“Why is NIO not releasing city navigation data when others are?” This is a common question.
In this particularly restless era of intelligent driving, some have claimed, “intelligent driving can be done nationwide.” However, NIO, a new car-making force under much scrutiny, has never made such aggressive claims.
Some say NIO spends too much on hardware, “wasting money on meaningless things.” Has NIO really misallocated its funds in intelligent driving?
What is NIO’s actual intelligent driving capability? Why hasn’t NIO done what everyone else is doing? How will NIO approach intelligent driving? Will the newly released ONVO brand use NIO’s intelligent driving functions? With a series of questions, TMTPost recently had an exclusive interview with Ren Shaoqing, NIO’s Vice President of Intelligent Driving R&D.
It was a weekend morning at NIO’s Haidian office in Beijing. We talked from 10 AM until 12:30 PM. If not for other meetings, the conversation could have gone longer.
Before meeting him, someone told me, “He is a charming genius.” Indeed, Ren Shaoqing, elegant and composed, exuded a unique techie aura, “preferring to talk about technology rather than tell stories.” His colleagues call him “Shaoqing,” and we will refer to him as such.
Ren’s genius reputation may stem from his influence in AI. Before joining NIO, he was a co-founder and director at Momenta. He graduated from the University of Science and Technology of China and the Microsoft Research Asia PhD program; in 2022, he ranked tenth on the AI 2000 Most Influential Scholars list and was listed among Young Chinese AI Scholars. In 2023, he won the Future Science Prize in Mathematics and Computer Science.
Ren prefers not to dwell on past achievements but shares what he is currently working on and his future plans.
Before the Beijing Auto Show, Ren appeared in William Li’s livestream for the first time. He rarely uses Douyin (TikTok), saying, “William Li (NIO Chairman) brought me out.”
Despite being his first intelligent driving livestream, Ren wasn’t nervous, stating, “You have tested NIO’s intelligent driving many times, and you understand its capabilities, so there’s nothing to be nervous about.”
During the livestream, there was an impressive 0.5 intervention, where the system didn’t disengage despite a small steering wheel adjustment. Ren explained that the system tolerates some steering input, facilitating better human-machine interaction.
In simple terms, within a lane, you can input your intention to the car, and the car will input its intention, achieving co-driving. The system doesn’t disengage upon minor steering input.
Ren sat in the back during the livestream and moved to the front passenger seat on the return trip. He asked William Li in the back, “Do you think the car or the human is driving now?”
This indistinguishable experience for passengers is NIO’s goal, emphasizing “human-machine co-driving” for safer driving and a semi-autonomous experience.
On April 30, NIO pushed its NOP+ city navigation feature to all NT2 platform models, becoming the second company after Huawei to deliver this capability. Covering over 240,000 users, NIO leads in user scale compared to competitors Li Auto and Xpeng.
William Li stated that based on user scale, verified usable range, and verified mileage, NIO is undoubtedly in the top tier of intelligent driving. As of April 20, NIO’s NOP+ city driving is available in 726 cities nationwide, covering 99%, with over 840,000 km of verified mileage, significantly surpassing the 400,000 km planned for June.
This progress is due to continuous iteration of foundational intelligent driving capabilities and NIO’s unique swarm intelligence, which accelerates its intelligent driving development.
Swarm intelligence involves user vehicles verifying each road, with distributed and centralized optimization, ensuring safe validation of each route and extensive real-world testing before user deployment.
Swarm intelligence is unique to NIO, with each NT2.0 platform model equipped with four Orin chips, one dedicated to swarm intelligence. Competitors either have fewer chips or only equip high-end models, limiting their swarm intelligence capabilities. This explains NIO’s longer foundational development period and rapid intelligent driving deployment.
Swarm intelligence significantly enhances NIO’s ability to identify risky scenarios, unlike industry norms relying on limited test vehicles. This proactive safety approach addresses user concerns about extreme scenarios.
“We continuously mine high-value data through swarm intelligence, ensuring safer navigation and active safety,” Ren stated.
Despite swarm intelligence’s advancements, current intelligent driving remains at L2+, achieving partial L3 functionality but still requiring driver involvement. Thus, human-machine co-driving remains the primary feature.
Achieving this involves two key aspects. First, making reminder functions more humane. Traditional driver monitoring systems (DMS) use cameras to monitor driver behavior, issuing alerts for unsafe actions. However, Ren believes this is insufficient without vehicle and external environment monitoring.
Ren advocates for a unified model integrating driver, vehicle, and environment states, issuing accurate alerts only when necessary to reduce driver disturbance.
NIO developed an ADMS model that significantly improves effective risk scenario alerts compared to traditional DMS, ensuring safety in human-machine co-driving, reflecting NIO’s original intention for its intelligent driving system.
Second, NIO introduced an intelligent driving score system to guide safe driver behavior through scores based on focused safe driving, proper system response, and good intelligent driving experience.
Before enabling city navigation, users must complete tasks and achieve 100 km of NOP+ driving. The intelligent driving score then controls access to full city navigation features based on safety scores.
Ren believes the score system aims to make users more conscious of their driving habits, enhancing safety.
ADMS and the score system enable ideal human-machine co-driving, ensuring user comfort and safety. Ren emphasizes that safety should precede city navigation speed.
Ren notes a trend towards cost reduction in intelligent driving, with self-developed chips being a key method. At NIO Day 2023, NIO introduced its first self-developed chip, NX9031, targeting efficiency and cost advantages over existing flagship chips.
Ren anticipates rapid industry changes in 2024-2025, driven by advancing hardware and software capabilities.
For longevity, NIO aims to lead hardware by two generations (six years) and software by nine to ten years.
NIO prioritizes safety, aiming to reduce “airbag” costs through technology.
Unified software and hardware architectures, like in the Le道 brand, ensure data sharing and continued software updates, critical for sustained intelligent driving development.
Ren asserts that achieving over 90% capability upon release is essential for user value.
Subscription models, while not widely accepted, are considered the right approach for continuous optimization, beyond one-time hardware and software sales.
Ren also discusses the hype around large models in intelligent driving, emphasizing the need for foundational capabilities before pursuing such models.
NIO’s intelligent driving formula (general capability + road verification/optimization/operation = NOP+) highlights its bottom-up approach, leveraging technologies like OCC 2.0 for high-precision perception.
Ren compares intelligent driving development to Ford’s T-model assembly line, emphasizing the importance of defined interfaces and end-to-end integration for solving complex real-world problems.
Overall, Ren highlights NIO’s focus on foundational capabilities, continuous iteration, and user safety in intelligent driving.
Credit: Tidu
