In the wave of smart cars, smart driving technology is becoming the core driving force to promote the reform of the automobile industry.
With their innovative ability and keen insight into market demand, some cutting-edge technology companies have gradually come to the fore in the industry and become an important force in leading the trend.
The horizon is undoubtedly one of the best.
On April 24th, Horizon held the launch of “Journey to the High”– 2024 Intelligent driving Technology products Conference.
Based on the era of intelligent driving, Horizon, relying on the forward-looking prediction and deep accumulation of the concept of software and hardware combined with full-stack technology, has launched a new generation of in-vehicle intelligent computing solution Expedition 6 series and Horizon SuperDrive ™full-scene intelligent driving solution.
The emergence of Horizon, Expedition 6 and SuperDrive Advanced Smart Drive is the phased revelation of the full stack technical strength of Horizon soft and hard combination and the strategic layout of large-scale mass production, which will further consolidate the leading position of the first camp of its smart driving technology company.
The “new journey” of the horizon, throughout the research and development of the family calculation scheme of the horizon journey, actually has a relatively clear context, which is deeply in line with the trend of the development of the industry.
Journey 2, focusing on intelligent driving assistance and active safety.
journey 3, leading large-scale mass production of high-speed NOA and parking.
journey 5, leading the market of high-level intelligent driving domain-controlled computing solutions.
journey 6, covering low-to high-order full-scene intelligent driving mass production demand, realizing the key transition from single-point to serialization.
The advent of Expedition 6 series is just the right time.
According to the data from the intelligent driving configuration database of the Global Automotive Research Institute, the standard number of L2 (including L2 + and L2 +) in the new energy passenger car market in 2023 will reach 3.
7 million, with a market penetration rate of more than 51%.
When the high-level smart drive ushered in the inflection point, the intelligent driving market actually presents a state of “mixed” development: there is not only the internal competition of urban NOA technology, but also the momentum of L3-level technology.
at the same time, the low-level smart driving function is also continuing to open up the market.
The coexistence of multi-level and multi-technology routes reflects the diversified needs of the market and users.
For automobile companies, it not only requires that the computing scheme can flexibly adapt to different environments and provide corresponding perception and decision support, but also needs to be open and easy to use.
it can efficiently support the development and deployment of various intelligent driving applications for different positioning vehicles.
The most important thing is to strike a balance between performance and cost-effectiveness.
There are six versions of the journey 6 series, including journey 6B, journey 6L, journey 6e, journey 6m, journey 6H, journey 6p, from all-in-one machine to domain control, from active security to full-scene intelligent driving, it can cover all-round intelligent driving full-scale applications, and flexible configuration of computing schemes for different intelligent driving scenarios.
Map source: horizon, in which Expedition 6B is mainly aimed at the low-level smart driving market and is committed to building the most cost-effective active security all-in-one solution in the industry.
For the mid-tier smart driving market, Horizon offers a range 6m to support cost-effective urban solutions, and a range 6e to support high-speed NOA, as well as AEC-Q104-compliant SiP modules and Matrix 6 domain control reference designs to achieve lower power consumption and better system costs with ultra-high integration.
For the high-end smart driving market, Horizon launched range 6P, which is designed for the new generation of full-scene intelligent driving, which has six product features: high integration, high computing power, high efficiency, high processing power, high access capability and high security.
A single flagship of Journey 6 can support full-stack computing tasks such as perception, planning, decision-making and control, as well as full-scene NOA functions.
If there are only many kinds, it is not enough to be called “serialization”.
More importantly, Expedition 6 has a unified hardware architecture, a unified tool chain and a unified software stack.
It has the characteristics of generational consistency, intergenerational compatibility, system optimization, and relies on a platform-based scalable computing architecture, as well as a consistent, complete and mature intelligent driving mass production development platform.
Let’s look at it one by one: a unified architecture means that car companies can more easily integrate and upgrade their systems, helping to reduce development costs and improve production efficiency.
The unified tool chain and software stack provide a stable development environment for developers, reduce the compatibility problems in the software development process, and improve the development efficiency.
Same-generation consistency and intergenerational compatibility means that in the same generation of products, different models can share the same technology and resources, while ensuring a smooth transition of technology between different generations.
