Unlike other car companies’ annual press conferences, for XPeng, the annual tech day on October 24th is the most important event.

On October 12th, XPeng released a poster titled “1024 is coming” to warm up for this year’s tech day. Unlike previous years, this year’s tech day has many new elements.

The Secret of the Poster

Let’s start with the ground. Under the poster is an elevated road that goes from far to near, and the light strips on the side of the road are classic elements that interpret high-precision maps. In addition, the car rushing through the middle indicates XPILOT 3.+, more precisely, the classic interpretation of NGP intelligent navigation assistance driving.

Speaking of which, NGP was the main feature of XPeng’s 2020 tech day, and it was precisely NGP that enabled XPeng to establish a leading position in differentiation among the new car makers in China by tearing apart the path of intelligence.

On both sides of the road are XPeng supercharging stations. In March 2018, XPeng’s first supercharging station was officially put into operation, and today XPeng has become one of the car companies with the largest investment in supercharging construction in China.

As of now, XPeng brand supercharging stations have reached 439, including 27 high-speed supercharging stations. He XPeng said at XPeng’s Q2 financial conference that XPeng will deploy more than 500 supercharging stations by the end of this year, not only continuing to cover the Beijing-Shanghai, Beijing-Guangzhou, and Beijing-Hong Kong-Macau expressways, but also accelerating the charging network layout in third and fourth-tier cities.

Above the poster is a flying object similar to a UFO, but upon closer inspection, it can be identified as XPeng’s Huitian Traveler X2, which just made its debut in July this year.

XPeng has been involved in the field of flying vehicles as early as 2013. So far, it has jointly developed five generations of intelligent electric manned flying vehicles. In addition to two verification machines, XPeng’s flying family currently has three members: the Traveler T1, X1, and X2.The Traveler T1 was launched in 2019 and is mainly used for aerial sightseeing. The Traveler X1 was launched in 2020 and is mainly used for emergency situations like police patrols and rescue missions. Clearly, the positioning of these two products is more inclined to the B-end market.

So, what about the Traveler X2? It can carry two passengers, has a maximum load of 200 kg, a range of 35 minutes, a maximum flying height of below 1000 meters, and a maximum speed of 130 km/h. It can be operated with one-click takeoff and landing and automatic driving.

The above parameters and functions are significantly different from the first two models, and the functionality is much simpler and faster, meaning that the X2 is directed more towards the C-end market.

However, we can see from the picture that it is still an aircraft and is far from the flying car we imagine. Indeed, the X2 has not yet achieved the integration of a car and an aircraft. But according to Zhao Deli, President of XPeng Huitian, the successor to the X2, the next generation of products currently under development have four wheels.

He XPeng revealed during XPeng’s Q2 earnings teleconference that flying cars will be positioned in a higher-end market (500,000-1 million RMB). It is believed that this year’s XPeng Tech Day will bring more news about flying cars.
From cars to intelligent driving, supercharging stations, and flying cars, XPeng is constantly expanding the boundaries of intelligent technology.

After NGP

On October 24, 2019, XPeng opened the first 1024 XPeng Tech Day when it only had one car, the XPeng G3, and the intelligent driving system was only XPILOT 2.5.

But even under the background of XPILOT 2.5, XPeng has made some impressive achievements, such as the industry’s first visual fusion automatic parking. Today, the visual fusion automatic parking is rapidly popularizing throughout the industry, which to some extent reflects XPeng’s foresight.

However, in addition to automatic parking, XPILOT 2.5’s automatic lane change assistance and lane centering maintenance were somewhat unsatisfactory at that time. At that time, the lane centering maintenance feature had restrictions such as the activation vehicle speed not being lower than 60 km/h (since then, OTA), and the turning ability was mediocre. The automatic lane change assistance also had slow lane change times and a low success rate, and the experience was not good enough.

It is precisely these imperfections that have left tremendous room for development for the science and technology day one year later.# Time comes to October 24th, 2020, based on P7 powered by XPLIOT 3.0, XPENG shared the progress of NGP, which is also China’s first integrated navigation-assisted driving system.

On the previous day, Classy Channel conducted an exclusive interview with Wu Xinzhou, Vice President of Autonomous Driving of XPENG and the direct person in charge of NGP.

In the exchange, Wu left many mysteries, such as “the attitude towards LIDAR is open” and “the power reservation of NVIDIA Xavier platform is sufficient and there is no need to worry”, which have been verified in XPILOT 3.5 installed on XPENG P5.

Of course, the most impressive point raised by Wu Xinzhou is still the statement that “if everyone’s ability and efficiency are similar, global companies will not be able to beat companies that only focus on single markets for the time being.” As for this statement, a clearer conclusion may not be reached until five or even ten years later.

