Parallel Domain Launches Virtual Training Technology for Self-Driving Vehicles
(1 month ago)
MENLO PARK, Calif: Parallel Domain introduced its virtual world generation technology to enable the large-scale training and testing of driverless cars before they hit public roads.
The company makes its public debut armed with its Seed funding round of $2.5 Million led by Costanoa Ventures and Ubiquity Ventures, with participation from RRE Ventures, and Bessemer Venture Partners.
Autonomous vehicles and ground transportation are projected to grow to multi-trillion-dollar markets at scale. However, for vehicles to demonstrate an acceptable level of safety, they will need to drive an estimated 11 billion miles to surpass current human standards and be prepared for critical corner cases like a pedestrian bolting in front of a moving car.
"We are at a turning point in the industry. Driverless cars need massive quantities of challenging training miles in order to learn how to drive safely, but these real-world miles are dangerous, expensive, and inflexible. State of the art simulations alleviate these bottlenecks while providing essential interactivity, control, and reproducibility," stated Kevin McNamara, Founder and CEO of Parallel Domain. "Our software automatically generates the environments and scenarios that feed into simulators, making it safe and fast for autonomous vehicles to learn from their mistakes, accelerating time to safety for all vehicles."
Autonomous vehicle companies often build training worlds for their simulators by hand. This is prohibitively slow, often taking weeks to craft a few city blocks. Parallel Domain's platform can generate multiple realistic, highly detailed city blocks in less than a minute, unlocking the necessary simulated training and testing environments for autonomous vehicles.
Parallel Domain's platform encompasses multiple ways to generate new worlds quickly. Real-world map data can be used to automatically reconstruct a photorealistic, living world complete with traffic, pedestrians, time of day, and more. Any element is adjustable and programmable, from the number of lanes to the condition of the asphalt. Road curvatures, locations of mountains, and hundreds of new cities can be generated with a few simple clicks. Further, procedural growth algorithms and generative models can create a massive variety of fictional worlds, whether they are completely new or similar to real locations.
Parallel Domain is launching with NIO as a customer, a global startup company that designs, develops, and produces smart, high performance, premium electric vehicles for the Chinese market. Parallel Domain's approach works for any company currently developing driverless car technology. They ensure that car sensors and machine-learning algorithms all work for the full spectrum of simulated driving scenarios before a car ever hits the road.
"Autonomous vehicles need to be taught to drive, and Parallel Domain offers the best solution," said Mark Selcow, Partner at Costanoa Ventures. "Every day, we see headlines questioning the safety of autonomous vehicles, but Parallel Domain's approach can help eliminate many of those concerns in virtual worlds before those vehicles reach public roads."