Key Features

What you can do

Physically Based Simulation

Uses NVIDIA PhysX to simulate joint friction, actuation, rigid and soft body dynamics, and velocity for realistic robot behavior.

Synthetic Data Generation

Includes Isaac Replicator for generating synthetic datasets with domain randomization to train robot perception models.

Sensor Simulation

Simulates sensors such as RGB-D cameras, RTX-Lidar, Radar, and IMU with support for OpenCV lens distortion models.

ROS 2 and Python Integration

Supports ROS 2 (including Jazzy) and MoveIt2 workflows, enabling integration with common robotics software stacks.

Modular Architecture

Allows creation of custom USD-based simulators and integration into existing pipelines for various robot types including humanoids and AMRs.

Containerized Deployment

Provides containerized versions for Linux with headless mode and livestream client support for flexible deployment.