Foretify Generate
Sensor Simulation
- High-fidelity synthetic data generation, grounded in physics and actors’ behaviors, to train and validate perception and E2E AI models in diverse conditions
- Improve dataset richness with enriched and varied real-world environmental conditions
- An integrated solution leveraging NVIDIA Omniverse and Cosmos Transfer
Real-World Variations
- Create new and relevant variations of the recorded real-world drive logs for training and validation, reducing the dependency on costly physical driven miles
- Improved variability and coverage while maintaining a high-level of realism
ODD Coverage
- Generate targeted scenarios to validate the AI-powered AV stack and expand testing to close coverage gaps
- Accelerate expansion to new geographies with automatic generation of relevant scenarios and distributions
Edge Cases
- Generate rare and dangerous scenarios safely in a virtual environment
- Generate unscripted tests that expose high priority, unknown critical bugs that would have taken millions of driven miles to encounter
- Improve productivity with automatic generation of realistic test and training scenarios
Capabilities
Automated Scenario Generation
- Automatically generate diverse, valid, and critical test scenarios with randomized attributes such as locations, lanes, and actor behaviors
- Replay simulation and real-world logs and generate behavior and physical variations to train and test scenarios optimized for your specifications
- Reduced manual effort with a constrained-random test generator that ensures parameter consistency with system constraints and physics
- Generate scenarios on new maps and ODDs with no change to abstract scenario definition
Closed-loop Simulation
- Use reactive actors to dynamically adapt scenario execution at runtime to ensure test conditions meet the intended scenario intent
- Reduce manual inspection efforts and optimize compute resources by adjusting rich behavior models in response to AV stack actions
- High-fidelity synthetic sensor simulation, grounded in physics and actors’ behaviors leveraging NVIDIA Omniverse and Cosmos Transfer
Abstract Scenario Libraries
Over 200 abstract scenarios for urban driving, highways and more, generating millions of concrete scenarios to accelerate AV development while maximizing productivity
Open Platform
Seamless integration with industry-leading simulators and proprietary test platforms to preserve existing investments
OpenSCENARIO DSL Support
- Define abstract scenarios in a modular and reusable format using an industry-standard language
- Enable seamless scenario reuse across maps, ODDs, and test environments while ensuring formal consistency in scenario parameters, constraints, and metrics
