicle performance.