Researchers test self-driving cars by running their driving software through repeatable virtual scenarios, changing conditions such as traffic, weather, lighting and sensor setup, and examining how the software responds. Simulation makes it practical to probe rare or hazardous situations, but it does not prove a vehicle is safe on its own: it is one part of a broader process that also includes software checks, closed-course tests and public-road testing.
How a simulation test works
- Define the system’s operating limits. Researchers identify where and under what conditions the system is intended to drive, then specify the behaviors it must handle. NHTSA’s preliminary testing framework describes a progression through modeling, simulation, track testing and open-road testing; it does not treat simulation as a stand-alone approval. NHTSA’s automated-vehicle research summaries provide the agency’s framework.
- Choose or create a scenario. A team may use recorded real-world driving as a starting point, replay it with changes, or build a wholly virtual case. Researchers can also generate adversarial variations designed to expose weaknesses that routine driving may not reveal. Waymo’s testing overview describes its use of simulation, while NVIDIA researchers explain scenario generation and characterization in their 2021 study.
- Set the test conditions. The team configures the road, route, other road users, weather, time of day and simulated sensors. Changing one factor at a time—or deliberately combining several—lets researchers compare responses under controlled conditions.
- Run the driving software in a loop. In closed-loop simulation, the virtual environment supplies input to the driving stack, and the stack’s actions affect what happens next. Researchers can inspect perception, decisions and vehicle control, then rerun the case after changing a condition or the software.
- Review the outcome and progress to physical tests. Teams examine whether the system handled the scenario as intended and use results to refine software or tests. Simulation is complemented by unit and integration checks, track work and road testing, rather than replacing them.
What researchers put into virtual scenarios
Simulation can combine ordinary driving conditions with events that would be difficult, costly or unsafe to stage repeatedly on a real road. The open urban-driving simulator CARLA was designed for controlled experiments with configurable sensors and environmental conditions. Its 2017 paper describes settings and interactions such as intersections, construction, pedestrians and conflicting behavior by road users. The CARLA paper is an example of the kinds of variables researchers can manipulate, not a claim that one simulator represents every real-world condition.
As an Amazon Associate I earn from qualifying purchases.
- Road and route: layout, intersections and route complexity.
- Other road users: vehicles, pedestrians and interactions among multiple agents.
- Environment: weather, lighting and time of day.
- Sensors: sensor types and configurations feeding information to the driving software.
- Unusual or adversarial events: variations intended to probe failure modes and the difficulty of avoiding a collision.
One approach to difficult cases is to assess possible safe driving trajectories at different starting times and characterize how hard accident avoidance would be. NVIDIA researchers describe this kind of scenario analysis in their published work. It helps prioritize demanding tests; it is not a universal safety score for every vehicle or simulator.
Free tools Windows power users keep installed
One-click scans. No signup required.
How simulation differs from track and road testing
Each testing method supplies different evidence. Simulation offers control and repeatability; a closed course lets a team test a physical vehicle in a managed setting; public roads expose it to the variability of real traffic and surroundings. NHTSA describes these as stages in a broader testing progression, and Waymo likewise describes simulation, closed-course scenarios and public-road testing as parts of its approach.
| Method | Control and repeatability | Rare hazards | Real-world variability | What it helps assess |
|---|---|---|---|---|
| Simulation | High: conditions can be configured and scenarios rerun. | Can explore hazardous or unusual cases without staging them on public roads. | Represented by the model and scenario; not identical to all real-world conditions. | Software behavior across controlled variations and difficult scenarios. |
| Closed-course testing | Managed physical setting, with scenarios designed for controlled testing. | Some challenging cases can be staged without exposing the public to them. | Physical vehicle and environment, but within a controlled course. | Vehicle behavior in physical tests under planned conditions. |
| Public-road testing | Less controlled and less repeatable than a designed simulation or course. | Some situations occur naturally, but should not be deliberately created unsafely. | Direct exposure to real traffic and surroundings. | Performance amid real-world conditions and variability. |
The table describes the methods’ general roles; it does not establish that every program uses identical procedures. NHTSA’s testing framework and Waymo’s company description both place simulation alongside physical testing.
What simulation can—and cannot—show
A simulator lets researchers systematically repeat a case, vary inputs and compare software responses. That is useful for investigating situations that are uncommon or dangerous to reproduce on a road. But a run in a virtual environment is evidence about the system under the modelled conditions, not proof that the model captures every detail of the real world.
A mileage total alone is not a safety threshold. Waymo reports “more than 20 billion miles in simulation” on its undated company webpage, accessed in 2026. That is Waymo’s own cumulative figure, not an independently verified result or a universal amount of testing that demonstrates safety. Waymo’s testing overview also reports more than 40,000 unique closed-course scenarios; that company-reported count is distinct from simulated miles and should not be read as a general benchmark.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →The practical interpretation is to treat simulation as a way to broaden and structure testing evidence, then evaluate the system through other methods, including physical tests. Neither a single simulation result nor a large mileage figure, by itself, establishes how a system will behave in every real-world condition.
Rank #3
- Product size: 6.69*2.36*1.97 inches Weight: 0.7 pounds Outer packaging: exquisite color translucent packaging box
- Material: Die-cast car model, made of zinc alloy, with plastic parts, rubber tires, detailed interior and exterior decoration.
- Function: Press the two wheels of the front car to trigger the sound of the engine and the lights, with upright car logos,the angel logo is stored in the hood. The details of the body interior are realistic. Place the car on a flat ground, press the body and pull it back, the car will drive forward. (With pull back function)
- This Rolls-Royce Phantom 1/32 car model pull back car is very suitable for children to play with. The paint is exquisite and feels good, and it is heavy to hold. It is also very suitable for car model enthusiasts as a decoration.
- Accessories and collections: This product is very suitable for static accessories and can be used as an accessory next to a computer. For those who are passionate about collecting model car models, this is a great art collection.
What “self-driving” means in safety discussions
“Self-driving” is common reader-facing language, but NHTSA cautions that it can describe a vehicle’s operating state without clearly identifying its capabilities. The agency uses “automated driving systems” for systems designed to perform the driving task without a traditional driver. That is different from consumer driver-assistance features, which do not make a vehicle fully self-driving. See NHTSA’s Automated Vehicle Safety information.
U.S. oversight and public-road testing
In the United States, NHTSA says companies test the automated-driving vehicles they build, must comply with Federal Motor Vehicle Safety Standards and certify that their vehicles are free of safety risks. The agency’s safety information page also describes limited state-permitted public-street testing, research and pilot programs monitored through its Standing General Order. These statements concern the U.S.; rules and testing arrangements differ by jurisdiction. NHTSA’s page references a 2025 USDOT automated-vehicle framework update, so regulatory details can change. The agency’s concise formulation is: “Vehicles are tested by the companies that build them.” NHTSA Automated Vehicle Safety.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →
