Guangzhou, China – XPeng Inc., a leading Chinese smart EV company, today unveiled a series of groundbreaking technical achievements and industry firsts at its fourth annual 1024 Tech Day held at its Company headquarters in Guangzhou. Showcasing XPENG’s in-house, full-stack R&D capabilities, this year’s event featured XPENG’s next-generation neural network-based perception architecture, closed-loop AI and data management platform, industry-leading voice and smart cabin technologies, and the latest developments in its mobility ecosystem, including latest robotaxi, flying car and robotics endeavors.
XNGP, China’s most advanced driver assistance covering all scenarios, non-reliant on maps
Introduced at the Flagship SUV G9 launch in September, XNGP is designed to be XPENG’s last step before full autonomous driving is realized, offering advanced driver assistance in the full spectrum of driving scenarios from highways and carparks to complex city roads and urban expressways. XNGP also achieved a technological breakthrough in expanding it to areas not covered by high-precision maps, nationwide.
XNGP is supported by significant hardware upgrades, including 508 TOPS of computing power, dual-LiDAR system, 8-megapixel HD cameras and a new software architecture XNet, backed by a closed-loop, self-evolving AI and data system.
XNet, XPENG’s next-generation neural network-based perception architecture
XPENG unveiled its next-generation neural network-based software architecture XNet on Tech Day. Different from the first-generation visual perception architecture, XNet adopts an in-house developed deep neural network for visual recognition with human-like decision-making capabilities, drawing from multiple cameras’ data. The neural network technology supersedes complicated manual processing logic to achieve a self-evolving data-driven algorithm.
XPENG’s XNet-empowered autonomous driving technical stack is backed by Fuyao, China’s largest supercomputing center for autonomous driving. Supported by Alibaba Cloud’s intelligent computing platform, supercomputing capability at Fuyao can reach 600PFLOPS, increasing the training efficiency of autonomous driving model by over 600 times. That’s to say, time spent can be reduced from 276 days to 11 hours for a model training.
With state-of-the-art deployment optimization, XNet uses only 9% of Orin-X chip’s processing power, compared with 122% before optimization.
Closed-loop self-evolving AI and data platform for autonomous driving
XPENG has established China’s first fully closed-loop autonomous driving data system, encompassing data collection, labeling, training and deployment. Supported by a full auto-labeling system (able to complete 2,000 people/year’s labeling work in 16.7 days), and Fuyao’s supercomputing capability, every detected corner case can be addressed by utilizing a targeted data collection approach to acquire a massive amount of training data from its mass-produced vehicle fleet.
XPENG’s high-efficient AI capabilities enable consistent and unsupervised machine leaning and rapid iterations in training models, resolving over 1,000 rare corner cases each year. This highly efficient closed-loop AI and data system has helped reducing incident rate for the Highway NGP by 95%.
XPENG robotaxi development achieves a milestone, obtaining Intelligent Connected Vehicle Road Test permit
XPENG’s G9 SUV became China’s first mass-produced commercial vehicle to pass the government-led Autonomous Driving Closed-field Test, a major step forward in XPENG’s robotaxi development. The test was completed in accordance with the requirements of the autonomous driving function test for intelligent connected vehicles in Guangzhou, a tier-one city in China.
“Obtaining the road test permit by our mass-produced commercial vehicles – with no retrofit – is a major achievement,” said Dr. Wu on Tech Day. “Our platform-based robotaxi development aims to generate significant cost benefits, and ensure product quality, safety and user experience.”