You can find the episode on autonomous driving and all other episodes of our podcast on Spotify, iTunes, Deezer, Google Podcasts, Pocket Casts, Radio Public, Breaker, Overcast, Castbox, Podcast Addict and Anchor.
As usual, the Munich Startup team was out and about at events, which we'll tell you about at the beginning of the episode: This time, it's the 8th anniversary of the WERK1, which was, of course, celebrated digitally. Florian Mann took the opportunity to talk with Britta Weddeling not only about the past of the WERK1 and their shared experiences in the Munich startup scene, but also to take a look into the future. You can also find the full conversation on YouTube.
Focus on autonomous driving
Elon Musk regularly inspires enthusiasm among fans of autonomous vehicles, most recently when he announced in early March that the beta tests for Tesla's Full Self Driving system would be expanded. However, appearances are deceptive, because Full Self Driving is not actually a fully autonomous system, but rather just an advanced Level 2 driver assistant. At least, that's what the Süddeutsche Zeitung a lawyer from Tesla. In addition, the SZ writes, Tesla has filed a complaint with the California Department of Transportation DMV One company hasn't applied for a permit to test driverless cars at all—six other companies have, however. And one company has even received approval for its autonomous vehicles from the DMV.
As the Tesla example shows, the degree of autonomy (and the correct naming) is of great importance in the discussion about autonomous vehicles. in six stages A distinction is made between levels: from Level 0, where nothing runs autonomously and the driver does everything himself, to Level 5, where there is no longer a driver in the true sense of the word, and no steering wheel or pedals. Level 2 autonomy means that the car can perform certain functions itself, such as lane keeping or parking. However, the driver must always supervise the entire process. This is a level that many car manufacturers are currently operating at.
Self-reliant in six steps
The first Munich startup, which we will take a closer look at, demonstrates what autonomous driving at Level 5 could look like: Sigra Technologies is developing a neural network called "Deep Einstein," which is designed to be able to navigate autonomously even through confusing traffic situations. To achieve this, the developers train the artificial intelligence similarly to how a driving instructor trains a student. The AI drives – in a controlled environment, of course – and the developers intervene if it makes mistakes. You can hear all the details about Sigra starting at minute 5:30.
Also works with AI Recognition (from minute 8:15). The startup, based in Munich and California, is working on an image-processing AI platform that enables the processing of sensor data from LiDAR, cameras, and radar sensors in real time and with low power consumption. The platform also includes powerful AI processors optimized for use in cars with Level 2 and higher autonomy.
A really long way
The startup Cognibit is also dedicated to a very special challenge for autonomous driving in the AI field. In order for manufacturers to prove that their vehicles are truly safe, they must complete many test kilometers without accidents. On the highway alone, this is expected to be around 6.62 billion kilometers Since this is difficult to achieve on the road, simulations are used. And it is precisely for these simulations that Cognibit (starting at minute 10:35) has developed software that simulates human behavior in traffic. The startup relies on the results of neurocognitive research rather than purely data-based machine learning approaches.
Sensors for autonomous driving
Of course, autonomous driving doesn't work without proper sensors. The Munich startup ecosystem has a lot to offer in this area, such as Blickfeld (from minute 13:35). The Munich-based company manufactures LiDAR sensors that can be used to optically measure distances and speeds. The startup is one of the pioneers in the field and has been working since its founding in 2016 to make the sensors increasingly robust, smaller, and more cost-effective.
Another technology is used in Toposens The startup relies on ultrasound, and its device mimics the sense of location of bats. The difference here compared to LiDAR lies primarily in the range. While the Blickfeld sensors can detect objects up to 150 meters away, depending on their size, Toposens focuses on shorter distances with a maximum of 4 meters, thus targeting other application scenarios, such as level measurement or collision avoidance in parking assistance systems. Furthermore, the price of the ultrasonic sensor is said to offer a significant advantage over other sensors. You can learn more about Toposens in the podcast starting at minute 17:00.
To conclude the topic, we also considered why autonomous vehicles are always electric vehicles. Our answer: According to a Study by Bain & Company The electric car will prevail much sooner than the self-driving car. Therefore, we conclude, the industry always plans autonomous vehicles as electric vehicles.
Rebike and STS Ventures
In the second part of the podcast (minute 21:15) we take a look at the mobility startup Rebike The Munich-based company sells and rents e-bikes via three different platforms and has already raised considerable capital. Recently, further 10 million euros in a Series BCV STS Ventures was involved from the beginning, and we will also introduce them in more detail.
