camera-only-vs-lidar

Purpose

This research examines how Tesla’s camera-only (Pure Vision) approach to autonomous driving compares to LiDAR-based systems like Waymo, including safety records, capabilities, and real-world performance as of December 2025.

Key Findings

Tesla FSD operates at Level 2 (supervised) while Waymo operates at Level 4 (unsupervised driverless)
Tesla’s crash rate is ~60% higher than Waymo’s per mile driven
Waymo shows 79-85% fewer injuries compared to human drivers
Tesla faces multiple NHTSA investigations covering 2.88 million vehicles
Tesla claims FSD will achieve unsupervised operation “in roughly three weeks” (as of Dec 2025)

Tesla’s Camera-Only Approach

Elon Musk’s Rationale

At Tesla’s 2019 Autonomy Day, Elon Musk declared:

“Lidar is a fool’s errand. Anyone relying on lidar is doomed. Doomed! [They are] expensive sensors that are unnecessary. It’s like having a whole bunch of expensive appendices.”

Musk argues:

Humans drive using vision alone, so cars should too
“You can absolutely be superhuman with just cameras”
LiDAR is a “crutch” that creates dependency
Camera-only enables mass-market scaling

Tesla Vision Hardware

Component	Specification
Cameras	8 exterior cameras providing 360° coverage
Current Sensor	Sony IMX963 (5 megapixel)
Previous Sensor	1.2 megapixel (HW3)
Radar	Removed from production in 2021
LiDAR	Never used

Key Decision: In 2021, Musk removed radar from production vehicles against the recommendations of Tesla’s own engineers, opting for “Pure Vision” using cameras and AI alone.

Cost Advantage

Tesla’s camera-only approach provides significant cost benefits:

Hardware cost per vehicle is substantially lower
Enables faster mass production scaling
No dependency on expensive LiDAR suppliers
Waymo’s Generation 5 robotaxi costs ~$200,000 per vehicle

Waymo’s Sensor Fusion Approach

Hardware Stack

Sensor Type	Quantity
Cameras	29
LiDAR Units	5
Radar Sensors	6 (front, side, rear)
Audio Receivers	Multiple

Operating Model

Current Autonomy Level: Level 4 (fully driverless)
Operational Regions: 5 U.S. regions including SF Bay Area (1,000 vehicles) and LA (700 vehicles)
Total Fleet: 2,500 robotaxis (as of late 2025)
Total Miles Driven: 96+ million rider-only miles

Safety Comparison

Crash Rates Per Mile

Metric	Tesla Robotaxi	Waymo	Difference
Average Miles Per Incident	~61,000	~98,600	Waymo 60% better
Total Crashes Reported	N/A	1,267	-

Injury Statistics

Waymo (Peer-Reviewed Study - 56 million miles):

85% fewer suspected serious injuries vs human drivers
79% reduction in airbag-triggering crashes
80% lower injury crash rate overall
88% fewer property damage claims

Tesla FSD Claims (Self-Reported):

1 crash per 6.69 million miles with Autopilot
~5 million miles between major collisions with FSD
~1.5 million miles between minor collisions with FSD

Methodological Concerns: Tesla’s data conflates Autopilot with FSD statistics and relies on self-reported metrics.

Fatalities and Serious Incidents

Tesla (as of November 2025):

772 total Tesla deaths recorded
65 Autopilot-related deaths
2 FSD-specific fatalities

Waymo (2021-2024):

47 total injuries: 3 serious, 6 moderate, 38 minor
Over 50 million driverless miles
Most serious crashes caused by other drivers

NHTSA Investigations (2025)

October 2025 Investigation

Vehicles Affected: 2.88 million Tesla vehicles with FSD
Incidents Identified: 58 crashes
Resulting In: 14 fires, 23 injuries

Focus Areas:

Vehicles running red traffic signals
Lane changes into opposing traffic
Multiple incidents at same intersection (Joppa, Maryland)

August 2025 Investigation

Tesla accused of waiting months to report crashes
Required reporting window: 5 days
Investigation into crash reporting inconsistencies

Legal Outcomes

August 2025: Florida jury awarded $243 million in wrongful death case
Tesla found 33% liable for Autopilot-involved crash
First wrongful death trial loss for Tesla’s ADAS systems

Current Autonomy Levels (December 2025)

Company	Autonomy Level	Status
Waymo	Level 4	Fully driverless robotaxis in 5 regions
Tesla FSD	Level 2	Requires active driver supervision
Mercedes Drive Pilot	Level 3	Conditional automation (highway only)

Tesla’s Austin Robotaxi Pilot

Launch: June 22, 2025 (10 days delayed)
Fleet Size: Started with 10 cars
Current Status: Safety monitor required in passenger seat
Musk’s Claim: Unsupervised operation “in roughly three weeks” (December 2025)

Limitations:

Geo-fenced to specific Austin area
Uses “plenty of teleoperation” for support
Cannot scale to customer vehicles in current form

Expert Consensus

Arguments Against Camera-Only

“Current expert opinion is that the camera-only approach will fail to deliver unsupervised self-driving, and that sensor redundancy is the only logical way to have a safe self-driving solution at scale.”

Key Criticisms:

2D image processing has fundamental limitations
Requires massive computing power and unproven algorithms
No redundancy for sensor failures
Poor performance in adverse conditions (fog, dust, sun glare)

Arguments For Camera-Only

Dramatically lower hardware costs
Massive data advantage (10 million miles/day vs Waymo’s 200,000)
Simpler hardware enables faster iteration
If successful, would enable global scaling

Scalability vs Safety Trade-off

Factor	Tesla	Waymo
Data Collection	10 million miles/day	200,000 miles/day
Fleet Size	Millions of customer vehicles	~2,500 robotaxis
Per-Unit Cost	Low (consumer pricing)	~$200,000
Geographic Coverage	Global presence	5 U.S. regions
Autonomy Level	Level 2 (supervised)	Level 4 (driverless)
Safety Record	Multiple investigations	Peer-reviewed validation

Verdict: How Is Tesla Doing?

The Reality

Tesla’s camera-only approach has not yet delivered on promises:

Still at Level 2 after years of “next year” predictions
Elon Musk has missed unsupervised FSD deadlines for 6+ consecutive years
Higher crash rates than LiDAR-equipped competitors
Multiple ongoing federal investigations
First wrongful death legal liability finding

The Potential

Tesla maintains significant advantages IF camera-only eventually works:

Unmatched training data volume
Lower costs enable mass deployment
Already installed in millions of vehicles
Rapid software iteration capability

Bottom Line

As of December 2025: Tesla is significantly behind Waymo in delivering safe, unsupervised autonomous driving. Waymo operates true robotaxis with validated safety improvements. Tesla’s FSD still requires constant driver attention and faces regulatory scrutiny.

The fundamental question remains unanswered: Can cameras alone achieve the reliability and safety required for Level 4+ autonomy, or is sensor redundancy necessary?

LiDAR Technology Overview - Why LiDAR is considered critical for autonomous systems