Cognibit: Making Autonomous Driving Safer

The Munich startup Cognibit aims to make autonomous driving safer, which is why the founding team has developed software that simulates human behavior in traffic. It’s an important issue, because people in traffic are not all that easily predictable. Co-founder Isabelle Garzoz answers our Seven Questions

Munich Startup: Who are you and what do you do? Please introduce yourselves!

Isabelle Garzorz, Cognibit: We are the Cognibit founders, Lukas Brostek (40), Alex Knorr (32) and me, Isabelle (31). Our shared vision is to make autonomous driving safer. As neuroscientists, all three of us conducted research on motion perception and behavioral modeling at the Center for Sensorimotor Research at LMU Munich. After his postdoc, Lukas worked for a large car manufacturer where he gained many years of experience as a development specialist in the field of autonomous driving. He also worked on the subject of “virtual development,” which is the computer-based simulation of traffic situations. It was against that background that the idea emerged – based on neurocognitive findings – to create simulation models of human behavior in traffic. That was how the business idea emerged, and we applied together for the Exist start-up grant, which has supported us since March this year.

Human behavior as a challenge for autonomous driving

Munich Startup: What problem does your startup solve?

Isabelle Garzorz, Cognibit: Cognibit stands for “Cognition and behavior in traffic,” and that is exactly what we want to make possible with our software – the simulation of human behavior in traffic. Why is that necessary?

An autonomous car has to drive approximately 18.5 billion miles for a car manufacturer to be certain that it’s safer than a human driver. Due to that magnitude, tests are now predominantly conducted as simulations. But that only works when the conditions in the simulation are realistic – after all, the car should also face difficult traffic situations. And human behavior in traffic, which the tested vehicle should interact with, presents the biggest challenge. Here at Cognibit, we realistically simulate that behavior.

Out of the black box

Munich Startup: But that’s nothing out of the box!

Isabelle Garzorz, Cognibit: Driver models do already exist. But we take a new approach that is based on neurocognitive research and offers many advantages compared to conventional machine learning approaches based solely on data. While the competition is trying to extract and simulate human behavior in traffic from traffic data based on their “black box” approach, we explicitly model this behavior.

We simulate the entire sensorimotor processing chain, which means our driver model also takes human limitations into account. For example, we factor in elements such as a limited field of vision, different reaction times or distractions, and that is extremely important. With our transparent models, situations that are critical to safety, such as accidents or close calls, can be covered and reconstructed in the simulation. Data-based approaches would need an infinite number of data sets to realistically simulate the many variations of critical traffic situations.

Munich Startup: What have been your three biggest challenges so far?

Isabelle Garzorz, Cognibit: Our biggest challenge has certainly been the current corona crisis, which caused many trade fairs and conferences to be cancelled – and for us, the opportunity to talk directly with customers, partners and investors. Nonetheless, virtual alternatives have worked amazingly well so far.

In five years: safe autonomous vehicles on the road

Munich Startup: Where would you like to be in one year, and where in five years?

Isabelle Garzorz, Cognibit: In one year, we would like to have impressed our first pilot customers with our product and to continue developing our product with them. By then, we will hopefully also have one or two new employees to support us in the fields of software development and sales.

Our vision for Cognibit in five years? By then, the aim is for our technology to be used across the board by manufacturers of autonomous vehicles to put safe autonomous vehicles on the road.

Munich Startup: What do you think about Munich as a startup location?

Isabelle Garzorz, Cognibit: We are a very young, newly founded startup, which means we don’t have much experience yet. But we have absolutely no complaints about Munich so far. As a mobility hub, Munich is the perfect place for us to be in Germany. Many of our potential customers and partners are located right here.

Munich Startup: Risk or security?

Isabelle Garzorz, Cognibit: I think our product reflects that very well: Security is always of top priority.