The long-running debate over how to best achieve full vehicle autonomy—through cameras alone or a combination of multiple sensors—has resurfaced, with RoboSense founder Steven Qiu firmly taking a position against Tesla CEO Elon Musk’s “vision-only” approach.
Speaking to Business Insider on the sidelines of the FutureChina Global Forum in Singapore, Qiu argued that a vision-only system, such as the one Musk champions for Tesla’s Full Self-Driving (FSD) vehicles, cannot guarantee the safety required for higher levels of automation. Instead, he insists that the future of self-driving cars depends on a multi-sensor fusion model that integrates LiDAR, cameras, radar, and other sensing technologies.
“There’s been a lot of debate over whether a vision-only or multi-sensor approach is better when it comes to self-driving vehicles in the past 10 years or so,” Qiu said. “But by now, it is clear that everyone understands that a vision-only approach is not safe enough. There are a lot of corner cases that a vision-only system cannot account for.”
LiDAR—short for Light Detection and Ranging—is a laser-based sensing technology that measures distances by emitting beams of light and calculating the time it takes for the reflections to return. It builds a detailed 3D map of a vehicle’s surroundings, allowing for precise detection of obstacles, vehicles, and pedestrians—even in poor visibility conditions.
Qiu, whose company RoboSense has emerged as a global leader in automotive LiDAR production, explained that without LiDAR and other complementary sensors, autonomous driving systems would struggle to handle complex driving environments and unpredictable real-world conditions.
“Let’s say you are cruising on an expressway. If there is a white car that has stopped in front of you, it would be challenging for a vision-only system to tell if it’s a car or a white cloud in the sky,” he said. “Similarly, if you are driving toward a tunnel, the system may not be able to tell if there’s a black car driving ahead of you.”
According to Qiu, such “corner cases” make it impossible for camera-based systems to achieve higher levels of autonomy beyond Level 2 under SAE International’s standards. Level 2 systems, such as Tesla’s current FSD, require continuous human supervision. To reach Level 3 or Level 4—where vehicles can operate without driver input under certain conditions—manufacturers must integrate LiDAR and radar alongside cameras.
The SAE automation scale ranks vehicles from Level 1 (basic driver assistance, such as lane keeping or adaptive cruise control) to Level 5 (full autonomy in all conditions). Tesla’s FSD remains a Level 2 system, even as Musk repeatedly promises that full autonomy is on the horizon.
Qiu’s criticism comes amid a growing sentiment within the automotive industry that Tesla’s reliance on cameras alone—without radar or LiDAR—could limit its progress toward genuine self-driving capability. While Musk argues that artificial intelligence and vision-based systems will eventually outperform human perception, many engineers and automakers disagree.
RoboSense’s rise underscores this divergence in strategy. Founded in 2014, the company has become the world’s leading supplier of LiDAR sensors for passenger vehicles, according to Yole Group, a market research firm that published its findings in March 2024. RoboSense’s technology is already deployed in a wide range of vehicles and robotics applications—from Waymo’s autonomous taxis to household robot vacuums and even smartphone cameras.
Musk, however, has long maintained his opposition to LiDAR. During Tesla’s “Autonomy Day” event in April 2019, he called the technology “expensive and unnecessary,” insisting that visual recognition powered by neural networks would eventually be enough to safely drive a car.
“In cars, it’s friggin stupid. It’s expensive and unnecessary,” Musk said at the event. “Once you solve vision, it’s worthless. So you have expensive hardware that’s worthless on the car.”
He did acknowledge that SpaceX’s Dragon spacecraft uses LiDAR for docking maneuvers at the International Space Station, but he maintained that the technology had no place in consumer vehicles.
In recent years, Musk has reiterated that view, dismissing LiDAR as an inefficient cost burden. But the economics of LiDAR have changed drastically. Qiu pointed out that prices have fallen from around $70,000 per unit to just a few hundred dollars, while the technology’s performance has improved dramatically.
That cost reduction has made LiDAR systems more accessible to mass-market automakers. Ford, Volvo, Mercedes-Benz, and several Chinese electric vehicle makers have already incorporated LiDAR sensors into their semi-autonomous models.
Earlier this month, the U.S. National Highway Traffic Safety Administration (NHTSA) opened a sweeping federal investigation into potential safety defects in Tesla’s FSD system, after a growing number of crashes and traffic violations allegedly linked to the technology.
According to the NHTSA’s Office of Defects Investigation (ODI), the probe follows 44 reported incidents in which Tesla vehicles using FSD were said to have run red lights, veered into oncoming traffic, or committed other unsafe maneuvers that led to collisions, some resulting in injuries.
Ford CEO Jim Farley has openly contradicted Musk’s stance, calling LiDAR “mission critical” at the Aspen Ideas Festival in June.
“A reflection on the back of a truck or the sun in the camera’s eyes where the camera will be completely blinded—the LiDAR system will see exactly,” Farley said, highlighting how LiDAR provides a fail-safe layer of detection when cameras fail.
Similarly, Li Xiang, CEO of Chinese EV manufacturer Li Auto, emphasized that local driving conditions in China make LiDAR indispensable. At Li Auto’s “AI Talk” event last year, Li argued that Musk’s perspective overlooks the complexities of Chinese roads.
“If you drive in China at night, you will often see trucks with broken taillights, or even trucks without working taillights, just parked on the road,” he said. “Existing camera systems would not be able to detect these trucks from afar. I believe that if Musk were in China, and driving on various highways late in the night, he would choose to include LiDAR as well.”
Rivian CEO RJ Scaringe also made it clear earlier this month that the company’s approach to autonomous driving sharply differs from Tesla’s, emphasizing that LiDAR remains a critical and safer technology in building reliable self-driving systems.
