XPeng introduces latest mobility technologies at Tech Day 2022

XPeng, one of main Chinese electric vehicle (EV) startups, unveiled its latest technical achievements at its fourth annual Tech Day held on Monday, including the company's next-generation neural network-based perception architecture, closed-loop AI and data management platform, industry-leading voice and smart cabin technologies, as well as the latest developments in its mobility ecosystem that encompass robotaxi, flying car and robotics endeavors.

"Our strong technological R&D capabilities allow us to fulfill customer needs with innovations,” said He Xiaopeng, Chairman and CEO of XPeng. "Each component we originate for our mobility ecosystem strengthens our proprietary platform and competitive advantage, while providing a springboard for our significant future growth.”

XNGP

According to XPeng's introduction at the event, XNGP is designed to be the company's last step before the realization of full autonomous driving, enabling advanced driver assistance in multiple driving scenarios from highways and car parking to complex city roads and urban expressways. It was introduced first at the XPeng G9 flagship SUV, which hit the market in September this year.

Photo credit: XPeng

XNGP is supported by a slew of hardware upgrades, including 508 TOPS of computing power, dual-LiDAR system, 8-megapixel HD cameras, as well as the new software architecture XNet backed by a closed-loop, self-evolving AI and data system.

XNet

XNet is XPeng's next-generation neural network-based perception architecture.

Unlike the first-generation visual perception architecture, XNet utilizes the startup's in-house developed deep neural network for visual recognition with human-like decision making capabilities. The neural network technology replaces the complicated manual processing logic to achieve a self-evolving data-driven algorithm.

The XNet-enabled autonomous driving technical stack is backed by the Fuyao supercomputing center for autonomous driving. Supported by Alibaba Cloud's intelligent computing platform, Fuyao's computing capability can reach 600PFLOPS, allowing the time span for a model training to be reduced from 276 days to 11 hours.

Closed-loop self-evolving AI and data platform for autonomous driving

XPeng said it has built China's first fully closed-loop autonomous driving data system capable of data collection, labeling, training and deployment. By virtue of a fully automatic labeling system and Fuyao's supercomputing capability, every detected corner case can be handled by using a targeted data collection approach to acquire a great deal of training data from a mass-produced vehicle fleet.

The company also introduced that its highly efficient AI capabilities allow incessant and unmonitored machine learning and rapid iterations in the models trained, addressing over 1,000 rare corner cases per year.

Robotaxi development

XPeng's G9 SUV became China's first mass-produced commercially-used vehicle to sail through the government-led autonomous driving closed-field test, a major step forward in XPeng's Robotaxi development, said the company.

Photo credit: XPeng

"Our platform-based robotaxi development aims to generate significant cost benefits, and ensure product quality, safety and user experience," said Dr. Wu Xinzhou, Vice President of Autonomous Driving of XPeng.

In-car voice system

Unveiled earlier, XPeng's all-scenarios voice system 2.0 version can recognize commands from every passenger in the cabin and understand speech instructions across multiple conversation streams with an accuracy exceeding 96%. It takes the newest voice system less than one second to perform execution after receiving commands as the system can respond to instructions without an activation command or even internet connection, delivering interface feedback in only 245ms and voice feedback in less than 700ms.

Photo credit: XPeng

The company stressed that it would make its whole-car, always-on voice system as a standard feature on all of its new vehicles.

Upgrades of AI-powered robot prototype

Powered by AI capabilities, XPeng's equine robot prototype has been upgraded with a new design to better support multi-degree-of-freedom (MDOF) motion, and locomotion capabilities.

Photo credit: XPeng

Upon completion of the upgrades, the motion control system of the aforesaid robot prototype has been better adapted for more complex indoor and outdoor terrain conditions, such as stairs, steep slopes and gravel roads.

Newest version of eVToL flying car

Developed by XPeng's ecosystem affiliate XPENG AEROHT, XPeng's eVToL flying car has been optimized in design with the last year's horizontal dual-rotor structure replaced by a new distributed multi-rotor configuration, so as to further improve flight safety and reliability.

Photo credit: XPeng

In driving mode, the flying car's latest version is comparable with any conventional car in terms of functionality and measurement. In flight mode, it can be piloted with a steering wheel and controlled by a right-hand gear lever to move forward and backward, make turns, ascend, hover and descend.