dSPACE, a provider of solutions for the development of networked, autonomous, and electrically powered vehicles has acquired the start-up company understand.ai in order to strengthen its expertise in AI.
Under the umbrella of the dSPACE group companies, understand.ai will invest in the core tasks ‘artificial intelligence (AI) applications’ and ‘cloud-based tools’, further develop its existing products as an integral part of the dSPACE product range, and use the dSPACE global sales network to market its products and services.
‘Understand.ai is an AI technology leader with a focus on automated data analysis, data annotation, and extraction of simulation scenarios for autonomous vehicles. With these key technologies, we are strategically enhancing the dSPACE portfolio to offer our customers a unique, integrated development and test solution for autonomous driving,’ says Martin Goetzeler, CEO of dSPACE.
‘We see dSPACE as the ideal partner for growth as understand.ai takes its next development step. Testing is the bottleneck in the development of autonomous driving, and dSPACE, as a leading development partner of the automotive industry, will give us a strong momentum with its expertise and network,’ explains Marc Mengler, co-founder and CEO of understand.ai.
‘Marc Mengler and I are looking forward to working with dSPACE to improve local customer service, accelerate our international growth, and further expand our global leadership position in solutions for providing training and validation data,’ adds Philip Kessler, co-founder and CTO of understand.ai.
In the development and introduction of autonomous vehicles, it is crucial to detect the environment of the vehicle realistically and without faults. Other road users, traffic signs, lanes, the static roadside structures, and open spaces must be reliably identified.
For this purpose, self-learning (machine learning) algorithms, in particular deep neural networks (DNNs) based on artificial intelligence, are used in these autonomous vehicles. These algorithms must be trained and validated efficiently. Therefore, it is a requirement to analyse, annotate, and also anonymise a tremendous amount of recorded (camera, lidar, and radar) sensor data.
The quantity, quality, and diversity of this training and validation data determine the quality of the resulting DNNs. The annotation process, also called labeling, is required for classifying the objects as a reference for machine learning. Today, this process carried out manually, which is mostly time-consuming and does not always ensure the highest quality level.
understand.ai has proprietary expert knowledge that enables automating this process to the greatest extent possible. The company also uses self-learning algorithms to process high-quality training and validation data. The underlying key technology is also based on artificial intelligence and ensures efficient data analysis as well as precise data annotation, which guarantees high quality training data for AI-based driving algorithms. understand.ai develops AI- and web-based tools for this area of application. The underlying know-how is also used to extract simulation scenarios from sensor data.
