On October 24th, the annual Xiaopeng Automotive Technology Day was held as scheduled. Xiaopeng Automobile has brought out a wealth of technological "dry goods" and announced its latest progress in multiple fields such as intelligent driving, intelligent cockpit, electric drive system, electronic and electrical, intelligent manufacturing, flying cars, robots, etc.
Xiaopeng Automobile Chairman and CEO He Xiaopeng stated that the automotive category will gradually evolve from software defined cars to the era of AI defined cars. Xiaopeng Automobile has deeply applied large models to multiple business sectors and technology research and development fields. It is worth mentioning that cost reduction under AI empowerment is also a topic that has been constantly mentioned at this press conference.
Intelligent driving: achieving full network coverage of XNGP with light radar and heavy vision by 2024
At this press conference, Xiaopeng Automobile once again strengthened its intelligent driving label. As urban intelligent driving enters the stage of urban racing, Xiaopeng Automobile announced its intelligent driving goals for 2024 and the technical foundation to achieve this goal.
According to Li Liyun, the person in charge of autonomous driving at Xiaopeng Automobile, Xiaopeng's architecture for full scene intelligent driving, XBrain, consists of modules such as deep visual neural network XNet 2.0 and neural network-based regulation XPlanner. XNet 2.0 integrates high-precision pure visual occupancy networks, achieving the integration of dynamic BEV, static BEV, and occupancy networks. It is the industry's first perception architecture with spatiotemporal understanding capabilities for application large models. At the planning and control level, XPlanner based on neural networks can combine time series continuous analysis of motives at or above the minute level, and adapt to the surrounding environment information in a timely manner to generate the optimal motion trajectory.
Based on this, XNGP urban intelligent driving using a light map solution will not be limited by the scope of use and update time of high-precision maps. At the same time, Xiaopeng has also set the goal of light radar, reducing reliance on LiDAR, and will remove two forward radars from the upcoming Xiaopeng X9, reducing overall costs by half, but the XNGP capability will not decrease. According to the plan, by 2024, XNGP will achieve full coverage of the road network in major cities across the country (including Class 1-4 roads), and by the end of 2024, NGP will cover the European market. Previously, Xiaopeng Motors had announced that XNGP urban navigation assisted driving coverage would reach 25 cities in November and open to 50 cities by the end of December.
On the evening of the event, Xiaopeng's "AI Valet" launched a small-scale test and will be fully opened in the near future. After activating the AI proxy driving function, users can set the starting and ending points nationwide. After manual driving, a memory map can be formed. When selecting the travel route later, the "AI proxy driving" function can be used to achieve city navigation assisted driving for individual routes/specific scenarios.
Smart cabin: XGPT Lingxi large model boarding
At this press conference, Xiaopeng Motors released Xiaopeng's fifth generation intelligent cockpit system XOS. Tianji System XOS Tianji is a cockpit system developed by Xiaopeng, which deeply combines intelligent driving capabilities and next-generation intelligent cockpit application scenarios, serving the era of human-machine co driving. The system will be equipped with Snapdragon 8295 chip for the first time, with a new visual upgrade, dynamic upgrade, and significantly improved interaction efficiency. Supporting split screen multitasking, XDock, full scene car sensing SR, and intelligent voice access to the XGPT Lingxi big model developed by Xiaopeng. The XOS Tianji system will debut on the Xiaopeng X9, and G6G9P7iP7 will also be upgraded to XOS Tianji through OTA.
It is worth mentioning that Xiaopeng Automobile has independently developed the XGPT Lingxi large model and integrated it into the voice system. In addition to using large models for training and improving intelligent driving ability, as well as enhancing the understanding and reasoning ability of intelligent speech, Xiaopeng Automobile applies large models to vehicle design and code development, empowering multiple business sectors and research and development fields to further improve efficiency.
Among them, Xiaopeng Automobile has released its first cockpit integrated on-board computing center, with the XEEA 3.5 electronic and electrical architecture as the core, integrating intelligent driving, cockpit, instruments, gateway, IMU, power amplifier and other functions. It is designed for the future era of autonomous driving and general artificial intelligence, with a 50% improvement in performance compared to the previous generation of central computing architecture and a 40% cost savings.
Intelligent Manufacturing: Next generation integrated die-casting technology reduces costs by 40%
In terms of intelligent manufacturing, according to Xiaopeng, Xiaopeng Automobile is the only brand in China that has integrated front and rear die-casting body production. It has the only integrated die-casting machine with a tonnage of up to 12000 tons in the world, and is about to expand a 16000 ton die-casting machine to support the production of larger die-casting parts. It is researching and exploring the integrated floor die-casting technology in CIB+, which will increase the vehicle's endurance by 5% while reducing the vehicle's manufacturing cost by 10% -30%.
It should be noted that in 2019, Tesla began exploring integrated die-casting technology and successfully applied it in the production of Model Y rear flooring, reducing 80 parts to 1, reducing body weight by 10%, and reducing costs by 40%. The use of integrated die-casting technology enhances the safety of vehicles and reduces the frequency of daily maintenance. The integrated die-casting technology mentioned by Xiaopeng this time will also achieve further improvement in vehicle endurance and a further decrease in vehicle manufacturing costs in the future.
Travel: Mass production of "Land Aircraft Carrier" in the next two to three years
Besides cars, Xiaopeng has been exploring more possibilities for future modes of transportation. At the press conference, Xiaopeng Huitian clarified two development paths for flying cars, namely land air integrated flying cars and split flying cars.
Among them, the land air integrated flying vehicle adopts a dual mode cockpit for land travel/flight, and has been further optimized based on the design announced in the previous two years. It adopts a supercar design, which has a more technological and aesthetic appearance; Flight systems such as rotors and arms can be fully folded and stored inside the vehicle, greatly reducing the wind resistance coefficient during land travel.
The split type flying vehicle is composed of a land-based vehicle and a flying vehicle. The flying car adopts an extended range hybrid power system, which can continuously replenish energy to the aircraft; 6X6 all wheel drive, rear wheel steering; The land-based body is equipped with a flying body that can automatically separate/combine, and the land-based body can provide energy to the flying body. Xiaopeng Huitian revealed that this product is planned to be mass-produced and delivered within the next 2-3 years.
Biped robot: with terminal tactile perception ability
At this press conference, Xiaopeng Automobile also brought a brand new self-developed bipedal robot. Intelligent robots possess bipedal walking and obstacle crossing abilities, as well as a pair of ultra lightweight humanoid arms and humanoid dexterous hands. The maximum load on a single arm is 3kg, and the maximum terminal line speed is 1m/s. They can provide different grasping and wrapping postures for objects of different shapes, and have terminal tactile perception ability.
This intelligent robot will apply intelligent vehicle homologous technologies such as XNGP, XEEA electronic and electrical architecture, XPower, XGPT Lingxi Big Model, XOS, and attempt to apply them to scenarios such as factory production and sales services.