Xpeng unveils its mobility technologies at company’s Tech Day 2022
On the occasion of Tech Day 2022 held at its headquarters in Guangzhou, China, Xpeng has unveiled in detail its new mobility technologies, including: neural network, AI, robotaxi, eVTOL flying car, robot and more.
Here’s a closer look at each of the technologies unveiled at the event, which is the company’s fourth of its kind.
XNGP, the last step before full autonomous driving
One of the most anticipated developments is its XNGP autonomous driving system, which was unveiled at the launch of the Xpeng G9 100 per cent electric flagship SUV last September. Xpeng claims that XNGP is China’s most advanced driver assistance.
According to the company, XNGP is designed to be Xpeng’s last step before full autonomous driving arrives, offering advanced driver assistance in all driving scenarios, all without relying on maps. Xpeng’s Vice President of Autonomous Driving, Xinzhou Wu, said:
“Our launch of the ‘City NGP (Navigation Guided Pilot)’ pilot programme in September represented a decisive milestone in the second half of the race towards autonomous driving. Improving the customer experience through innovation is our ultimate goal.
We are deploying the most advanced driver assistance in series production vehicles in complex urban driving environments, a huge leap forward as we work to build our autonomous driving capabilities.
The underlying strength of this holistic platform will support our exponential growth, expected in the next phase of the race.”
The XNGP features significant hardware upgrades: 508 TOPS of computing power, dual LiDAR system, 8-megapixel HD cameras and a new XNet software architecture, backed by an AI system and closed-loop, self-evolving data.
XNet, the next generation software architecture based on neural networks
XNet adopts a self-developed deep neural network for visual recognition with human-like decision-making capabilitiesbased on data from multiple cameras, unlike the first-generation visual perception architecture.
XNet is backed by Fuyao, China’s largest supercomputing centre for autonomous driving. Supported by Alibaba Cloud’s intelligent computing platform, Fuyao’s supercomputing capacity can reach 600 PFLOPS, which they claim: “increases the efficiency of the autonomous driving model by more than 600 times. That is, the time taken can be reduced from 276 days to 11 hours to train a model.
With next-generation deployment optimisation, XNet uses only 9% of the Orin-X chip’s processing power, compared to 122% before optimisation.
AI platform and closed-loop data
Xpeng has created China’s first closed-loop autonomous driving data system, covering data collection, labelling, training and deployment.
Xpeng’s AI capabilities enable rapid iterations in training models, solving more than 1,000 cases per year. This closed-loop AI and data system has helped reduce the road NGP incident rate by 95%.
Robotaxi development gets test permit
Xpeng has also reported a new milestone in relation to the development of its robotaxi, as it has managed to obtain permission for road testing. Wu, commented:
“Obtaining the road test permit for our mass-produced commercial vehicles – without the need for retrofitting – is a great achievement.
Our platform-based robotaxi development aims to generate significant cost benefits and ensure product quality, safety and user experience.”
The Xpeng G9 is the first series-production vehicle in China to pass government-mandated closed-loop autonomous driving tests.
Voice Assistant
Xpeng becomes the first automotive company to use self-developed underlying voice technologies, and also the first to incorporate multiple-input multiple-output (MIMO) technology for its next-generation in-car voice system.
Version 2.0 of Xpeng’s voice system recognises commands from all passengers in the cabin and understands instructions across multiple conversation streams with an accuracy rate exceeding 96%, without the need for an activation command. They claim the system takes less than a second from command receipt to execution, providing interface feedback in 245ms and voice feedback in less than 700ms.
The brand has said the system will be standard on all new vehicles in China.
Latest version of eVTOL flying electric car unveiled
Xpeng used the event to unveil the latest version of its eVTOL electric flying car, developed by its subsidiary Xpeng HT Aero. Last year’s horizontal twin-rotor structure has been optimised to a new distributed multi-rotor configuration.
The test flying vehicle has successfully completed its maiden flight, as well as multiple single-engine failure tests. From Xpeng, they explain the eVTOL’s performance:
“In driving mode, it is comparable to any conventional car in terms of functionality and measurements. In flight mode, the flying car is piloted using the steering wheel and right-hand gearshift as controls for forward and reverse, turning, climbing, gliding and descending.”
AI robot enhancements
Xpeng’s AI-integrated robot, popularly known as ‘Pony’, has been updated with an improved design, achieving enhanced movement and locomotion capabilities.
The new design also features a more sophisticated mechanical structure with higher transmission efficiency, a stronger robotic actuator, a high-end automatic-grade computing platform, and a thermal and battery management system.
The original text of this article has been extracted from this link. Thanks.