Software-Defined Vehicles Need New Development Tools

Tesla has had a positive impact on the auto industry in two unquestioned ways: making EVs appealing to mainstream consumers and introducing the concept of the software-defined vehicle. The rest of the auto industry is now scrambling to catch up on both of these fronts, and while EVs may not do much to boost profitability in the long term, the potential for new revenue streams enabled by software-defined vehicles has executives and financial markets salivating. However, creating and supporting these products will require a new way of working and much more sophisticated tools. 

Software in cars is not a new concept. The emergence of microprocessors in the early 1970s enabled meeting new fuel economy and emissions standards and later, safety regulations. For the first several decades, the software in these electronic control units was written directly into the hardware and largely isolated from other systems in the car. However, as more functionality has been added over the years, the interactions have become vastly more complex. 

Exponentially More Data and Functions

With hundreds of gigabytes per day flowing through vehicle networks and safety-critical systems interacting with connectivity and entertainment systems, it’s more important than ever to understand the interactions and dependencies. Unfortunately, these systems have become so complex that traditional methods of human code review and static code checkers cannot possibly keep track of everything. 

A new generation of tools that incorporate machine learning and other AI techniques are now required to help developers gain a semantic understanding of what is happening across the vehicle. These tools parse through all of the code, mapping all of the interactions and dependencies. Tools also have to be trained to understand more than just which functions are called from where. Tools must track which aspects are safety-critical, how the subsystems running at different cycle rates may impact other areas of the code, and much more. 

The needs also go well beyond the initial creation of the software. Up to this point, automakers have followed a path of designing, building, and selling vehicles and then moving on to the next generation of products. Once a vehicle left the factory, it remained largely unchanged for its lifecycle, apart from fixes for defects or regulatory compliance issues. 

Expectations Are Changing

Tesla has created customer demand for new features to be enabled by over-the-air software updates, improving the vehicle over its useful life. Understanding what features to develop or how to upgrade existing features can be aided by telemetry data gathered from vehicles. Parsing through that data to get relevant insights is another area that can benefit enormously from this new generation of AI-based tools. Guidehouse Insights has collaborated with Aurora Labs to produce a new white paper, Vehicle Software Intelligence: Adopting the AI Required to Create a Software-Defined Vehicle, that examines the issues around creating, updating, deploying, and supporting modern vehicle software development efforts.