We use cookies on this site to enhance your user experience.
By continuing to browse our site you accept our cookie policy. Find out more. OK, I agree
New Mobility World @IAA Pkw 2017
September 14 - 17, 2017 Frankfurt am Main
Automated Driving

“Highly Precise and Intelligent Maps Are a Crucial Foundation for Autonomous Driving.”

By Editorial Team on July 24, 2017

In this new interview series, we discussed with Manuela Rasthofer, CEO and cofounder of TerraLoupe . TerraLoupe is a German-based artificial intelligence startup. The company recently closed a seven-digit round of financing backed by Bayern Kapital and private investors.

How will we move from A to B in 2037?

TerraLoupe firmly believes in the future of autonomous driving. But rapid growth will not happen overnight. The combination of different trends will occur. Key trends from our perspective will be autonomous driving, vehicle electrification, and ride sharing (Uber), —which together can create a compelling economic case with cutting travel costs by 60% according to BCG research and most convenient form of transportation.

We are thoroughly excited on the advantages that autonomous driving is bringing, as we will all benefit from less congestion in the cities, safer driving with fewer accidents and cleaner air.

What role will new maps play for autonomous driving?

Autonomous cars will require maps that must be different in several important ways from the maps we use today for turn-by-turn directions. New maps need to be high-definition. Meter-resolution maps are maybe good enough for GPS-based navigation, but autonomous cars will need more accurate maps that can tell where a lane marking or road edge is within a few centimetres. To sum up, highly precise, intelligent maps are a crucial foundation for autonomous driving.

Today mapping is done from a car point of view. This needs to be accompanied by aerial mapping in the future. Satellite images are not very detailed. Aerial photographs from airplanes and in the future UAVs can be far more thorough delivering the required resolutions and oblique views.

However, if you want to tag all the traffic lights, construction and filling stations, you need a lot of time, patience and human resources – or smarter – that a machine takes over the work. And this is exactly what our company Terraloupe is doing. We provide complete, accurate information of objects from an aerial point of view, using artificial intelligence to speed up the process significantly.

That enables the exhaustive recognition and processing of objects and infrastructures: road networks precisely. The information has an accuracy of only a few centimetres so that autonomous cars would be able to react in their environment.

How will your technology change autonomous vehicles?

TerraLoupe generates highly accurate HD Maps for autonomous cars via deep learning using aerial images. Our core competence is the automatic recognition of objects in aerial images via deep learning algorithms.

Interesting objects can be classified and described by type, size, position, and more. The extracted digital road information can improve the overall accuracy of other existing data sources. We provide complete, accurate information of objects from an aerial point of view. This enables TerraLoupe to deliver high-quality data at a faster pace and lower cost for large geographic areas. A self-propelled car must not only perceive immediate obstacles but also must know about the general environment to expect.

TerraLoupe offers digital information of roads in cm-accuracy. The absolute accuracy of the exact location of all recognised objects is less than 10 cm. We offer digital route information in a variety of data formats such as Shape Files, XML, NDS, and OpenDRIVE data formats and coordinate reference systems such as WGS84 / UTM Zone 32N – EPSG:32632. Today, some big challenges of mapping for autonomous driving are still lying in:

• Safety and comfort that requires redundancy of mapping data and sensor data
• Safety via quality need – high accuracy mapping of road objects
• Positioning problem of autonomous vehicle
• Large scale solutions for all roadways
• Massive data use
• Quality assurance for Mapping Data
• Dynamic map updates needed

TerraLoupe can address most of the challenges, and we can claim that we will increase safety, comfort and efficiency of autonomous cars by digitising all relevant objects in the real world. If you are asking yourself how? Here are some answers:

• Digital representations of various real-world objects in cm-accuracy
• Enabling highly precise vehicle positioning
• Superior performance through faster and more robust automatic object recognition via deep learning algorithms
• Quality assurance process for highest levels of automotive requirements
• Flexible integration with existing mapping solutions
• Optimised data size for efficient computations

How can self-driving cars win the public trust?

Today, technology tends to be more trusted when it is under human control – even when evidence suggests that the human element creates most of the problems. Looking at the current statistic from KPMG, 93% of road accidents are caused by human error; by 2030, the introduction of driverless vehicles could save 2,500 lives and reduce the number of dangerous accidents by 25,000 each year.

Safety will be key for public trust and TerraLoupe’s technology will improve safety through adding the aerial perspective to what I autonomous vehicle “knows”. Much of the technology is already available, and the benefits are clear enough. But to avoid a bumpy road to an autonomous future, strong collaboration is required. For gaining Public trust, we all need to communicate the outset about the social and long-term benefits of autonomous driving.


###

About TerraLoupe

TerraLoupe is an award-winning German artificial intelligence start-up. Founded in 2015, our belief is that understanding accurate geo image data can fundamentally improve decision making for businesses. Our core competence is the automatic recognition of objects in aerial images and the full digitalization of the environment via deep learning. We use a unique approach to extract information of objects in aerial images and thus, can quickly provide customizable maps.

Our AI algorithms allow for a more precise and economic planning. Among other industries, we have successfully supported companies in the field of automotive, infrastructure, and insurance. In the automobile sector, we successfully digitalized highway and road tracks for leading German automotive OEMs.

Hero image source: TerraLoupe