aiSim World Extractor | High-Fidelity 3D Environment Generation

Neural simulation

Accurate environments at scale with neural reconstruction

aiSim 4

Ultra high-fidelity 3D environments at scale with neural reconstruction

Virtual validation of complex automated driving solutions requires a diverse set of 3D scenes with a variety of assets to ensure that sensor models are similar to real-world conditions. With the aiSim World Extractor, you can generate accurate environments at scale with neural reconstruction in a fraction of the time and cost compared to traditional 3D modeling.

AI-based reconstruction & rendering

Generate maps and environments from real-world recordings across various ODDs within a day, compared to asset creation that requires months of manual work from 3D artists.

Real-time performance with "real-like" environments

The aiSim World Extractor supports 3D Gaussian Splatting and Neural Radiance Fields (NeRF) to accommodate both real-time performance and the highest possible fidelity.

Augment the scenes with a diverse set of assets

Add various 3D vehicle models, vehicle dynamics, and traffic events with full OpenSCENARIO support. Use aiFab to create a large number of diverse and interactive scenarios. Decorate the maps with the thousands of traffic furniture available in aiSim's asset library.

Simulate a variety of physically accurate sensor models

Simulate complex sensor setups, including cameras and lidars, in a physics-based and deterministic simulation environment. aiSim's performance-efficient neural rendering allows real-time Hardware-in-the-Loop (HiL) testing to evaluate the performance of your target ECU.

Automated processing from real-world recording to reconstructed scenes

The aiSim World Extractor is an end-to-end pipeline for neural reconstruction, using complex neural networks to understand the recordings and generate a simulation environment. The accuracy of scenes removes the domain gap between simulation and reality.

Benefits of aiSim World Extractor

• Single-click, automated pipeline creating an aiSim simulation 3D environment from raw recordings
• Neural reconstruction generated detailed and highly accurate 3D scenes in a fraction of the time compared to traditional scene creation technologies.
• Save costs by relying on an automated process instead of the manual work of 3D artists.
• Achieve an exceptionally detailed quality of the scenes, removing the domain gap between simulation and reality.