The recent Xiaomi SU7 accident in Tongling, Anhui Province, has ignited fervent discussions not only about Xiaomi itself but also about intelligent driving (hereinafter referred to as "intelligent driving").

In the realm of automotive safety, a crucial metric known as TTC (Time to Collision) measures the duration from normal driving to an impending collision.

According to Xiaomi's partial driving data released regarding this incident, the vehicle's NOA system issued an obstacle warning 2 seconds before the collision. The driver, in a state of panic, took control of the vehicle but still collided, with the vehicle speed at 97 km/h at the time.

In essence, at a speed nearing 100 kilometers per hour, the driver had a mere two seconds to react.

This incident has reignited the spotlight on LiDAR, an active light-emitting device that uses laser beams to scan the surrounding environment. Unlike visual systems that rely on cameras for passive perception, LiDAR is impervious to ambient light conditions. When FSD entered the Chinese market, Li Xiang suggested that Elon Musk consider incorporating LiDAR into Tesla's technology:

"Currently, cameras can only see a distance of just over 100 meters (effective detection range) in the absence of light. However, LiDAR can detect objects up to 200 meters away, even in complete darkness. At a speed of 130 kilometers per hour, LiDAR can still enable AEB (Automatic Emergency Braking)."

Bohu Finance has analyzed this and previous intelligent driving accidents and identified two common misconceptions among the public regarding this technology.

Firstly, many consumers perceive automakers' intelligent driving systems as highly mature and capable of fully assuming the driver's responsibilities. Numerous online videos depict individuals deceiving automakers' safety designs and completely entrusting driving to intelligent systems.

Secondly, the intelligent driving systems of all current automakers are at the L2 level, more accurately termed intelligent assisted driving. However, this distinction often goes unnoticed in promotional materials.

Furthermore, due to cost constraints, LiDAR, an important safety redundancy, is not widely adopted but rather reserved for a few high-end models. As an affordable alternative, the pure vision route has become the mainstream for budget-friendly models.

While the pure vision approach aspires to an ideal technical state, its implementation is driven by more pragmatic needs. The original intention of autonomous driving is to enhance safety, but market pressures have led to intelligent driving systems falling short of their advertised safety standards.

01 Technological Idealism

Elon Musk's decision to eliminate LiDAR and pursue a pure vision route is rooted in two primary reasons.

Cost is one factor. In the early market, a 64-line LiDAR from American company Velodyne could cost as much as $80,000. However, thanks to increased supply and cost reduction efforts in China, the price of LiDAR has plummeted to $200. Behind this price drop lies the fact that the LiDAR industry itself is not highly profitable. According to a Caijing report, LiDAR manufacturer Hesai Technology stated:

"In the Chinese passenger vehicle market, LiDAR is not a particularly profitable industry. Our currently good overall gross margin is partly due to Hesai's substantial robotics business, which has higher margins. In fact, the gross margin for passenger vehicle-related businesses is not that high."

More importantly, Musk discovered the BEV+Transformer+occupancy network solution for intelligent driving. Theoretically, the best intelligent driving system should mimic human driving, which relies heavily on visual perception.

Previously, the pure vision approach struggled to obtain 3D images through continuous scanning like LiDAR, making accurate distance judgments challenging. However, through large models and occupancy networks, Tesla's pure vision approach can stitch two-dimensional images into a three-dimensional dynamic scene, achieving functionality similar to LiDAR.

Theoretically, the pure vision route for intelligent driving could eventually evolve to mimic human driving. Nonetheless, safety cannot accommodate idealism.

This is not to suggest that the LiDAR solution is superior. According to industry insiders quoted by Jiemian News, while LiDAR's current recognition distance can reach 200 meters, and the theoretical upper limit of sensor perception capability is higher, actual performance hinges on the software algorithm system. "If the algorithm is not well-developed, even with LiDAR, the effectiveness will be limited."

Rather, from a consumer perspective, in complex environments such as night, rain, or heavy fog, solutions equipped with LiDAR offer greater reliability.

Currently, intelligent driving has become a significant factor influencing consumers' car-buying decisions. This underscores the importance of promoting intelligent driving when selling vehicles. This year, the concept of equal rights in intelligent driving has gained prominence.

Many automakers fail to adequately warn consumers about the limitations of their systems during promotions. Some consumers told Bohu Finance that frontline salespeople often claim that their vehicles are infinitely close to L3 capabilities. At press conferences, automakers' overly optimistic promotions can also shape users' perceptions, fostering a misconception of unwavering "safety" and "convenience".

However, it's crucial to note that according to the Ministry of Industry and Information Technology's "Classification of Vehicle Driving Automation," L2 is categorized as "combined driving assistance," requiring the driver to monitor the environment throughout the journey and be ready to take control at any time. In other words, unless it's an L3-level intelligent driving system recognized by relevant authorities, the driver remains the primary responsible party in the event of an accident.

Currently, there are no L3-level intelligent driving systems in China.

02 Navigating Compromises with the Market

Ultimately, the mission of autonomous driving is to replace human drivers due to the various factors that can interfere with human driving ability. A mature intelligent driving system promises safer and more rule-abiding driving, freeing humans from the role of drivers and opening up new industrial opportunities.

Even from the current perspective, intelligent driving offers significant assistance in certain driving scenarios.

In terms of assisted driving, Bohu Finance interviewed users who have utilized intelligent driving systems. They unanimously agreed that these systems significantly reduce fatigue during high-speed driving. For most of the time, they only need to hold the steering wheel and be ready to take over if something doesn't meet their expectations. Some experienced drivers even claimed that in urban areas, intelligent driving systems might perform better than inexperienced or unqualified drivers.

With the growing importance of intelligent driving systems, the concept of equal rights in intelligent driving has come to the forefront. To make intelligent driving systems accessible in lower-priced models, more automakers are adopting the pure vision approach.

A prime example is Xpeng. As a leading automaker in China for intelligent driving systems, before the launch of the Xpeng P7+ in July last year, multiple models, including the G9 and X9, were equipped with LiDAR. However, the P7+ and subsequently launched G6 and G9 2025 facelift versions have switched to pure vision.

More brands choose to differentiate their intelligent driving routes based on different price points. For instance, BYD's Tian Shen Zhi Yan system offers two schemes: A and B, which include LiDAR, while the basic version C scheme employs pure vision.

This is because compared to LiDAR-equipped solutions, pure vision solutions can reduce costs by 30%. Whether or not to install LiDAR and how to integrate it have become strategic choices for automakers.

The good news is that the strength of China's supply chain and the intense competition in the Chinese market are accelerating the popularization of LiDAR. In April last year, RoboSense released a new automotive LiDAR product, the MX, priced at only $200. Shortly after, Hesai announced the ATX, which is sold for less than $200. This move was seen by the industry as the dawn of the "technological equal rights" era for automotive LiDAR.

The combination of cost reduction in technology and scale effects has enabled domestic LiDAR companies to gradually find their footing. In 2024, Hesai Technology reported an overall gross margin of 42.6% for the year. Another competitor, RoboSense, achieved an overall gross margin of 17.2%, up from 8.4% in the same period last year.

We are on the cusp of a safer future, one that balances technological idealism with market realities.

References:

Caijing: LiDAR Monopoly Pattern Stabilizes, How Will Leading Players Fight the "Growth Battle"?

The cover image and accompanying images are owned by the respective copyright holders. If the copyright holder believes that their work is not suitable for public browsing or should not be used free of charge, please contact us promptly, and we will immediately make corrections.