news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

on august 28, the jiyue end-to-end ai intelligent driving conference with the theme of "evolution·ai reshaping intelligent driving" was held in chengdu. the new v2.0 version of the software was officially released and will be first installed on the jiyue 07. the v2.0 version of the software brings the most important update since the launch of the jiyue model, covering more than a dozen major modules such as end-to-end large-scale intelligent driving, large-scale intelligent cabin, and intelligent car connection, with a total of 947 new functions and optimization upgrades. the pre-sale price of jiyue 07 was officially announced, starting from 215,900 yuan, and the new car will be officially launched on september 10.

ppa intelligent driving has been upgraded to asd (apollo self driving), marking the entry of jiyue's pure visual high-level intelligent driving into the end-to-end large model era. jiyue has become the world's first smart car brand to apply baidu apolloadfm, a large model for l4 autonomous driving. the buyout price of jiyue's asd high-level intelligent driving package is 29,900 yuan, the limited-time discount price is 4,999 yuan, and the latest monthly subscription price is 599 yuan/month. a three-month asd high-level intelligent driving package is also given free of charge to all new and old car owners. for old car owners who have bought out pure visual high-level intelligent driving, jiyue will "make up the difference" on the basis of the current limited-time discount price, truly achieving "increase in quantity, reduce prices, and do not backstab". in addition, jiyue has also kicked off the free prelude of pure visual high-level intelligent driving, giving all users a lifetime free high-speed elevated navigation assistance, opening the era of "high-level intelligent driving for all people".

xia yiping, ceo of jiyue, said: "this is the first time that l4 autonomous driving technology has enabled family cars! it is the deep integration of baidu adfm and jiyue, the world's first large-scale autonomous driving model that supports fully unmanned driving applications. baidu has been continuously researching and developing ai and autonomous driving for more than ten years, and has continuously invested more than 150 billion yuan, which is the key to jiyue's continued leadership. without 50 billion yuan, intelligent driving cannot be done well, and intelligent driving made with only a few billion yuan is a 'road killer'."

comprehensively upgraded asd, the pure end-to-end vision leading the industry by one generation

compared with the traditional intelligent driving of "lidar + hd map + rule algorithm", the new intelligent driving has opened a new stage of intelligent driving development with "pure vision + end-to-end big model". with the rapid evolution of ai big models, the "crutch" role of lidar and hd map will be gradually replaced, and pure vision will continuously improve the level of intelligent driving with its powerful self-learning ability.

based on baidu apollo's l4 autonomous driving atomic capabilities and safety system, jiyue has achieved the only pure vision high-level intelligent driving in china and the only one in the world, firmly ranking in the first echelon of the industry. in this v2.0 version of the software, jiyue ppa intelligent driving has been upgraded to baidu's most cutting-edge intelligent driving technology product asd, which has achieved pure vision + end-to-end large model, leading the industry by one generation. asd, built using the large model apollo adfm of l4 autonomous driving, takes into account the safety and generalization of the technology, and is several times safer than human drivers, achieving full coverage of high-speed and complex urban scenes.

baidu is the world's leading ai big model company and a leader in autonomous driving technology. it has accumulated more than 10 years of experience in ai technology and autonomous driving, and its technical capabilities are world-leading. at present, apollo's l4 autonomous driving mileage has exceeded 100 million kilometers, and a high-computing training cluster of over 5.5eflops has been established, which can achieve efficient training and rapid iteration of models. jiyue will achieve "monthly evolution" based on this, providing an increasingly advanced, intelligent and safe smart driving experience.

the perception capability of jiyue pure vision advanced intelligent driving even surpasses mainstream laser radars. its 3d accuracy can reach the centimeter level, with 360-degree coverage of the vehicle's surrounding environment, a perception distance of 250m, and the ability to identify mineral water bottles within 200m. the accuracy of identifying the speed of moving obstacles is also controlled within an error range of 0.1 m/s. it can also distinguish between green plants, temporary construction fences, faulty vehicles parked on the side of the road, predict the movement of pedestrians, etc. even in rain, snow, fog, haze, or at night, in tunnels and other low-light environments, it can still accurately perceive and identify.

at a time when intelligent driving is reaching new heights, "experience is king". the asd of the jiyue will bring users a smooth intelligent driving experience. with functions such as automatic shifting and automatic turn signals, users can get in and go, making driving easier, just like "buy a jiyue and get an ai driver for free". asd includes a total of 106 new functions and capability upgrades, including navigation assistance, valet parking, advanced lane keeping, and active safety, to achieve comprehensive evolution of the three capabilities of driving, safety, and parking.

