Cognata, a provider of autonomous vehicle (AV) and ADAS data generation technologies, has launched DriveMatriX, a new technology designed to support AV development by generating alternate versions of pre-conducted test drives that account for different weather and lighting conditions.
Driven by generative AI technology, its real-data solution Cognata is aiming to provide customers with a unique combination of real-data augmentation and photorealistic synthetic data generation for perception training and validation.
DriveMatriX allows companies to use their existing, real data from test drives to create multiple, high-fidelity, versions of these same tests under varying weather and lighting conditions. Through it, developers at these companies can more easily generate complex scenarios based on existing clear weather data – such as rain, snow – or different times of the day to ensure the more comprehensive, more accurate, validation of AV and ADAS perception capabilities.
While synthetic data has historically represented a key part of Cognata’s technology portfolio, the provider’s DriveMatriX real-data technology aims to look ahead – enabling a new approach for the use of real-world test data. Here, the technology ensures comprehensive testing across different scenarios in a way that allows developers to train and validate perception systems with real data for real-world accuracy, while synthetic data provides additional flexibility in exploring edge cases and less common conditions.
In allowing users to leverage existing test drives, DriveMatriX enables organizations to reuse their data and apply the technology’s augmentation capabilities to create several test environments. While reducing the need for repeated physical testing, Cognata says that this also helps reduce the time, money, and resources that its customers may spend on AV development and ensure that their AV are exposed to a wide range of on-road scenarios.
To demonstrate the capabilities of DriveMatriX, Cognata rolled out a complimentary dataset, through which it intends to help developers and researchers train and validate their perception systems. Through it, the provider is ultimate aiming to demonstrate DriveMatriX’s ability to address real data needs and support AV development.
