Simulation & Software Development
Engineering embodied intelligence requires far more than models alone. It needs a complete stack around simulation platforms, simulation assets, world construction, software middleware, debugging tools, and deployment pathways. This chapter organizes those pieces into a single engineering loop: where robots run, how worlds are built, how physics is configured, and how software is integrated end to end.
Contents:
- Simulation Platforms — Positioning, selection, and reading entry points for mainstream simulators
- Simulation Assets — How robots, objects, scenes, sensors, materials, and lighting are modeled, produced, imported, and validated
- Simulation World Building & Physics Rules — World hierarchies, frames, contact/friction/constraints, integration stability, randomization, and validation
- Simulation Tool Comparison — Comprehensive simulator comparison
- NVIDIA Ecosystem — Omniverse, Isaac Lab, Cosmos, GR00T
- ROS2 Ecosystem — ros2_control, MoveIt 2, Nav2 deep dive
- Open-Source Frameworks — LeRobot, robomimic, robosuite, ManiSkill
- Development Toolchain — URDF/MJCF, BT.CPP, Foxglove
Recommended Reading Order
If you are new to this section, the most practical sequence is:
- Start with Simulation Platforms to map the simulator landscape.
- Then read Simulation Assets to understand how robots, objects, scenes, and sensors become usable assets.
- Next read Simulation World Building & Physics Rules to understand how those assets are assembled into trainable, testable, transferable worlds.
- Follow with Simulation Tool Comparison and NVIDIA Ecosystem to complete platform and ecosystem selection.
- Finally connect the stack to implementation through ROS2 Ecosystem, Open-Source Frameworks, and Development Toolchain.