By Bobby Carlton
Built on the Omniverse platform, NVIDIA Drive Sim provides engineers with a complete end-to-end simulation solution, and allows them to train their neural networks and perform motion control simulations.
The development of self-driving cars requires immense amounts of data. Engineers must analyze and label all of this information in order to train their neural networks. Through this data, they can then test and validate the systems that are designed to drive autonomous cars.
One of the biggest factors is the accuracy of the data, and simulation becomes a crucial tool in this process, as accuracy is often the determining factor in its success.
To help researchers collect realistic and physically accurate data, they are turning to simulations through the use of the NVIDIA Drive Sim platform, which is built on Omniverse. This solution is designed to provide engineers with a complete end-to-end simulation solution, and allows them to train their neural networks and perform motion control simulations.
AV sensors can be categorized as follows:
- Passive: Cameras
- Active: Lidar, radar, and ultrasonic
Through the use of the Nvidia Drive Sim platform, we can now confidently deliver accurate lidar models that match the real world. In a recently published whitepaper, Nvidia talked about the process that enables them to achieve this goal.
The image below shows the active sensor data that’s moving through the Drive Sim pipeline. The first step is to create a representation of the data that’s sent to the Omniverse World model. The second step is to send the data to the NVIDIA RTX sensor API plug-in.
With this tool engineers are able to collect important and accurate data that will have a significant impact in many industries.
You can read Nvidia’s Drive Sim/LiDAR validation white paper here.
