The days of automakers designing vehicles with a hardware-centric focus are over. Hardware features can feel underwhelming to the modern consumer and often become archaic at some point during the vehicle's lifespan. Software-defined vehicles (SDVs) are the new standard in the automotive industry, enabling original equipment manufacturers (OEMs) to securely add new features throughout vehicle ownership via over-the-air (OTA) updates. Offering new driver experiences years after a vehicle leaves the lot translates to increased customer satisfaction and long-term revenue streams.
To successfully transition to SDVs, automakers must account for upgradeability, flexibility and customizability. Lacking the required in-house software skills, the transition to an SDV architecture requires OEMs to foster relationships with semiconductors that can help integrate key enabling technologies -- embedded compute, middleware, hypervisors, applications, developer toolchains and the cloud.
Addressing SDV compute needs
Compute is the lifeblood of any software application. In the SDV context, OEMs should prioritize a few things in a potential silicon partner.
First and foremost, SDVs require more embedded computing resources than needed for the feature set at the time of purchase. The purpose of an SDV design is to ensure that a wide range of automotive applications can run properly in the future. This will carve out new revenue paths that legacy vehicles could never attain with their hardware-focused nature. Therefore, the semiconductor partner must offer excess compute capacity (up to 30% to 50%) for hassle-free OTA updates.
A heterogeneous approach to the SDV platform is also important. OEMs should select a semiconductor partner with a broad range of intellectual property (IP), including a central processing unit, graphics processing unit, digital-signal processing and neural network accelerator. Partners with a strong IP portfolio are better equipped to support future SDV applications and accommodate shifting consumer demand for electronics.
Embedded AI, a key innovation in the mobile and PC space, can be applied to SDV platforms. While cloud AI is perfectly capable of handling low-latency functions, mission-critical applications require embedded AI. For instance, embedded AI is essential for autonomous-vehicle perception, sensor fusion and motion planning. If you were to offload these functions to the cloud, you risk high latencies and, therefore, a dangerous driving experience. A hybrid approach is the best option, as it combines embedded AI's reliability with cloud AI's flexibility.
Middleware and hypervisors provide much-needed abstraction
Middleware serves as an abstraction layer between hardware within the SDV platform and consumer-facing applications. It promotes harmony between hardware and software development, which is necessary to establish a software-led design culture and improve multi-partner collaboration.
The traditional approach of adding a new electronic control unit (ECU) for every feature has made vehicle systems heavy and costly. As passenger vehicle production moves toward a software-defined architecture, automakers can simplify these systems by using fewer, more powerful controllers. Controller consolidation is more cost-effective and space-efficient than using multiple ECUs for the SDV platform.
And to maintain safety, it is important to separate critical safety functions from non-critical ones. Using technology such as a hypervisor helps achieve this by allowing shared resources while keeping safety intact. A hypervisor sends compute resources to domains based on demand and safety prioritization. This dynamic allocation of compute allows automakers to reduce costs and manufacture lighter vehicle models.
Bringing novel SDV applications to market quickly
To capitalize on the lucrative opportunities presented by SDVs, automakers must be able to develop new applications regularly and easily. As a baseline, the semiconductor selection process should prioritize partners with an extensive library of automotive applications that can be customized quickly. For example, an OEM should be able to quickly fine-tune an application to its branding. With resources already stretched thin, expediting app development/integration is essential. It allows carmakers to spend more time developing new, distinctive SDV features and less time on run-of-the-mill applications.
Fast time to market of new automotive applications is also supported by software toolchains. This software enables manufacturers to develop, test and verify SDV applications in a timely fashion.
Lastly, software development kits (SDKs) are key in SDV design. They simplify the process of using third-party applications to develop internally made features. Moreover, SDKs help ensure that engineering resources are not exhausted when developing new applications.
Fostering SDV innovation with the cloud
Building an SDV platform on the cloud is fundamental to future-proofing automotive strategies. An SDV cloud layer facilitates the fleet management of connected vehicles and allows automakers to scale software applications throughout the vehicle's life cycle.
The following SDV applications are built on the cloud platform layer:
Choose your silicon partner wisely
Today's software-defined automotive market starkly contrasts with the hardware-defined nature of years past. Automakers used to focus on simply selling as many vehicles as possible and hoped customers would return when they needed a new one. While maximizing vehicle sales is always a goal, the post-sales experience is now front and center. Brand differentiation will come from the OTA software updates customers receive throughout the vehicle life cycle.
Great silicon is a core aspect of seizing the SDV opportunity. Semiconductor suppliers enable automakers to create revolutionary driving experiences that will drive fresh post-sales revenue streams. The automotive industry is at a crossroads, requiring a pivot in vehicle design philosophy and business models. To ensure a smooth transition to SDVs, automotive OEMs must select a silicon partner that adequately supports the full stack of enabling the technologies identified in this article.