LiDAR vs. Vision-Only: Why a Multi-Sensor Approach Is the Future of Autonomous Vehicles
The race to perfect autonomous driving has ignited a global debate between two major technological philosophies: vision-only systems championed by Tesla and multi-sensor solutions led by companies like RoboSense. Steven Qiu, founder of RoboSense, a leading Chinese LiDAR manufacturer, strongly believes that multi-sensor technology is the safer and more effective path toward fully autonomous vehicles.
LiDAR—short for Light Detection and Ranging—works by emitting laser beams to scan the surrounding environment and measuring the time it takes for the light to return. This creates a highly accurate 3D map of a vehicle’s surroundings. While the technology is already used in robotaxis, robot vacuums, and even smartphone cameras, its role in autonomous vehicles is increasingly becoming essential.
During the FutureChina Global Forum in Singapore, Qiu shared with Business Insider that the debate between vision-only and multi-sensor systems has been ongoing for over a decade. However, he believes the evidence is now clear: relying on cameras alone is not enough to guarantee safety.
“By now, it’s clear that a vision-only approach is not safe enough. There are too many corner cases that a camera system simply cannot handle,” Qiu explained.
The Limits of Vision-Only Systems
Tesla CEO Elon Musk has long been vocal about his confidence in vision-based systems, arguing that cameras—combined with advanced AI—are sufficient to enable full self-driving capabilities. Tesla’s Full Self-Driving (FSD) software currently operates at Level 2 automation, which still requires human supervision.
However, Qiu points out that achieving Level 3 or Level 4 automation—where vehicles can handle most driving scenarios without human input—requires more than just cameras. The SAE International automation scale ranges from Level 1 (basic driver assistance) to Level 5 (fully autonomous driving in all conditions). Vision-only technology, according to Qiu, hits a wall before reaching those higher levels.
He offers a practical example:
“If you’re driving on a highway and a white car stops in front of you, a vision-only system might confuse it with a cloud or a light reflection.”
“Similarly, if you’re approaching a tunnel, it may not detect a black vehicle ahead due to poor visibility and lighting transitions.”
LiDAR’s precision in detecting objects regardless of lighting conditions is what makes it indispensable. Unlike cameras, it doesn’t rely on visual clarity—it measures distance directly through laser pulses, making it ideal for unpredictable real-world driving scenarios.
LiDAR’s Growing Role in Autonomous Vehicles
Founded in 2014, RoboSense has rapidly become a global leader in LiDAR technology. In 2024, it held the largest market share for passenger car LiDAR systems, according to Yole Group. The company’s sensors are increasingly being integrated into production vehicles, signaling a shift in industry trends.
Musk has often called LiDAR “expensive and unnecessary,” famously declaring in 2019, “Once you solve vision, LiDAR is worthless. It’s expensive hardware that adds no value.” But the reality has changed dramatically.
Qiu notes that LiDAR’s cost has plummeted over the years—from nearly $70,000 per unit to just a few hundred dollars. Meanwhile, its performance has improved significantly, making it a practical and scalable option for automakers worldwide.
Why the Industry Disagrees with Musk
Musk’s skepticism isn’t widely shared among industry leaders. Ford CEO Jim Farley, for instance, described LiDAR as “mission critical” at the Aspen Ideas Festival in June. He highlighted scenarios where cameras may fail—such as glare from sunlight or reflective truck panels—while LiDAR continues to perform accurately.
Chinese EV maker Li Auto CEO Li Xiang also pointed out the difference between American and Chinese driving conditions. In China, poorly lit roads and trucks with broken taillights are common at night, creating hazards that vision systems struggle to detect early. LiDAR, however, identifies such obstacles reliably.
“If Musk were driving in China at night on a highway, he’d likely choose to include LiDAR,” Xiang remarked.
This growing consensus suggests that while vision systems play a crucial role, they work best when combined with other sensors. Multi-sensor fusion—using cameras, LiDAR, radar, and AI—may ultimately deliver the reliability needed for safe autonomous driving.
The Road Ahead
As autonomous technology advances, the debate is no longer about whether LiDAR is necessary, but how best to integrate it efficiently. The cost barrier is falling, and industry adoption is accelerating. Companies investing early in multi-sensor solutions are positioning themselves to lead in the next wave of autonomous mobility.
Trenzest Insight
At Trenzest, we explore groundbreaking technologies shaping the future of transportation and mobility. From LiDAR innovations to AI-driven automotive trends, we deliver insights that keep you ahead of the curve.
Visit Trenzest.com for more tech news, industry analysis, and future-driven stories.
