2024/09/24

AI & GenAI condense ‘code to road’ timelines for the Software Defined Vehicle

An illustration showing a woman in a car.

IBM’s Hans Windpassinger explores how AI technologies are driving productivity enhancements in SDV R&D.

In the automotive industry, the race is on to win in the Software Defined Vehicle (SDV) market –  forecast to reach a value of $250 billion by 2032 (a CAGR of 22.1%). Pressure is mounting to get new features from “code to road” as fast as possible, upping the ante on already stretched R&D teams. In this context, AI and GenAI offer significant potential to increase engineering productivity – far beyond code generation – and are already being exploited in a wide range of use cases to enhance and accelerate SDV R&D.

Enabling in-vehicle AI applications

SDV R&D teams have been developing and deploying in-vehicle AI for many years. Indeed ADAS/AD (Advanced Driver Assistant Systems and Autonomous Driving) or the chatbots many luxury cars now feature wouldn’t be possible without it.

These have so far been created using traditional AI technologies such as Machine Learning and Deep Learning. Today, however, OEMs are considering bringing in small Large Language Models (LLMs) to their vehicles to power onboard chatbots. These models, also called consequentially Small Language Model (SLM), can run in the vehicles thanks to increasingly high-performance compute units. Further model minimization, necessary for the deployment of models in strongly confined hardware environments (in-vehicle), can be achieved by cutting-edge technology from IBM Research, which can optimize models automatically to adjust to the diverse boundary conditions (memory, flops, storage) without any loss in inference fidelity and latency.

What’s needed in addition are strong capabilities in new areas such as LLM lifecycle management and GenAI governance. Here, significant time can be saved via collaboration between car makers and IT service providers, sharing learnings and experience, as well as advanced platforms like IBM® watsonx.governance, built to direct, manage and monitor the AI activities of an organization.

Supporting software and systems engineering

In Software Engineering, meanwhile, GenAI can support various activities including generating user stories, acceptance criteria, UML models, source code, make files and different kind of tests.

These uses can also be enlarged to cover Systems Engineering, generating out of descriptions in natural language semi-formal SysML V2.0 models, for example, as well as checking and improving the quality of engineering artifacts. Whilst it has long been possible to check requirements and identify improvements using traditional Natural Language Programming (NLP) technology, with GenAI and LLMs, such quality checks are much comprehensive and can be done in seconds.

At IBM, we are already using all of the above mentioned topics to streamline customer projects, with solutions like IBM Consulting Advantage, an AI Services Platform and Library of Assistants.

Enhancing knowledge management

Knowledge management is another critical area in which AI and GenAI can revolutionize R&D. This is particularly crucial at a time when so much car-making experience is being lost as skilled engineers reach retirement age.

To help build up this knowledge in the younger generation, companies need to enhance the systems they use to collect, create and share organizational insights. This requires building a Q&A resource from a broad internal or external knowledge base.

With the help of GenAI, it is possible to analyze multiple documents and data inputs, provide effective responses based on real-time information feeds and improve documentation quality, ensuring that your company’s legacy knowledge remains a competitive advantage.

Solving complex problems

Many engineering problems are far easier to solve with the help of AI and GenAI.

In the truck industry, for example, flexible configurations are a major differentiator, but can create high complexity and lead to non-buildable trucks being discovered during production.

To avoid the high costs generated by manual verification and re-work, we recently collaborated with a European truck manufacturer to implement a pipeline to ingest data from manufacturing and engineering systems and predict the buildability of variants.

Inspiring improved vehicle operation

AI and GenAI can also play a key role in generating and interpreting advanced insights from vehicle operation and feeding it back to the R&D team who can analyze it immediately.

This is important as driving habits differ in ways that are not always easy to predict. We worked with a Japanese car maker, for example, on solutions allowing them to use telematic data to make safety improvements as well as design a personalized steering experience.

AI-powered insights can also be critical for unlocking predictive maintenance efficiencies, and an unlimited range of patterns and clues that would never be found in a manual search.

Streamlining R&D operations

R&D is an organization like any other enterprise, spanning procurement, a supply chain, IT, HR – all functions where productivity can be enhanced by AI.

Let’s take Patent Management as an example. Before filing, existing patents must be read to determine whether the features are indeed new.  This requires lot of time and is error-prone as analysts can write queries too specifically or too broadly that leads to either too few or an overwhelming number of patent results.

By using explainable gen AI in patent research, IBM has been able to significantly reduce work efforts and times.  Work that typically takes several days can now be achieved in just a few minutes.

Driving scale

IBM’s best practices advise capturing the full productivity-boosting potential of AI & GenAI by first establishing an experienced “AI & GenAI R&D improvement team” to design and develop solutions that represent your company values and drive SDV success.

Secondly, we recommend leveraging an open, trusted and empowering platform, allowing you to run AI & GenAI in a hybrid environment, where certain sensitive artifacts and activities can run on-prem while others use public cloud.

Finally, AI and GenAI applications must be embedded into existing tool environments, for easy use. Most PLM and ALM tools allow the integration of client-specific plug-ins to start programs and utilities outside the tools.

What’s up?

Link copied Link copied?

What else drives us

Contact

Link copied Link copied?

Any questions?

Let's connect and find out how we can make things happen.

Hans Windpassinger
Business Development Executive | IBM Deutschland