ADAS & Autonomous Vehicle International recently caught up with Mr. Yufeng Zhang, vice-president of Horizon Robotics and president of the company’s automotive business unit, to discuss its rapid growth and future vision, as well as the latest on its partnership with Volkswagen to develop driving-assist functions for the auto maker’s electric vehicles.
Please introduce your company and explain its vision/ethos?
Founded in 2015, Horizon Robotics is a leading provider of energy-efficient computing solutions for advanced driver assistance systems (ADAS) and automated driving (AD) for consumer vehicles, with the mission to make human life safer and better.
Before Horizon was founded, we saw two generations of computing paradigms that have already reshaped the human society in a profound way: the personal computer and the smartphone. We thought the next generation would be computing for robotics and that’s how we named our company Horizon Robotics. We are now focusing on the computing for autonomous driving for passenger vehicles, as we think cars will be the first type of robot that impacts everyone’s daily life.
Horizon Robotics is committed to enhancing next-generation driving experiences by integrating hardware and software, including low-power hardware computing solutions and open software development tools. Shipments of our Journey series automotive-grade computing solutions have already passed four million.
Over 100 industry partners and more than 25 auto makers have forged partnerships with Horizon Robotics, signing cooperation on over 150 car models from prominent Chinese automotive OEMs like Li Auto, Chery, BYD and Geely. More than 50 of these models have already been launched and reached full mass-production in China.
What have been some of the company’s most important milestones to date?
In 2016, we launched our first-generation ‘Brain Processing Unit’ (BPU). It is our independently developed computing architecture, specifically for the era of smart vehicles. Powered by BPU, we introduced the Journey series of automotive-grade computing solutions, now in its third generation. Journey 5, launched in 2021, is our latest high-performance, energy-efficient product.
In 2020, our Journey 2 product was deployed in Changan’s popular UNI-T model, marking the beginning of mass production for China’s domestic automotive-grade computing solutions. Journey 5 is now featured in all Pro and Air trim models of Li Auto’s L series, marking a new era of mass produced high-level automated driving functionality.
In October 2022, Horizon and Volkswagen Group announced a partnership. As part of the cooperation, Horizon and VW’s software company CARIAD are establishing a joint venture in China, to speed up the customization of VW’s automated driving solutions for the Chinese market. We are increasingly becoming the partner of choice for domestic and international automotive companies to promote the development of automated driving in China.
You recently released Nash, the latest version of your computing hardware architecture. What are some of its key features and how does it differ from other solutions?
As Yu Kai, founder and CEO of Horizon Robotics, said at the 2023 WNEVC (World New Energy Vehicle Congress) during IAA Mobility, our company is very unique. We are not like a typical hardware design company or software algorithm company. Instead, we take the approach of software and hardware co-design and co-optimization.
BPU is our proprietary data-driven computing architecture specifically for autonomous driving. Leveraging cutting-edge deep learning technology, our BPU delivers optimal efficiency and flexibility and is a powerful engine to accelerate the development of software-defined smart vehicles.
Our latest generation of BPU Nash is designed for high-level automated driving solutions especially in scenarios like urban NOA (Navigate on Autopilot). In urban streets, especially in China, the traffic environment is very complex and requires negotiation between all the traffic participants in real time. Therefore, our BPU Nash is based on interactive decision making to better ensure the realization of urban NOA and the safety of drivers and passengers.
What represents one of the biggest breakthroughs in recent years, regarding neural networks – and what challenges still remain?
One of the biggest breakthroughs, from my view, is the end-to-end autonomous driving framework based on transformer. Traditional autonomous driving systems adopt a modular deployment strategy, in which each function is developed individually and integrated into the vehicle. In contrast, an end-to-end autonomous driving system combines perception, prediction and planning into a single model that can be jointly trained. In this way, the whole system is optimized toward the ultimate task and the computational efficiency is increased.
Of course, there are still challenges for end-to-end autonomous driving, such as interpretability and generalizability. One of the solutions that the industry is exploring is to combine detection, tracking, mapping, trajectory, occupancy prediction and planning all in one transformer-based framework, which not only ensures safety and reliability, but also helps accelerate the commercial implementation of autonomous driving technology.
You also offer driving assistance solutions. Can you tell me about them?