Platform scalable computing architecture design allows customers to expand functions and services as needed in the future, ensuring long-term sustainability and the possibility of technology upgrading.
A consistent, complete and mature development platform provides a comprehensive development environment, which enables automobile companies to carry out efficient application development and deployment on a mature platform.
The demand of smart car products shows differential development, and the essence of volume value and volume price comes from the bottleneck of efficiency, and the key to solving the problem lies in the reconstruction of the computing base.
The computing scheme based on unified hardware architecture, tool chain and software stack is the ultimate answer.
Based on the above characteristics, Expedition 6 will support its customers to rapidly develop and deploy smart driving applications, respond quickly to market changes, speed up time to market, and gain a head start in a highly competitive market.
At present, China’s intelligent driving market has entered the next competition cycle with “mass production on a large scale” as the core.
For local suppliers, the next few years will be the key period for L2 smart driving products to win on a large scale and to compete with the L2 smart driving market, which will compete with rapid mass production and cost reduction, and achieve better functional experience at a limited cost.
If the journey 6 series corresponds to the mass production demand of intelligent driving in various scenarios in today’s market, then under such diversified needs, the urgent question that the industry needs to consider is what exactly is the smart driving experience that users really need? And how to build the next generation of more universal high-level intelligent driving system.
Because of the diversity and complexity of user needs, most of the solutions on the market are not satisfactory.
Especially in the current urban NOA volume, the urban NOA of most car companies is still in its infancy and can not meet the needs of users for smart driving.
Generally speaking, the vast majority of current users do not really need to achieve self-driving, safe and comfortable, cheap and reliable, intelligent and efficient experience is the core demand.
Standing at the important node of intelligent driving moving towards the large-scale popularity of the whole scene, the horizon believes that the intelligent driving system will experience from available to goodUse the three stages of development of love.
However, the current mass-produced high-level intelligent driving system represented by urban NOA has not fully reached the standard of “availability” in the eyes of users.
Source: horizon, another highlight of the Horizon launch-SuperDrive, the full-scene intelligent driving solution, provides a new solution to create an “easy-to-use” smart driving solution.
According to reports, SuperDrive focuses on the personification experience breakthrough.
With the end-to-end perception architecture of dynamic, static, OCC (Occupancy occupied network) and data-driven interactive game algorithm, SuperDrive can take into account scene pass rate, traffic efficiency and behavior personification in different road environments.
It can also calmly deal with complex urban scenes such as congestion confluence, intersection interaction-dynamic Driveline, comity to cyclists, congested lane change, urban roundabout traffic, and so on.
Since 2023, “end-to-end” has become the absolute “industry top” in autopilot.
Tesla’s latest version of FSD Beta V12 upgrades the city street driving stack to an end-to-end neural network.
At present, the bottleneck in the development of end-to-end autopilot technology is still the traditional three major problems-algorithm, computing power and data.
Among them, the computing platform needs to support the network in the end-to-end autopilot system on a larger scale.
the data level needs to be massive and very diverse, which can cover most of the case data in ordinary users’ life scenarios.
and the algorithm needs to explore an end-to-end autopilot system with interpretability and strong intervention.
Based on the end-to-end architecture of dynamic, static and occupancy, the occlusion call rate is increased by 70%, the number of dynamic lines of code is reduced by 90%, and the network load is reduced by 50%.
It can effectively solve the problems of high latency, many rules and heavy load in the current industry-aware architecture.
In addition, Horizon accumulates an end-to-end deep learning algorithm based on Monte Carlo tree search, which greatly improves the perception and interactive game ability of intelligent driving system in complex traffic environment.
In the congestion scenario, the success rate of lane change is increased by 50%, and the crossing rate is increased by 67%.
, map source: horizon, Horizon’s expectation for SuperDrive is to create the best combination of software and hardware with the flagship version of Expedition 6 under the blessing of top software algorithm technology, so that the easy-to-use urban NOA scheme can speed up large-scale production and landing.
For the industry as a whole, SuperDrive is committed to providing high-level smart driving model rooms, supporting the improvement of R & D and delivery efficiency through the flexible opening of full-stack technical capabilities, and accelerating the industry to the end of high-level smart driving.
