A Bristol-based embedded software developer has launched technology to help driverless cars, drones and industrial robots find their way around without using satellite navigation.

Kudan uses 3D-recognition technology from augmented reality (AR) to provide real-time tracking accuracy between 1mm and 1cm from an ordinary camera, a major technological achievement. Because it uses the same AR technology used in mobile phones, KudanSLAM (Simultaneous Localisation And Mapping) ) uses just 5% of the computing power of a mobile processor.

The tech uses computer vision algorithms to process, analyse and understand the digital images from the camera to map the environment in 3D, identifying objects, and understand its location within it. You can see how a driverless car uses KudanSLAM in the video below:

 

“We are very proud that KudanSLAM is now ready for the market and has begun providing it to a number of industries, including self-driving cars, advanced driver assistance systems, drones, robotics, and of course AR,” said Tomo Ohno, managing director and founder of Kudan. “With its precision tracking, high speed, low CPU usage and overall robustness, it can be a powerful tool in these particular industries,” he added.

Making maps

The KudanSLAM software can be combined with sensors and a LiDAR laser ranging system for autonomous vehicles, building a 3D map using a car’s camera. This allows the car to navigate when it cannot get sat-nav data, such as inside a parking structure.

For a drone, the software runs on a low-end processor with a low-cost camera but can still be used in low lighting to find its way around.

The software can process a frame in 1ms, compared to 15 or 30ms for competing open-source algorithms, and provides accuracy down to 1mm, compared to 10 to 30mm. This can be used for positioning of robots in factories and warehouses.

Kudan has a team of engineers at its Bristol headquarters and a sales team in Tokyo. It has raised over $3m for its technology development.

There are more details at the Kudan website.