Products on Show
DAY 2: Automated driving full software stack with reusable, model-space-based perception
aiMotive is showcasing the latest generation of its automated driving full software stack, aiDrive 3.0, featuring in-house, model-space-based perception, which the company says is the foundation for higher-level automation but achieves much more than that. It is an innovative system for the data-driven evolution of both direct model-space-based perception and multi-sensor fusion, supported by a fully automated annotation pipeline.
The company is showing a front-camera ADAS application of aiDrive 3.0 running on Texas Instruments’ TDA4. All ADAS/AD features require a 3D representation of the vehicle’s surrounding environment – the model space – to perform decision making and path planning. Model-space-based perception supports this basic requirement by creating a 360° 3D model space directly from the fusion of data from multiple cross-modality sensors, meaning the overall ADAS/AD feature stack achieves greater clarity. “This represents a fully data-driven evolution, including multi-sensor fusion, which reduces complexity and computing resource needs,” revealed Gábor Pongrácz, aiDrive senior vice president, live at the expo.
aiMotive claims to provide more than just software, preferring to think of itself as a reusable and efficient data and simulation pipeline with the primary goal of decreasing an automotive player’s time-to-market. aiSim, its end-to-end validation suite, accelerates the evolution of aiDrive’s features, while aiDrive ensures aiSim can validate complex automated driving solutions. In addition, aiData, the company’s proprietary data pipeline, provides the foundation on which its other solutions can build, and reduces the complexity of processing by using high automatization while ensuring automotive quality.
Validating the safety of any system requires processing vast amounts of validation data, but aiMotive’s fully automated data pipeline, paired with its perception solution, reduces overall effort and the cost of validation data management, ensures the capability of continuous improvement and a fast time-to-market while keeping the cost for manual annotation as low as possible.