Simulation Platform

ubt_sim is a robotics simulation platform built on NVIDIA Isaac Sim (Isaac Lab 2.2.0). It supports four robots: Walker TienKung Pro, Walker TienKung Dex, Walker S2 EDU, and Walker C1 EDU. A ROS2-ZMQ bridge publishes simulation state and images on exactly the same topics as the real robot, so teleoperation and data-collection scripts switch between simulation and hardware without modification, enabling seamless sim-to-real transfer. Main use cases:

• Simulation data collection: generate HDF5 training data in batches without occupying the real robot
• Algorithm validation: low-risk testing of imitation learning and control policies before real deployment

Architecture

Three processes (Python versions must not be mixed):

Isaac Sim (Py3.11)  ──ZMQ──►  ROS2-ZMQ Bridge (Py3.10)  ──ROS2──►  teleop / collection scripts

• Simulation /isaac-sim/python.sh: loads the task scene selected by UBT_SIM_TASK (the task name also determines the robot type) and sends camera images directly over ZMQ.
• Bridge /usr/bin/python3: bidirectional ZMQ↔ROS 2 forwarding, started automatically by start_sim.sh.
• Collection /usr/bin/python3: publishes commands and records 15 Hz HDF5.

Simulation preview

Walker TienKung Pro

Walker TienKung Dex

Walker S2 EDU

Walker C1 EDU

Documentation

Section	Description
Container setup	Shared container build, startup, and run.sh commands
Walker TienKung Pro	Parlor apple-picking scene, launching, batch collection, camera test, preview
Walker TienKung Dex	Inspire/BrainCo hands, apple picking, batch collection, playback
Walker S2 EDU	Part-sorting scene, batch collection, robot controller tools
Walker C1 EDU	Parlor pick-and-place, online IK collection, reset and playback

After collection, continue with the Imitation Learning Platform. For full playback options, see Playback.