Software is still waiting for metal, and software-defined vehicles are paying for it

Software-defined vehicles (SDVs) promise faster development and continuous improvement. Many automakers are still building them with processes designed for a different era, one where the prototype sits at the center of the program and software waits for physical hardware to exist.

The cost of that approach is easy to see. Integration problems and real-world edge cases surface late, when they are most expensive to fix. Teams stitch together stacks from many vendors, work in silos with their own tools and each builds its own version of the world. Meanwhile, software recalls are rising. An SDV is never finished. It is trained, tested and updated continuously, and treating software as a component bolted on at the end means shipping late and fixing in the field.

In a recent interview with Automotive World, Julian Gonsior, Director of Product Management, Automated Driving at HERE Technologies, argued that the fix starts earlier than most programs expect. His answer to when software architecture should begin shaping vehicle decisions is blunt, and it challenges the usual order in which electrical/electronic (E/E) architecture, sensor sets and compute budgets get decided.

The conversation moves from the problem to a practical alternative. Gonsior explained how running software on virtual ECUs (electric control units) in the cloud, pre-integrating supplier components and generating test scenarios from real-world location data lets engineers compare architecture options with working code rather than slideware. He also put figures on the potential gains in development speed, research and development cost, and testing capacity. Those figures are worth seeing in context.

He is equally clear about the limits. Some validation still demands photorealism, full vehicle dynamics and centimeter-level map accuracy, and some confidence only real miles can provide. His view on where physical hardware earns its place is a useful counterweight to the usual virtual-first enthusiasm.

The discussion also covers why a single shared location foundation matters from development through simulation, validation and production, and what happens when training, simulation and the production vehicle each work from a different view of the road. Gonsior described the closed loop he believes will separate the leaders from the laggards, and hints at what HERE plans to unveil at CES.

Read the full story at Automotive World