Key components
Neural reconstruction
Reconstruct maps, worlds, and scenarios in simulation automatically from collected drive logs. Transform a manual artist process into an automated workflow with neural reconstruction. Reduce complex scenario creation timelines from weeks to hours.
Novel view synthesis
View recorded real-world data from entirely new perspectives. Enable log re-simulation with pose divergence (novel agent positioning). Transfer collected logs to entirely new sensor suites and vehicles (e.g., log data collected on a sedan vehicle can be used by a truck with different sensors).
Intelligent behaviors
Utilize intelligent behavior models that leverage drive data to learn and emulate real-world driving patterns. Built-in driving models simulate physics-based dynamics for cars, trucks, tractor-trailers, and more.
Train, test, and validate with AI
Train end-to-end systems with a performant, scalable, closed-loop simulator suitable for high-throughput reinforcement learning. Test and validate using neural simulations that can be linked to requirements, ODDs, and safety case trees.
.webp)
.webp)