, soft and hard integration, all-round players, release range 6 and SuperDrive, the horizon is undoubtedly very big.
Generally speaking, Horizon wants to act as a bottom enabler of the intelligent driving track, with the full stack technology of the combination of software and hardware, to accelerate the innovative development of the intelligent driving industry, and the core is to achieve large-scale production.
After several twists and turns in the development of smart driving, the industry has understood that the evolution of intelligent driving ability must rely on large-scale mass production in order to effectively verify the feasibility and reliability of technology.
In fact, observing the various characteristics of Journey 6 and SuperDrive, as well as the previous product R & D layout of the Horizon, all have a clear direction, which is also mass production.
For the field of high-level intelligent driving, the scale production capacity of software and hardware combined with full-stack technology is the key factor to determine the success or failure of enterprise competition.
Large-scale mass production verification is the premise of promoting the coordination of software and hardware and promoting the evolution of intelligent driving ability, while the combination of software and hardware and an open and flexible business model is the guarantee of large-scale mass production.
With the continuous progress of smart driving technology, the trend of integration of software and hardware will be more obvious.
The competition in the future lies not only in a single software or hardware capability, but also in how to tap the full stack technical capability of software-hardware cooperation and establish the core competitiveness of enterprises.
Especially in the era of software-defined cars, traditional development models are also changing.
The mainframe factory and some Tier 1 not only need to coordinate hardware and software suppliers to solve many specific pain points such as compatibility and integration difficulties, but also strengthen supplier management in the whole life cycle of products.
Therefore, the industry prefers partners with comprehensive software and hardware capabilities.
Horizon has always been known by the market as a “software algorithm company under the cloak of hardware”, and it is also positioned as a “market leading provider of advanced auxiliary driving (ADAS) and advanced self-driving (AD) solutions for passenger cars”.
Horizon founder and CEO Yu Kai said: “Horizon starts from the end, and looks forward to laying out the full stack technology of the combination of software and hardware, with software as the traction and hardware as the driver, to create a software-hardware combination intelligent driving system facing to the whole scene and full ecology, and to achieve the ultimate efficiency with the ultimate openness to reach the end of smart driving.
” Horizon’s insistence on soft and hard full stack technology is consistent.
Or end-to-end, Horizon proposed the end-to-end evolution concept of autopilot as early as 2016 and the end-to-end algorithm Sparse4D in 2022.
In 2023, the industry’s first publicly published end-to-end self-driving model UniAD, written by Horizon Scholars, won the best paper of CVPR 2023.
At the same time, in terms of hardware technology, the new generation of intelligent computing architecture BPU Nash, which is specially developed for large parameter Transformer, can create industry-leading computing efficiency with a high degree of software-hardware cooperation, and accelerate the popularization of end-to-end and interactive game algorithms.
Map source: horizon, today, end-to-end has become the focus of high-level smart driving algorithm technology competition, the horizon once again catch up with the times, but also based on the forward-looking layout of many years ago.
The comprehensive capability of the horizon’s software and hardware can also be comprehensively reflected from the products it can provide.
Horizon has always emphasized that its own positioning is Tier2, which can not only provide software and hardware product solutions, but also authorize customers to use Horizon’s algorithm, software and development tool chain.
Of course, only technical products are not enough, there is no matching business model, still can not win the favor of customers.
Horizon provides customers with different cooperation programs with high openness.
SuperDrive supports partners’ mass production delivery with full stack capacity, which is actually the embodiment of the business philosophy that Horizon has always adhered to.
It is precisely because of the diversified software and hardware product portfolio, flexible and open business model, whether car companies or Tier1, whether hardware or software partners, can find a suitable business cooperation model in the development environment and diversified product portfolio provided by the horizon.
The line can build a mature smart driving ecology and constantly expand partners.
Today, the horizon has produced 200 + fixed-point models, and the journey series has delivered more than 5 million pieces, linking more than 30 car companies and 200 upstream and downstream partners to build a mature, efficient and open business ecosystem.
Continue to promote different levels of smart driving system performance improvement and cost optimization.
