Developing AI-driven in-vehicle interaction. KARLI & Audi.

Within KARLI, Audi is working on AI-based interactions in the vehicle with regard to person- and context-sensitivity. The goal is for the human-machine interface to behave adaptively based on the user and system state of the individual operating history.

Author: Doreen Engelhardt

Henry Ford once said, “Any customer can have a car painted any color that he wants, so long as it is black.” In the early days, this was very much following the understanding of industrial automobile production. Today, however, technology gives us the opportunity to develop true individuality instead of “one-size-fits-all solutions.” We want our customers to feel that their Audi is individually customized for them.

At KARLI, Doreen Engelhardt of Audi works with a team from different departments to implement this adaptive individualization. “We focus on designing and developing AI-based interactions in the vehicle for person and context-sensitivity,” she describes.

Our goal within the KARLI project is for the human-machine interface to behave adaptively based on the user and system state of the individual operating history. Furthermore, it should act according to the contextual factors in the vehicle and outside, the current automation level, and the ongoing task.

The focus is primarily on the user and thus also on the question: What are the user-specific preferences to which the system should adapt on a situation-specific basis?

To find that out, we cannot rely on large data sets, i.e., Big Data approaches, because they are too generalistic. We use user-centered methods here, such as innovative active Question Answer agents developed by Manya Sahakyan.

The principle is to trigger a proactive conversational interaction by using multi-agent query reformulations to get answers from the user considering all the required contextual information. We combine this new technique from Conversational AI with operating routines derived from user interaction.

The project KARLI is particularly interesting for Audi because it provides us with a framework for a holistic view of AI-based interaction. More specifically, KARLI allows us to address user interface design as well as machine learning algorithm development and validation.

For this reason, the following three departments are involved in the project: Development Innovations User Interface/User Experience, Innovation Management Product and Experiment Setup, and the Innovation Center.

In addition to the broad opportunities KARLI opens up for us, the consortium offers an exciting exchange and collaborative, iterative learning.

Audi’s KARLI team consists of the permanent team members Manya Sahakyan (technical development Question Answer Agent and learning algorithms), Dr. Lena Rittger (design User Experience), Martin Satzke (overall vehicle integration), and Dr. Vahid Hashemi (AI assurance). Besides the content work for the design of the user experience, Doreen Engelhardt takes over the project management at Audi as well as the application management for the work package AP-400 “AI interaction”.

In this work package we are looking forward to collaborating with the partners paragon semvox GmbH, INVENSITY, studiokurbos, TWT, and Uni Stuttgart IAT.

Leave a comment