In March of this year, XPENG launched an NGP expedition, starting from Guangzhou and heading north to finally reach Beijing, with a total distance of 3,675 kilometers. NGP was used for more than 3,145 kilometers in the trip. According to XPENG’s official data, the average take-over rate of NGP per 100 kilometers was 0.48 times, with a success rate of 97.36% for lane-changing and overtaking, 93.29% for entering and exiting highways, and 98.33% for tunnel passing.

In mid-2021, XPENG’s intelligent driving iteration further accelerated. On June 8th, XPENG launched the Valet Parking Assist (VPA) parking memory function. Without relying on parking lot renovation/navigation/GPS signals, XPENG VPA memory parking can remember up to 1,000 meters and 100 parking lots.

Until now, XPENG is still the only car manufacturer in the industry with the capability of memory parking technology.

Of course, there are still some imperfections in memory parking. For example, after watching the video, you may find that VPA can only operate on a single floor and cannot cross floors.

At the launch event of XPENG P5, He XPeng stated that VPA will be upgraded to VPA-L cross-floor parking memory function through OTA upgrades in the future. Undoubtedly, this is a more close-to-user-scenarios and more closed-loop function iteration.On this year’s October 24th highlight reel, Xpeng P5, equipped with two lidars, will undoubtedly replicate last year’s NGP performance on highways and elevated roads, and paste it onto open city roads.

Prior to this, Xpeng had released a test video of its City NGP, achieving zero handover on open city roads in a journey of nearly 30 minutes.

At this point, the landing and evolution path of Xpeng’s intelligent driving has become clear: starting with the relatively simplest (although not actually simple) NGP on highways and elevated roads, then the memory parking covering semi-closed scenes in parking lots, and the ultimate challenge is the zero-handover assisted driving on open city roads with their much greater challenges.

Combining parking lots, city roads and highways together, which is the real point-to-point intelligent driving, this is an important step toward full automation and the Robotaxi road.

It must be pointed out that although they are all open roads, the challenges in city road scenarios are much greater than those on highways and elevated roads, which also makes me more eager to see the performance of the City NGP, which will be shown on October 24th.

The Third Space

So, what’s next after achieving point-to-point intelligent driving?

At last year’s Xpeng Technology Day, Xpeng also displayed a killer feature alongside NGP: a full-scenario voice assistant. Xpeng’s voice assistant, Xiao P, supports continuous conversation, interruption at any time, semantic recognition and dual-tone zone recognition, greatly enhancing the voice interaction experience during driving.

As an example of semantic recognition, Xiao P filters out invalid sentences and extracts and executes valid commands. In dual-tone zone recognition, it accurately recognizes and ignores other vehicle occupants’ commands except for those of the driver and front passenger.

The usage rate of Xpeng voice assistant has maintained at 99% for a long time, to the extent that Xpeng has stopped announcing it after announcing more than ten consecutive months of 99% usage rate. In the recently released data, the single-month wake-up times of Xiao P in August surpassed 10 million times. Xpeng’s 100,000th vehicle had just rolled off on October 11th, which shows how frequent the usage of Xiao P is.

Why is Xpeng so focused on voice recognition? Voice interaction is a foundational capability that allows Xpeng to attempt to advance to the next level of automotive form, the Third Space, on the P5, based on multi-mode interaction, all-time online collaboration with lower-level operating systems, and rich sensor interfaces.On the first version of the XPeng P5, also known as the XPeng Third Space, the product managers defined five scene functions for users: X-Play entertainment scene, X-Sleep sleep scene, X-GO outdoor scene, X-Life life scene, and X-DIY scene.

Image of XPeng P5

Based on these five scenes, users can enjoy entertainment, sleep, work, eat, and even connect drones with the car system to create more interesting ways to play inside the car.

Today’s XPeng Third Space is similar to XPILOT 2.5, which was mentioned earlier, in terms of automatic lane changing assistance and lane center keeping. They have helped XPeng’s products become more unique, have more differentiated competitiveness, and most importantly, based on their self-developed and continuously iterated capabilities, they are almost certain to become even better over time.

If we expand our scope, let’s think about the X2 traveler on the poster.

He XPeng once discussed the definition of initiative in a public speech: thinking of something first is not enough, you must have the ability to seize the initiative. Without control and stability, you will never be able to do this well. To some extent, initiative is a combination of several abilities.

Therefore, could a XPeng traveler with Xiao P multimodal interaction, autonomous driving, and point-to-fly capabilities become the ultimate intelligent transportation tool for idle times on the road and taking off during traffic jams, and be He XPeng’s next phased goal?

This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.