in terms of driving capabilities, asd can achieve free movement from point a to point b. lane change and avoidance are combined into one, intelligent speed limit, automatic start and stop at traffic lights, autonomous detour, continuous lane change, courtesy to pedestrians, unprotected left turn and other functions make intelligent driving capabilities fully advanced, and can easily cope with various complex scenes on highways and urban roads. in particular, the super obstacle avoidance ability can accurately grasp the timing of avoidance and make more human-like decisions. at the same time, ai computing power is used to accelerate modeling. it can detect 5,000 collision risks and perform 5,000 distance calculations per millisecond to actively explore the best driving trajectory. the ability to efficiently pass intersections allows the car to learn to actively detour when encountering vehicles occupying the road ahead. when changing lanes and avoiding high-risk vehicles, asd will pay extra attention to keeping a safe distance to avoid them when changing lanes, which enhances the user's sense of security. by the end of 2024, asd can be used wherever there is a baidu navigation map, and it can be driven throughout the country.

in terms of safety capability evolution, the forward aeb capability has been further improved, expanding the effective range, helping users to decelerate to the maximum extent in a timely manner, gaining more safety time and reducing the degree of harm. the new two-stage braking, after optimizing the aeb algorithm strategy, first decelerates and then brakes hard, further improving the safety and experience of intelligent driving. the new cornering aeb function can accurately respond to sudden crossing behaviors such as pedestrians, express trucks, and takeaway trucks in the corners, helping users to stop in time.

parking capabilities have also been greatly upgraded. apa automatic parking assistance not only supports parking in and out of a variety of complex parking spaces such as vertical parking spaces, horizontal parking spaces, space parking spaces, reverse-slope parking spaces, and dead-end parking spaces, but also solves the problem of parking in low light and extremely narrow parking spaces, doubling the parking success rate. in addition, when parking, the rearview mirror will automatically retract to prevent scratches, demonstrating hard-core parking skills comparable to those of an "experienced driver".

the large-scale intelligent cabin has evolved again, with the industry's first lip movement recognition

as early as october 2023, simo took the lead in integrating baidu wenxin yiyan, becoming the world's first voice assistant to apply ai large models, which can be called the "ceiling" of voice interaction in the automotive industry. simo's recognition speed is not only as fast as 500ms, but also can interact smoothly even in weak network/no network environment, and can support the control of almost all functions of the whole vehicle. even solving olympiad math problems and making travel guides are easy, just like an "all-round butler", bringing a more human-like, more natural and smooth human-computer interaction experience.

at this conference, jiyue also released the industry's first native ai big model cockpit. based on the new native ai big model architecture, the underlying capabilities of all cockpits were reconstructed, and a set of end-cloud collaborative ai agent architecture was innovated to model all the cockpit's perception, understanding, execution and other capabilities. it not only retains the ability of the end side to respond quickly, but also integrates the powerful understanding, generation and decision-making capabilities of the cloud big model.

the newly upgraded simo 2.0 is smarter and more humane, subverting the traditional voice assistant's one-question-one-answer command-based stereotyped interaction. simo 2.0 has many capabilities such as fun exploration, role-playing, and scene mode. for example, while driving, you can ask simo about the housing prices in the nearby community, recommend an authentic "fly restaurant", find a mid-point to meet and chat with friends, etc. it is no problem to play the role of a tutor, a car expert, a strategy expert, etc. it also supports 10,000+ user-defined scene settings.

simo 2.0, which integrates a multi-modal large model, has realized the industry's only lip movement recognition function. by integrating and identifying the user's lip movement trajectory, the accuracy of speech recognition in noisy environments is increased to 90%. at the same time, simo 2.0 has also upgraded its dialect recognition capabilities and supports conversations in multiple dialects. with the support of multiple command recognition, multi-zone group chat, and full-scenario voice commands inside and outside the car, it creates a futuristic and immersive smart cockpit for users.

as a transitional product serving "human driving", hud has not been adopted in the jiyue model. the design of jiyue's intelligent cockpit fully considers the popularization of high-level intelligent driving and even the human-computer interaction experience of future unmanned driving. jiyue innovatively integrates lane-level navigation maps with intelligent driving sr information. through a complete reproduction of the physical world, it presents rich elements such as green plants/water systems, refined lane lines/zebra crossings/traffic lights, and even the taillights of the front car on the large screen in a 1:1 ratio, bringing users an immersive navigation experience, which can be called "the next generation of high-definition ar hud". the in-vehicle map has also been upgraded to baidu map v20, realizing "the same navigation for mobile phones and cars". the world's only 3d lane-level navigation covers cities across the country, and it can truly be used across the country.

the ui design is also a highlight of the smart cabin upgrade. the new future theme "era" has been added to the ultra-immersive leisure cabin. the more user-friendly ui design has brought about an improvement in user experience. in terms of entertainment ecosystem services and smart interconnection, the v2.0 version is the first to launch tencent start cloud games, allowing users to play dozens of popular games such as black myth: wukong and genshin impact with one click. the cockpit can also be interconnected with xiaodu ecosystem, smartphones/watch wearable devices, and switch screen projection. the fully renewed app3.0 makes it more convenient to prepare and find a car with one click, bringing users a cutting-edge experience of human-car-home interconnection.