As Horizon Robotics is committed to creating the computational foundation for the era of smart mobility, these driving assistance solutions are developed based on our products, specifically our Journey series computing solutions. Our Journey 2 and Journey 3 can support front camera ADAS (advanced driver assistance systems), and our Journey 3 and Journey 5 can support high-level automated driving functions like highway NOA, assisted driving and parking. These solutions have all been mass produced in Chinese car models, such as Li Auto’s L7, L8, L9 Pro, Neta S and GT, Geely Boyue L, Roewe RX5, etc.
To help our customers, automotive OEMs and Tier 1s, better understand the performance of our products, we demonstrated our market-proven road demos at IAA Mobility this year. We also presented several assisted driving use cases jointly developed with domestic and international industry partners. Together with our partners, we are committed to bringing consumers a safer and better driving experience.
Your company motto is ‘Journey Together’ – can you explain the thinking behind this?
No single company can do everything. Horizon’s business model is to create an open and flexible ecosystem.
We now collaborate with more than 100 partners, including software vendors, hardware component vendors, sensor makers, etc. Most of them are local Chinese companies, but we have also established partnerships with international Tier 1s such as Aptiv, Continental and ZF.
Horizon’s network of cooperation also includes such well-known auto makers as Volkswagen, SAIC, Li Auto, BYD, GWM, Chery and Nio. We work together and share our insights, even our challenges, as a community, hoping to make the daily driving experience for drivers and passengers in China a lot safer, more convenient, and more comfortable.
Just how important is your recent deal with VW/Cariad?
As announced, we’re in the process of establishing a joint venture between Horizon and VW’s software company Cariad, and hopefully we’ll have great news to share with everyone in the near future. This VW partnership is a significant milestone for Horizon and we’re confident that together we will speed up the customization of automated driving solutions for VW in the Chinese market.
We expect our open ecosystem, strong local-supporting capabilities, energy-efficient computing solutions, and user-friendly software development tools will create more opportunities for us to work with global OEMs, who are now competing to accelerate their customization of automated driving solutions for the Chinese market. Our collaboration with Volkswagen Group is a great start. I believe our local DNA positions us to support global OEMs succeed in China.
Finally, where are we currently on the journey to fully autonomous driving?
Over the last few years, the automotive industry has undergone unparalleled changes – from the rapid development of electrification to advances in automated driving. And China has emerged as the world’s most dynamic and progressive market for advanced smart driving solutions.
But Horizon always believes that the goal of technology is not to give power to machines, but to enable human greatness. Now, the primary demand of consumers, from our observation, is not to fully replace human drivers with autonomous technology, but rather to enhance their driving experience and make it safer and more convenient.
For example, in cities such as Beijing and Shanghai, highway NOA at an affordable cost can be a big help to effectively reduce driver fatigue, as most daily commutes are on highways or urban expressways. Full self-driving is still technically challenging and will take a few years to achieve.
StradVision’s SVNet available on Horizon Robotics Journey series
Horizon Robotics recently formed a strategic partnership with StradVision, an automotive industry pioneer in deep learning-based vision perception technology. The collaboration focuses on the production-quality implementation of StradVision’s perception technology, SVNet, on Horizon’s Journey series computing solutions, aiming to provide high-quality driving assistance solutions for Tier 1s and OEMs.StradVision’s SVNet provides cutting-edge vision-based object detection and recognition, even in challenging weather conditions, and has been optimized on Horizon’s Journey computing solutions via its easy-to-use Horizon OpenExplorer software development toolkit. With two market-competitive companies joining hands, the combined solutions offer best-in-class performance, power and cost efficiency in advanced driver assistance systems (ADAS) to Tier 1s and OEMs, and ultimately bring safety and convenience to the end-consumers and their families during their journey.
“The automotive landscape is undergoing a rapid transformation,” said Dr. Yu Kai, founder and CEO of Horizon Robotics, back in September 2023 when the partnership was announced. “StradVision’s production-proven perception solution enables market-competitive ADAS implementation. Through its optimization on Horizon’s energy-efficient computing solutions, we will together bring safer and more competitive products, and promote the digital transformation of automotive industry.”
“Horizon’s Journey computing solutions, renowned for their industry-leading compute efficiency and minimal power consumption, have been meticulously crafted for ADAS/AD applications,” added Dr Junhwan Kim, CEO of StradVision. “This strategic collaboration underscores StradVision’s commitment to our mission of delivering highly adaptable vision perception technology. We empower automotive players to seamlessly integrate our technology with a spectrum of hardware options, thereby expediting the evolution toward software-defined vehicles.”