Products on Show
Automated driving full software stack
The complexity of driving automation has advanced significantly since the development of assisted driving systems with L2 ADAS functions, such as adaptive cruise control or lane keeping. However, there have been challenges in realizing fully autonomous driving by 2025, alongside the availability of L3 systems, due to the requirements for robust perception and data for development and validation.
Perception is fundamental to any automated driving software. Reliable neural networks (NN) that form the heart of the perception pipeline require large, diverse, contextually rich annotated data sets for NN training and validation.
In Stuttgart, aiMotive will introduce the latest generation of its automated driving full software stack, aiDrive 3.0, featuring in-house, model-space-based perception, which it says is the most solid foundation for automated driving solutions. The stack can process inputs from multiple sensor modalities (camera, radar, or lidar), and cater for multiple tasks (for example, lane and object detection) – all of which are fed by the output of the automated data annotation pipeline.
aiDrive 3.0 has instigated the company’s journey towards data-driven neural network training, allowing for rapid, continuous evolution. Its model-space-based perception is the foundation of higher-level automation but includes several innovative advancements in comparison with traditional perception pipelines. For instance, it is reusable across different projects at any automation level, which reduces the need for repeated data collection and annotation, development costs and time-to-market.
The virtual sensor technology makes it lightweight: the whole stack only requires low-power chips (less than 2x30W) for full L3 functionality.
Its detection performance enables automated driving technology to realize its goal of making road transportation safer for everyone. This includes excellent results in low-light and adverse weather conditions, significantly improved distance estimations, recall and position metrics, alongside far-range detection.
Having enough of the right data is critical for such systems. Over the past few years, aiMotive has invested in the development of pioneering in-house data tools: an automatic annotation pipeline, supported by synthetic data generation (aiData), based on the company’s physics-based simulator, aiSim. These tools ensure endless data supply, a vast diversity of data, quick development cycles and the continuous adaptability of the software to meet changing market needs.