Towards a quality SdV (& EV) experience

From Code to Road? Check Engine Light.

In the midst of a difficult year, automakers are well set on the path to differentiate EV brands based on autonomous driving features, connected applications and (now at least multi-display) digital cockpits. To do so, they have converged on high-performance computers inside vehicles so software platforms (Operating Systems) can control a large part of the driving and in-cabin experience.?

And instead of acting as systems integrators that outsource component production, car makers are investing heavily to take more ownership of SdV (software-defined vehicle) platforms.?A winning combination of systems on a chip, advanced sensors, actuators, saftey-certified hypervisors, real-time operating systems, middleware and intelligent vehicle applications spanning assisted driving, digital cockpit etc. It's a lot of stuff--but when available will be the "digital chassis" and virtual driver that sits behind a brand.

Recent market entrants such as “new energy” and Asian EV makers have shown nimbleness (without the burden of an internal combustion engine past) to show off the benefits of a vertically integrated business model i.e. stay cost efficient and control one's destiny.?

While the rationale behind SdV platforms is rooted in aspirational revenue and growth a large part of the motivation to replace a hardware-led development model with?software design allows car makers to get ahead of an expected rise in vehicle complexity (see image below). To bend the cost curve is crucial because every part of the business is going to be affected from budgeting, R&D, operations, manufacturing, dealership and customer service. For example, the service lane experience and dealership business models are evolving as car makers gain influence in diagnosing software-related issues and applying over-the-air updates....saving time, effort and headache for the customer.

Quality at Every Turn

A significant challenge that needs to be discussed more will be sustaining quality from the start of the buying experience, before start of production, during customer support, part of software development and across all parties involved in the product life cycle.?While quality has always been part of the job and a priority for auto markers---this time it's different and because it's software.

Take for example a turn signal function. It takes different teams pushing lots of code into a shared (software) build system to make sure that a seemingly mundane turn signal works, all the time. Today a turn signal may interface with automatic braking, lane keep assistance and collision warning. A software bug will not only be harder to detect and isolate to a supplier but there will be more dependencies across a centralized architecture, with more open source contributors, in comparison to self-contained and supplier proprietary embedded code in Electronic Control Units.

When everything is controlled by a relatively new software platform the success of an EV model will be at stake. What's the impact of a glitch or many of them in a row? That will depend on the philosophy of the car maker and expectation of the customer. But, without going into statistics one can start to appreciate the gravity of the situation given the rise in recalls, warranty complaints, customer satisfaction reports and National Highway Saftey Administration interventions.

Come glitch, the version number of a Car’s OS and installed applications will be under regulatory scrutiny and the "SdV Platform" must be devoid (to a large degree) of instability, unexpected behaviors?and issues that will materially impact the brand.?

I Want a Quality SDV (&EV) Experience?

The agreed to customer expectation is a customizable (very digital) vehicle, tailored to a brand, meets locals geographic and regional requirements. All told, allowing auto makers to engage in selling, packaging and servicing a product that increases in value with updates.

I personally want a safe software-defined vehicle, that carries my precious cargo and is not built like a consumer electronics product.

When safety is concerned, quality is real and perceived by the customer, accountable by the car maker and a shared responsibility with all parties involved from regulators, R&D, engineering to suppliers and an expanding tech ecosystem.

Some guiding principles:

1) Quality programs should take a broad in scope systems and data-driven approach to achieve and sustain quality from inside the vehicle, to outside and in the cloud and related parts of the business.

2) While quality metrics are addressed by standards such as ISO26262 and SOTIF to guide the way, the weighted contribution of software regulation means fresh view to issue resolution, cost of quality, serviceability, stability of upgrades and safe feature velocity.?

3) Unlike hardware-led development, EV models will have a much higher degree of configurability (pre and post production) that cascade the impact of an issue. So version and variant management is crucial starting with vehicle intelligence. Whether that's an error in a line of code, security vulnerability, partial failure of a hardware element or sensor mis-calibration. For example, braking systems are now part of a vehicle dynamics domain controller and integrally linked with a vehicle's main operating system. An update to the braking logic should trigger? expansive quality checks (i.e. “type certification” in the EU & self-certification you in the US) because components are not as independent as before. When a software update is delivered to the braking logic there is change, and with change there will arise a (potential) population of issues that involve energy management, driving behavior or the instrument cluster. In many cases the issue will be temperamental and difficult to root cause because it occurs at a certain time, for a specific model, model year, software platform version or for a region.

4) The quality challenge is even greater for incumbents that have multiple brands, each with a different SdV platform or different milestone on the roadmap.?It just becomes an ambitious undertaking to take a company that traditionally bent steel to now fully organize around software. So moving to a feature-first mindset as the benchmark is going to be an ongoing effort in the change management and transitions.

5) Adopting more software updates that will predominantly be conducted over the air requires a layering of quality control frameworks and governance procedures to reduce risk of misconfigurations, model drift, introduction of error conditions and unknown security vulnerabilities. As SDVs become mainstream, there will be a need for multi-disciplinary frameworks that must be knitted together, where data is fused to get a precise picture of quality. For example, ann area of great public interest will be continuous type approval—as AI models are rolled, but encounter issues in the field.

Ship safety critical software continuously to end customers...? Let's plan well.

Contributors:

Philipp Wolf Jim Heaton Nitya Verma Bopaiah Mekerira