AI Online

Ai INNOVATION, SINCE 1895

Milestone for autonomous vehicle technology

Arriver is using artificial intelligence to detect driving conditions, including free space, road edge, lane markers and objects.

Progress continues to be made in the rollout of autonomous vehicles.

In early 2021, the Polestar 2 electric vehicle with Arriver’s Drive Policy software achieved top Euro NCAP safety results and a five-star rating. “This is a first proof-point of our scalable platform approach for deployment of Arriver software on the Snapdragon Ride Platform. In parallel we are introducing our new machine learning software for vision perception to demonstrate the complete capabilities of Arriver, enabling the best possible safety and convenience functionality for the driver,” says Salah Hadi, CTO of Arriver™.

A wholly owned unit within the Veoneer group, Arriver has started to optimize and deploy next generation software for vision perception and drive policy on the Qualcomm® Snapdragon Ride™ Platform. The Drive Policy software is being ported by Arriver to the Snapdragon Ride Platform with Lane Support System (LSS), Forward Collision Warning (FCW), and Autonomous Emergency Braking (AEB) as initial functionalities, with more to be added.

Arriver is using the latest artificial intelligence and deep learning techniques to deploy its 5th generation 8MP Vision Perception on the Snapdragon Ride Platform, allowing the testing and demonstration of road detection functionalities, including free space, road edge, lane markers and objects.

Deep learning utilizes Convolutional Neural Network (CNN), a machine learning method which enables vehicle recognition up to 400 meters and pedestrian recognition up to 150 meters. The next steps include the introduction of new scene content and context based holistic classifiers to enhance the understanding of the current driving situation. With support of active learning, the system will improve functionality over time with limited need for human interaction, according to Arriver.

The next generation CNN is approximately more than 50 times larger than the existing generation, and enables improvement in true positive (TP) rate, as well as significantly lowers false positive (FP) results with a factor ~x10-x50, given a fixed TP rate, allowing for the higher accuracy necessary for when the car takes over responsibility for the driver, as in Level 3 applications and beyond. “The Arriver software stack is developed to be an open, flexible and scalable solution which will allow automotive manufacturers and automotive suppliers to, based on a robust software platform, tailor and customize solutions to their specific needs based on their individual product development strategies,” says Giuseppe Rosso, Head of Arriver™.

Vehicles will in future be defined by their software stack, he believes. The Arriver 5th generation 8MP Vision Perception software will be scalable, flexible, and expected to be available to auto manufacturers and Tier-1 suppliers as part of Qualcomm Technologies’ future product offerings, allowing car manufacturers to create driving experiences tailored to their own specific needs.