Supervised vs. Driverless: The SAE Level Divide
The most critical distinction in the autonomous vehicle space is the gap between SAE Level 2 driver-assist and Level 4 driverless operation. As we assess Tesla FSD capabilities ahead of the robotaxi event, it is vital to separate marketing terminology from engineering reality.
SAE Automation Levels Explained
- Level 2 (Driver Assist): The system manages steering and acceleration, but the driver must remain fully engaged and responsible. (e.g., Tesla FSD Supervised, GM Super Cruise, Ford BlueCruise).
- Level 3 (Conditional Automation): The driver can disengage in specific conditions but must take over when requested. (e.g., Mercedes Drive Pilot in limited traffic jam scenarios).
- Level 4 (High Automation): The vehicle operates without a human driver within a defined Operational Design Domain (ODD). If the system fails, it must achieve a safe state on its own. (e.g., Waymo, Zoox).
- Level 5 (Full Automation): The vehicle operates anywhere, in any condition, without human intervention. This does not yet exist commercially.
Tesla's FSD (Supervised) remains a Level 2 system. Despite the name, the driver is legally and practically responsible for the vehicle. True robotaxi services, like Waymo and Zoox, operate at Level 4, meaning there is no one in the driver's seat to take over.
Tesla FSD (Supervised) Technical Profile
Tesla's approach to autonomy is unique in its reliance on a vision-only architecture. While competitors use LiDAR, radar, and HD maps, Tesla utilizes a suite of cameras and neural networks to interpret the environment.
Vision-Only vs. Sensor Fusion
| Feature | Tesla FSD (Supervised) | Waymo (Level 4) |
|---|---|---|
| Primary Sensors | 8 Cameras (Vision-Only) | LiDAR, Radar, Cameras |
| Mapping | Real-time perception | HD Geofenced Maps |
| Redundancy | Visual confirmation | Multi-sensor overlap |
| Weather Handling | Camera limitations in fog/rain | LiDAR penetration (better) |
The current hardware suite (HW4/AI5) processes over 1,000 frames per second, but the lack of active depth-sensing (LiDAR) remains a point of contention among safety researchers. Tesla's "end-to-end" neural network attempts to mimic human driving by learning from millions of miles of video data rather than following hard-coded rules.
Service Availability: Where Robotaxis Actually Operate
While Tesla owners can use FSD (Supervised) on almost any road, Level 4 robotaxi services are restricted to specific geofenced areas where they have been approved by local and state regulators.
| Provider | City/Region | Service Type | Availability |
|---|---|---|---|
| Waymo | Phoenix, San Francisco, Los Angeles, Austin | Level 4 (Driverless) | 24/7 Commercial |
| Zoox | Las Vegas (Strip), Foster City (Testing) | Level 4 (Driverless) | Limited Pilot/Employee |
| Cruise | Various (Post-2024 suspension) | Level 4 (Driverless) | Phased Re-entry |
| Tesla | N/A (Supervised use global) | Level 2 (Driver Assist) | Available with FSD Purchase |
As of 2026, you cannot hail a driverless Tesla. Any "robotaxi" event or announcement from Tesla refers to future plans for a supervised fleet or a future Level 4 vehicle, neither of which is currently operational for public hire.
Safety Data and Intervention Rates
Safety is the primary bottleneck for regulatory approval. We must look at miles per intervention (MPI) and incident reports filed with the NHTSA.
Real-World Safety Metrics
- Waymo: Internal data suggests one contact with a human remote-assistant every 15,000–20,000 miles. They have completed over 20 million driverless miles.
- Tesla FSD: Tesla claims one accident every 4-5 million miles with FSD engaged, but this includes supervised driving where the human prevents incidents. NHTSA data shows a high volume of Level 2 incidents, though this reflects market share as much as system safety.
- Human Baseline: The human average is approximately one police-reported crash every 500,000 miles.
For a robotaxi to be viable, it must prove it is significantly safer than a human in complex urban environments. Waymo's data suggests it is approaching that threshold, while Tesla's supervised data is less applicable to a driverless use case.
Pricing: Robotaxi vs. Rideshare Economics
The economic promise of robotaxis is the removal of the driver's wage, which typically represents 60-70% of a rideshare fare. However, the cost of Level 4 sensor suites and remote-assistance infrastructure keeps prices high.
| Service | Base Fare | Per Mile | Typical 5-Mile Trip |
|---|---|---|---|
| UberX / Lyft | $3.00 | $1.50 | ~$14.00 |
| Waymo (Standard) | $5.00 | $1.75 | ~$17.00 |
| Tesla (Future Fleet Projection) | $2.00 | $0.80 | ~$8.00 |
Currently, robotaxi pricing is comparable to or slightly higher than Uber/Lyft. Tesla's projected "Cybercab" economics rely on a vehicle cost below $30,000 and a vision-only sensor stack to undercut the market.
Current Limitations and Edge Cases
Every autonomous system has an Operational Design Domain (ODD)—conditions in which it is designed to function. When a vehicle leaves its ODD, it must either stop or hand over control.
- Weather: Heavy rain, snow, and fog obscure cameras and LiDAR, often triggering a "disengage" or service suspension.
- Construction Zones: Unmapped lane shifts and temporary signage remain a challenge for both vision and map-based systems.
- Unprotected Turns: Navigating complex, unmarked intersections with mixed pedestrian and vehicle traffic is a high-intervention scenario.
- Remote Assistance Latency: In Level 4, if the car is confused, it may wait for a remote human to provide a "path" or "permission," which can cause traffic delays.
The Regulatory Landscape
In the United States, the NHTSA sets vehicle safety standards (FMVSS), but the deployment of autonomous services is largely regulated at the state level. California (CPUC and DMV) and Arizona have been the most permissive. For Tesla to launch a driverless service, it would require a new FMVSS exemption if the vehicle lacks a steering wheel and pedals, as well as state-by-state operating permits.
Frequently Asked Questions
Is Tesla FSD actually self-driving?
No. Tesla FSD (Supervised) is an SAE Level 2 system. You must keep your hands on the wheel (or apply torque) and your eyes on the road. You are legally responsible for all driving actions.
Where can I hail a driverless robotaxi right now?
As of 2026, Waymo is the only provider offering widespread commercial driverless service in Phoenix, San Francisco, Los Angeles, and Austin. Zoox is in limited pilot phases.
How does Tesla's camera-only approach compare to LiDAR?
LiDAR provides precise, active depth mapping that works in low light, while cameras rely on ambient light and sophisticated AI interpretation. Most safety experts argue that LiDAR provides necessary redundancy for Level 4 operation.
What happens if a robotaxi gets confused?
Level 4 vehicles are programmed to "fail safe." This usually means pulling over to the curb or stopping in a lane with hazards on, then contacting a remote support center for guidance.
