Build a Robot Pick-and-Place Simulation with SimReady Assets
Overview
SimReady assets provide pre-configured OpenUSD files containing the necessary physical properties—such as mass, collision meshes, and joint drives—required for robotics simulation. By utilizing these assets within NVIDIA Isaac Sim, developers can accelerate the creation of complex tasks like pick-and-place operations. This workflow demonstrates how to assemble a UR17G robot arm and a Robotiq 2F-140 gripper to manipulate a beverage can, using an AI agent to generate the necessary control code.
Understanding SimReady Assets
SimReady Foundation defines the technical specifications required for an asset to function reliably in a simulation environment. These specifications are organized into three tiers:
- Profile: Defines the use case (e.g., a robot body).
- Features: Specifies required capabilities, such as controllable joints.
- Requirements: Sets validation checks for specific properties, such as joint drive settings.
Assets on SimReady Central include validation reports that confirm they meet these requirements, providing a standardized baseline for simulation developers.
Building the Simulation
To build a pick-and-place simulation, developers must combine validated assets with task-specific code. The process involves three primary steps:
1. Asset Acquisition
Download the required robot arm, gripper, and object models from SimReady Central. It is critical to keep the complete asset folders intact to ensure Isaac Sim can locate all supporting files and metadata.
2. Scene Assembly
Assemble the robot and environment within Isaac Sim. This can be performed manually or via an AI agent using Isaac Sim's Python server. The agent utilizes specific skills—such as isaac-sim-remote and usd-articulation—to link the arm and gripper into a single, controllable assembly. Verification of joint limits and drive settings is essential before proceeding to task execution.
3. Task Execution and Testing
Once the scene is composed, an AI agent can be prompted to generate and run the pick-and-place control logic. The task sequence typically includes approaching the target, performing a contact-based grasp, lifting, transferring, and releasing the object.
Simulation Considerations
When evaluating the performance of an agent-generated simulation, developers should account for specific constraints:
- Physics and Contact: Ensure the object rests correctly on the surface and that the gripper maintains contact during movement.
- Scripted Logic: The example uses a fixed drive-effort cap and predefined waypoints. It does not account for real-world variables like object deformation, liquid motion, or sensor-based path replanning.
- Validation: Always verify the accuracy of the pick-and-place sequence against success criteria, such as the ability to hold the object throughout the entire motion cycle.
For further development, users can contribute their own assets to SimReady Central by following the provided submission guidelines or learn to create SimReady assets from CAD files using frontier AI models.