Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizens Wu Yanzu in South China for the delivery of clues! CTOnews.com August 16 news, today, Zhiji held the "IM AD DAY Zhiji Intelligent driving Conference" in Tsinghua Science Park, during which it announced its NOA intelligent driving plan roadmap, which claims to take two years to complete the competitor's nine years of smart driving.

The first city of the expressway NOA:2023 landed in April of the year → radiated 333 cities nationwide in December 2023.

Urban NOA:2023 began its internal test in April, and → was officially tested in Shanghai in October 2023, entering the "urban NOA era" at the same time.

Go to Gaojing Map NOA:2023 started internal testing in April, → started public testing in September 2023, and entered the era of "no map NOA"

Commuter mode: open all cities in 2024

Door to Door (full scene commuter): landing in 2025, suitable for most scenarios.

According to reports, the NOA-assisted driving function, which does not rely on high-precision maps, will start public testing in September 2023, and urban NOA will be officially tested in October 2023; at the same time, the upcoming launch LS6 will have the ability to go to high-precision map NOA, and the delivery will be equipped with one-click scene driving function.

The official also announced the timetable of IM AD intelligent driving assistance products: in 2024, Zhiji IM AD's commuting model will cover more than 100 cities across the country, and it will go to Europe in the same year. It is perceived that European localization adaptability training has been completed, and the planning algorithm has been adapted to the local environment and driving habits.

At the same time, Zhiji Automobile also announced that it will launch a new Zhiji LS6 at the Chengdu Auto Show.

Zhiji LS6 will have the ability of "no map NOA", and the delivery will be equipped with "one-click scene driving" function, including "one-click sticking", "one-click tracking", "one-click relief", "one-click parking" and so on. The "one-click" function of more scenes is also under development.

CTOnews.com previously reported that SAIC Zhiji LS6 has now completed the declaration of the Ministry of Industry and Information Technology. According to the declaration information of Zhiji LS6, the length, width and height of this model are 4904 mm, 1988 tick 1669 mm, wheelbase 2950 mm, positioning medium and large SUV, 800V silicon carbide platform, 100-degree ternary lithium battery pack, and the dual-motor version officially lasts 706km.

In terms of power, Zhiji LS6 can choose single motor and dual motor version, with a maximum speed of 252km / h and zero acceleration of 3.5s. The battery is provided by SAIC era, a joint venture between Ningde Times and SAIC.

In addition, Liu Tao also released the charging data of Zhiji LS6 in the form of a personal test at the end of last month. He tested it on a 1000V / 600A liquid-cooled overcharge pile and finally took 19 minutes to replenish energy of 547km (SOC from 2% to 80%), during which the maximum charging power reached 388kW.

Based on the cutting-edge technological breakthroughs of intelligent driving, such as BEV+Transformer, Occupancy Network occupying network, non-graph NOA, commuting mode and so on, Zhiji Automobile and Momenta have launched three fully data-driven technology solutions: DDOD, DDLD, DLP:

DDOD (Data-Driven Occupancy & Object Detection, data-driven occupancy and object detection) DDOD can be understood as Zhiji's version of "Occupancy Network occupying the network".

It uses centimeter grid, without Lidar to solve the ground protuberance, road teeth and other general obstacle perception, can detect ground elevation information, omni-directional perception of the surrounding environment information, to provide users with a more extreme experience.

DDLD (Data-Driven Landmark Detection, data-driven landmark detection) DDLD is the key technology to realize "no map NOA" and "commuter mode".

Based on the data-driven method, it builds maps in real time in the process of vehicle driving, and on the basis of real-time mapping superposition, fuses multiple maps to form accurate recognition of road features, and can complement the road topology like human beings, improve the traffic capacity of complex intersections and other scenes, and get rid of the dependence on high-precision maps. At present, DDLD already has millions of kilometers of training data.

DLP (Deep Learning Planning, planning algorithm based on deep learning) DLP adopts deep learning, planning model based on Transformer architecture, and vehicle-cloud closed-loop iteration to establish an efficient understanding of the scene and other vehicle behavior, and significantly improve the ability to predict complex environmental changes.

DLP can plan smart driving behavior in advance, achieve more human-like intelligent driving, create a safe and comfortable experience, start and stop silky, change lanes decisively, without setbacks, and infinitely close to old human drivers.

Zhiji also announced a timetable for the application of the above technologies:

In April of DLP:2023, Zhiji Automobile and Momenta released the first D.L.P in the industry. Artificial intelligence model

In September, DDLD:2023 launched the public trial of the DDLD technical solution without high-precision maps.

In the year of DDOD:2024, it is expected to enter the application stage of DDOD technology scheme based on Occupancy occupied network.

According to reports, Zhiji IM AD is the only self-research manufacturer in the industry that is compatible with Xavier and Orin computing platforms, with efficient algorithm development capabilities:

-through algorithm optimization, the computing requirement is reduced by 90% on the premise of meeting the function growth. At the same time, through model fixed point, model compression, core algorithm automatic tuning and cross-layer fusion, the model efficiency is improved by 500%.

-Global visual fusion can be realized only on Xavier and Orin N limited computing platform.

Only rely on a single Orin X chip and a single lidar, you can cover a full range of city scenes and high-precision maps.

Based on the statistics of the frequency of road collision accidents in China, IM AD million km 0.6 times, human driving 1 million km 1.9 times, IM AD is 3.2 times of human driving safety; for safety class error braking frequency, IM AD million km less than 1, which is 5 times the level of the industry.

In terms of driving performance, the success rate of lane change of IM AD is 98%, which is 10% higher than that of head players; the number of uncomfortable deceleration is only 1.3 times per thousand kilometers, and the comfort is 2.3 times that of head players; the number of safety takeovers is 0.36 times per thousand kilometers, and the fluency is better than that of head players in the industry.

In terms of parking performance, the parking acceptance rate of IM AD is only 0.8%, which is much lower than that of the top players in the industry; the parking success rate is 97%, and the parking recognition rate is as high as 99.4%.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.