Reasons Why Connected AVs will be Software-Defined

Microchip maker Nvidia and car manufacturer Mercedes are planning to roll out software-defined autonomous vehicles by 2024.

They claim that the vehicles will be able to receive over-the-air (OTA) software updates to ensure that they are able to benefit from the latest autonomous features. However, there has been talk about OTA updates for quite a while now and so this raises the question about why connected and autonomous vehicles (CAVs) need to be ‘software-defined’.

Defining software-defined CAVs

In fact, what does this term really mean in practice? Danny Shapiro, senior director of automotive at Nvidia explains: “Currently, the vehicle’s software functions are powered by dozens of electronic control units, known as ECUs, that are distributed throughout the car. Each is specialized. One unit controls windows and one the door locks, for example, and others control power steering and braking. This distributed architecture makes it difficult to update or add new features to the car once it has left the factory. While many ECUs have firmware, they are generally fixed function.”

To address these issues, he says Nvidia is centralizing and unifying computing in the car with a single platform. This approach simplifies and facilitates the ability to integrate and update advanced software features as they are developed. An example is how people update their smartphone software, which is updated periodically either via a mobile connection, of via Wi-Fi. In essence software-defined vehicles will be able to do much the same, while adding new features and capabilities that, he says, “may not have been envisioned when the car was originally sold.”

Hardware-defined vehicles

Suhas Gurumurthy, industry analyst of connected cars at Frost and Sullivan, argues that if they are is such a thing as a software-defined autonomous vehicle, “there could also be a hardware-defined one too,” he said. He defines a hardware-defined autonomous vehicle: “In an unusual case, a hardware-defined autonomous vehicle will be one that has a predefined collection of features/functions, i.e. hardware restricts the functionality of the software.”

“In this context, therefore, we can define an autonomous vehicle to be software-defined when its features and functions are primarily software-activated. In most vehicles today, hardware and software components are coupled together and this forbids the vehicle features or functions to be customized or personalized as per consumer preferences.”

Decoupling software and hardware

This means, to be software-defined, he believes the software and hardware within CAVs much be decoupled to allow ECUs to communicate with each other at the software level. This permits new features and functions. “As an example of such a development, VW works closely with partners such as Continental to build a range of software-defined vehicles on its I.D platform,” he discloses.

To a certain extent he also finds that “today’s connected vehicles and autonomous vehicles are in some sense software-defined but the degree or level and complexity of the definition of software varies.” Advanced driver assistance (ADAS) systems are software-defined, which he explains which can be “remotely controlled through software-over-the-air (SOTA) or firmware-over-the-air (FOTA) updates.”

Shapiro adds that software-defined vehicles have a centralized compute architecture that is optimized for artificial intelligence powered functions. These functions include autonomous driving and intelligent cockpit. “As an example, the upcoming Mercedes-Benz fleet will be capable of driving regular routes from address to address autonomously and Level 4 automated parking, in addition to countless safety and convenience applications,” he explains.

“The hallmark of software-defined vehicles, however, is the ability to continuously add to and update the features throughout the life of the car. Teams of software engineers and AI experts will be supporting these vehicles, continually releasing new software to make them better and better.”

Autonomous capabilities

Varun Krishna Murthy, senior research analyst at Frost and Sullivan, argues that autonomous vehicles need to understand, analyze, decide and act with minimal or no supervision from a human driver, as they will operate in a constantly changing environment using sensors, processors and actuators to either make driving decisions, to alert a human driver to ensure that appropriate actions are taken.

The sensors, processors and actuators are governed by software. “Today’s vehicles featuring a few driver assist systems make decisions locally at a feature level (e.g. an automated emergency braking ECU activates the brakes to avoid imminent collisions).”

“However, to attain higher levels autonomy the industry is working towards bringing multiple features (ECUs) together in a central system, run by a software that controls all tasks of driving, which can be broadly called as software-defined vehicle.”

An example of this is the 2019 Audi A8, which features SAE Level 3 autonomous driving is a good example that Murthy says, “integrated a few features to offer autonomous driving on highways”. His colleague, Gurumurthy also confirms that most vehicles in the North American and European regions can update software inside CAV infotainment systems OTA.

Premium brand focus

The likes of BMW, Mercedes, Tesla, etc. already provide FOTA upgrades for multiple systems. He says that the premium branded automakers have also started offering what he refers to as ‘FoD’ [features/functions-on-demand] on their vehicles for connected services, and to support functional upgrades.

However, this kind of offering has yet to appear in vehicles made by the mass market brands. A few of the premium brands are also in the same situation at the moment. He therefore comments: “There are a vast majority of vehicles still dependent on dealership updates as these vehicles are from previous generation platforms designed to be partly software-defined.”

Mass market limitations

Shapiro confirms that over-the-air capabilities are “extremely limited”, meaning that vehicles with traditional architecture can only “update specific ECUs, like infotainment, and even then, the range of updates is incredibly narrow. Most vehicles also have extremely limited processing power and extremely limited programmability.” He argues that truly software-defined vehicles are highly programmable and they have the ability to unlock updates for the entire vehicle. So, in his view, software-defined CAVs have nearly limitless possibilities.

He explains: “This software-defined architecture is also opening up new business models for automakers. Features and capabilities can be purchased at any time during the life of the vehicle. Second and third owners can unlock capabilities that the prior owner may not have.”

Software-defined future

So, what is next? Well, he thinks the industry needs to “implement centralized compute and to develop architecturally coherent and programmable fleets”. He says this will allow companies with a growing installed base to offer “software upgradeable applications and subscription services for the entire life of their upcoming fleet.”

Murthy predicts that CAVs will be developed over the next few years in a similar way to smartphones. “Like other smart and connected devices, vehicles will be able to personalize comfort and convenience features for vehicle occupants thereby making the feel at home”, he comments.

He adds that the ability to build software-defined CAVs will enable OEMs and Tier 1 supplier to offer new features, applications and update existing ones without replacing the hardware or visiting the service centers. This will permit them to optimize cost, which offering efficient and improved product life to their end customers. Therefore, the future of CAVs is likely to be software-defined.