The newly launched Expedition 6 series already has 10 first batch of mass production cooperative car companies and brands, including SAIC, Volkswagen, BYD, ideal Automobile, GAC GROUP, Deep Blue Automobile, BAIC, Chery Automobile, Xingtu Automobile, Lantu Automobile, etc.
, as well as a number of Tier1, software and hardware partners.
It is reported that the first pre-production model delivery of the Expedition 6 series will begin in 2024, and more than 10 models are expected to be mass-produced by 2025.
At the same time, SuperDrive will cooperate with a number of top Tier1 and auto brands in the second quarter of 2024, launch a standard mass production plan in the fourth quarter, and deliver the first mass production cooperative model in the third quarter of 2025.
At this conference, Horizon has announced a cooperation with Zero Technology and Zero Technology to build a production-level SuperDrive domain controller.
Mass production on a large scale determines the future, and it is not easy for the horizon to achieve what it is today.
If it had not been the route design at the beginning of its establishment, guided by mass production and adhering to the combination of software and hardware, I am afraid it would not have been what it is today and there would be no future.
A very intuitive question: why can Journey 6 be favored by the market as soon as it is released? even when Expedition 6 may still be in the PPT stage, many manufacturers have expressed their willingness to cooperate? It is safe to say that the emergence of Journey 6 is inseparable from the first three generations of product development and the accumulation of 5 million sets of shipping experience.
Horizon, it is not easy to transfer the intelligent driving system with software and hardware from the lab to the mass production model.
This includes not only the ability of system integration, but also the management of cost, reliability and production cycle.
With regard to the development of the intelligent driving system, the horizon also pointed out: “from availability and ease of use to love, it is a comprehensive test of the ability of the software-hardware combination system, which not only requires the profound accumulation and application of forward-looking software algorithms.
” it also tests the mass production efficiency of software and hardware collaboration.
” It is this accumulation that has been verified by the market that makes Horizon gain the long-term trust of customers.
In the smart car market, there are many enterprises that boast themselves as “strength players”.
How many of them have squeezed out the water and put it into practice? The most important yardstick is whether customers and users are willing to pay.
We can’t make a definite judgment about the future of the horizon.
However, at least one thing can be determined that large-scale mass production is the core standard for testing technology leadership, and the advantages of the enterprises that take the lead in mass production are more obvious.
Relying on the rich intelligent driving ecology, the horizon has opened the closed loop of “R & D and application”.
Extensive cooperation between car companies and ecology brings a large number of intelligent driving scenes to the horizon, provides a large number of valuable, cutting-edge, complex and high-value feedback close to the needs of users, and inversely promotes the software development capability of the horizon.
It is precisely because of this closed loop that the horizon is able to evolve and polish both software and hardware in the same frequency, and eventually become an intelligent driving technology enterprise with the ability to combine software and hardware with full stack technology.
Technological research and development and mass production form a positive cycle, the scale of mass production continues to expand, and the leading position of the horizon will remain stable.
With the gradual deepening of the development of intelligent electric vehicles, the cooperation mode of the industrial chain has changed.
How to build a stable cooperative relationship is a difficult problem, which puts forward higher requirements for the stability and expansion of suppliers.
Because of the demand for the stability of cooperation, the industry is more inclined to enterprises with long-term viability.
The moat based on the horizon and the advantage of leading technology, at least in a short period of time, the horizon should not have too many survival concerns, which is the guarantee for the horizon to maintain its development.
Wang Chuanfu, chairman and president of BYD Group, said at the end of the press conference: “BYD started strategic cooperation with Horizon as early as March 2021, and this year, millions of cars have been equipped with Horizon Expedition 2, 3, 5 series, and we have become the largest mass production landing partner on Horizon.
Today, I announce that the BYD model will also carry Expedition 6, and we will continue to work with Horizon to create a safer and more reliable smart driving experience for consumers.
We will make every effort to promote the popularity of high-level smart driving and the development of intelligent cars in China.
” Horizon, the road to high-level intelligent driving has just reached a turning point.
As one of the head suppliers of domestic intelligent driving technology, Horizon has won tickets to the finals.
The “horizon” of the horizon may be much farther than we thought.
After all, what Horizon needs to do in the era of smart cars is Intel+Windows in the era of PC and ARM+Android in the era of mobile terminals.
, return to the first electric network home page >.
