Computer Vision / Video Analytics

How NVIDIA DriveWorks Ensures AV Sensors Stay Aligned

Lidars and cameras alone aren’t enough to put self-driving into action. Sensor diversity is a cornerstone of autonomous driving. However, it only works if every sensor is in alignment. The NVIDIA DriveWorks SDK makes it possible to perform sensor calibration both offline, before the vehicle hits the road, and while the vehicle is driving with self-calibration.

Developers can learn how to use this tool for their own development processes in an upcoming webinar led by NVIDIA experts, streaming on Dec. 9.

A common use case of sensor calibration is the process of bringing different sensors, such as camera, radar, lidar and IMU, together on one coordinate system. By fusing this real-time data together, autonomous vehicles can reliably perform critical functions such as localization and ego-motion.

DriveWorks delivers all the functions necessary for autonomous vehicle development, including sensor integration, data recording  and calibration capabilities. It’s modular and designed to be compliant with automotive industry software standards.

Getting on the Same Page

The variety in sensing data provided by camera, radar and lidar is what makes the sensors optimal aids and eventual replacements to human eyes — from visual information to distance to speed of surrounding objects, and more.

However, these different sensing modalities all operate on various coordinate systems. Additionally, manufacturing processes result in slight variations in the physical sensor itself compared to its specifications.

Discrepancies in sensor data can significantly affect the reliability of a vehicle’s perception. For example, if a camera and lidar sensor can’t agree on the location of a stop light, the vehicle won’t be able to react appropriately nor localize itself in the driving environment.

Since sensor alignment is so critical to safe autonomous driving, calibration is performed in the garage before the vehicle begins driving — what’s known as offline or static calibration — and while the car is driving — self-calibration — to ensure sensors that have been knocked or jostled are still operating on the same coordinate system.

Doing It All on DriveWorks

NVIDIA DriveWorks makes it easy for developers to perform both static and self-calibration for safe and robust autonomous vehicles.

The SDK contains self-calibration algorithms as well as key performance indicators with which to gauge sensor accuracy. Developers can use these building blocks to construct their own unique self-calibration solutions.

These calibration modules are also optimized to run on the NVIDIA DRIVE AGX platform, for high-performance, energy-efficient AI compute.

These tools and more will be covered in our Dec. 9 webinar. Register today to learn how to streamline the autonomous vehicle development process.

Discuss (0)