Experience Lab
Explore the realities
Physical AI learns from.
Robots and autonomous systems need experiences that are too dangerous, expensive or rare to collect in the real world. Explore how simulation creates those experiences, reveals what sensors miss and helps test what AI has learned.
Play, pause and compare recorded simulations. Each example connects the physical world to a learning or testing task.
Established physics.
More experience.
- Configure the simulator
- Vary the conditions
- Run scenarios in parallel
- Use the data to train and test
Learn to act.
Practice a decision, then compare what happens.
A grip can hold tissue, or damage it.
A robot needs practice with materials that behave differently. Virtual tissue lets it make mistakes before touching anything real.
Compare the grips. Watch how the tissue responds as each controller pulls.
Recorded simulation
Watch the grip change the outcome.
Compare a fixed grip with the learned controller on the same virtual tissue.
Recorded outputs, replayed in your browser. No new simulation is running.
A learned controller adapted to unfamiliar virtual tissue, completing more successful pulls than a fixed grip.
Simulation results. No clinical validation.
A clear route can still lose its signal.
Buildings create radio dead zones that a drone cannot see from the street layout. Simulation gives a model examples of where those gaps appear.
Compare routes, then switch between calculated coverage and the AI prediction.
Recorded simulation
Follow the signal along the route.
Compare the shortest route with a route chosen using predicted radio coverage.
Recorded outputs, replayed in your browser. No new simulation is running.
A route that accounts for radio coverage can keep a drone connected more often, even when it takes a longer path.
A recorded simulated mission. No real-flight validation.
A turbine changes the wind for the turbines behind it.
Turning a turbine can redirect its wake. Simulation lets a model learn the tradeoff between its own output and the rest of the farm.
Compare turbine settings, then see how the predicted gain holds up in the physics solver.
Recorded simulation
Check what the turbine settings deliver.
Choose a recorded wind condition. Compare the model's prediction with the power computed by the physics solver.
Recorded outputs, replayed in your browser. No new simulation is running.
A learned predictor helps select turbine settings. Fresh simulations reveal when the expected improvement holds and when it does not.
Recorded steady simulations. No operating wind-farm validation.
Predict what happens next.
Generate unfolding hazards that are too dangerous to recreate for every test.
Danger spreads beyond the room you can see.
Heat and smoke can make a familiar building impassable. Simulated fires reveal how hazards move beyond a robot’s sensors.
Move through the fire. Switch between temperature and visibility across the rooms.
Recorded simulation
Watch the hazards fill the rooms.
Move through the recorded fire. Switch between heat and visibility.
Recorded outputs, replayed in your browser. No new simulation is running.
Varied simulated fires create experience for learning how hazards spread and testing responses to unfamiliar conditions.
This replay shows a physics simulation. Learned predictions are evaluated separately.
Seeing a fire now does not tell you where it goes next.
A spreading fire can close a route before anyone reaches it. Simulated fires give a model examples of how the danger develops.
Compare the physics model with the AI forecast. Watch the fire spread and reveal the routes.
Recorded simulation
See where the fire reaches next.
Move through the recorded fire. Compare the physics model with the learned forecast.
Recorded outputs, replayed in your browser. No new simulation is running.
Simulated fires on real terrain teach a model to anticipate how a fire could spread.
The simplified physics baseline was faster in this example. These routes are not validated for evacuation.
The reality stack
Give your model
more experience.
Laplace turns established physics simulators into engines for generating training and testing data. Our ambition is to build the reality stack intelligent machines learn from.
Explore earlier work in materials, weather, chemistry and hidden structures.
Visit the Research Hub ↗Research Hub
Follow the process.
See what holds up.
From a difficult question to a simulated world, a comparison and a result worth questioning.
Robotics · touch and control
Robotic finger design simulation
Simulate robotic fingers before fabrication. Compare skins, touch sensors and controllers across manipulation tasks.
Explore the simulation ↗Energy · connected-system resilience
Large-scale cyberattack simulation
Simulate cyber disruption and physical failures at a fictional energy company. Compare how defensive responses affect electricity delivery and business operations.
Explore the simulation ↗Energy · finding the cause of a fault
Reactor fault diagnosis simulation
Create hidden reactor faults with similar readings, then test whether an intervention, a longer wait or another measurement reveals the cause.
Explore the simulation ↗Energy · testing diagnostic reliability
Reactor diagnosis stress-test simulation
Stress-test reactor diagnosis across concurrent faults, uncertain plant conditions and different physical models.
Explore the simulation ↗