GPU-accelerated physics simulation
Uses NVIDIA PhysX to provide high-fidelity, accurate physics simulations including support for deformable objects, enabling realistic modeling of robot interactions.
Modular architecture
Allows customization of workflows with robot training environments, tasks, learning techniques, and integration of custom libraries such as skrl, RLLib, and rl_games.
Comprehensive sensor simulation
Provides RTX-based cameras, LIDAR, and contact sensors for accurate sensor simulation within the training environment.
Flexible deployment options
Can run locally or be distributed across cloud infrastructure to support large-scale robot learning deployments.
Multiple learning approaches
Supports reinforcement learning, imitation learning, and motion planning workflows to accommodate different robot training methodologies.
Reduced sim-to-real gap
Employs GPU-accelerated PhysX to provide accurate physics simulations that minimize discrepancies between simulation and real-world robot performance.