Onshape to robot examples

Here you will find several examples of robots that were design to be exported properly using the onshape-to-robot tool.

Have a look at the design-time considerations to understand better the constraints to export your robot.

Before testing, in a nutshell, you need onshape-to-robot:

pip install onshape-to-robot

This repository contains both config.json file and resulting export. You can use the following commands to visualize the result:

# Start PyBullet simulation
onshape-to-robot-bullet directory
# Start MuJoCo simulation
onshape-to-robot-mujoco directory

Robots

2 wheels

Description: Very simple design and minimalistic configuration

• Give it a try
  • onshape-to-robot-bullet 2wheels_urdf
  • onshape-to-robot-mujoco 2wheels_mujoco
• Onshape design
• config.json
  • urdf
  • mujoco
• Notes:
  • The joint properties are set to use continuous with URDF format, and velocity controlled in MuJoCo
  • wheel2 has the _inv suffix, which changes its direction making both wheels spin in same direction when same sign of order is given**

Adjustable height arm

Description: Robot with 4 degrees of freedom, one prismatic (linear) joint and 3 rotations

• Give it a try
  • onshape-to-robot-bullet -f adjustable_height_arm_urdf (note: -f is passed to fix the base to the ground)
  • onshape-to-robot-mujoco adjustable_height_arm_mujoco
• Onshape design
• config.json
  • urdf
  • mujoco
• Notes:
  • The robot is fixed using a "fixed" Onshape feature
  • There is one frame (frame_tip) attached to the tip of the arm
  • Collisions mesh are ignored using no_collision_meshes

Robot-soccer-kit omnidirectional

Description: an omnidirectional robot from the robot-soccer-kit project.

• Give it a try
  • onshape-to-robot-bullet rsk_urdf
  • onshape-to-robot-mujoco rsk_mujoco
• Onshape design
• config.json
  • urdf
  • mujoco
• Notes:
  • In MuJoCo, passive wheels are marked as unactuated.
  • In URDF, fixed links are added to the base, mostly to preserve colors in pyBullet.
  • Each small wheel is a degree of freedom. This approach is tedious and likely sub-optimal to simulate such robot.

Quadruped

Description: 12 DOF quadruped robot, design similar to the Metabot open-source project using XL-320 servos.

• Give it a try:
  • onshape-to-robot-bullet quadruped_urdf
  • onshape-to-robot-mujoco quadruped_mujoco
• Onshape design
• config.json
  • urdf
  • mujoco
• Notes:
  • Full pure shape approximation (OpenSCAD) for collisions
  • Frames in the trunk (trunk_frame) and in the tip of one leg (tip_frame)
  • A frame is attached to the tip of a leg (leg_tip)

Dog

Description: 12 DOF dog robot, made for fun with MX-64 servos (was actually built)

• Give it a try:
  • onshape-to-robot-bullet dog
  • onshape-to-robot-mujoco dog
• Onshape design
• config.json
• No pure shape approximation (pure STLs)
• Dynamics is overridden for MX-64 and MX-106, since it is provided by constructor

Sigmaban2019

Description: 20 DOF humanoid robot, snapshot of 2019 Sigmaban model from team Rhoban, used at RoboCup kid size

• Give it a try:
  • onshape-to-robot-bullet sigmaban2019_urdf
  • onshape-to-robot-mujoco sigmaban2019_mujoco
• Onshape design
• config.json
  • urdf
  • mujoco
• Notes:
  • Full pure shape approximation (OpenSCAD) for collisions
  • Frames in the trunk (trunk_frame), in feet (right_foot_frame, left_foot_frame), the camera (camera_frame), the head base (head_base_frame) and some location to use Vive tracker for ground-truth (vive_frame`).
  • In the URDF format, STLs are merged and simplified

Orbita SPM

Description: Orbita is a type of spherical parallel manipulator with 3 DOFs.

• Give it a try:
  • onshape-to-robot-mujoco field_mujoco
• Onshape design
• config.json
  • mujoco
• Notes:
  • The based is fixed using a fixed constraint
  • Two loop closures are used here in order to ensure the mechanism is fully constrained
  • Collisions are disabled in this example

Field

Description: This is not actually a robot, but an environment that is designed in Onshape and can be converted into an URDF file

• Give it a try:
  • onshape-to-robot-bullet field_urdf
  • onshape-to-robot-mujoco field_mujoco
• Onshape design
• config.json
  • urdf
  • mujoco
• Notes:
  • Since it is a static environment, no_dynamics is passed
  • There are multiple root nodes in this robot. Thus, add_dummy_base_link option is added in URDF to ensure there is only one root node.
  • This is a model of RoboCup humanoid soccer kid-size (2019) 6 x 9 m field
  • In URDF, the use_fixed_links option is set to true so that the colors of different sub-part are kept

Re-running the imports yourself

Instructions

Read the instructions from the onshape-to-robot repository, especially be sure to have the API key registered and set as environment variable before running onshape-to-robot.

To get the same results, you might need openscad and meshlab installed.

sudo apt-get install openscad meshlab

Else, pure shape approximation and meshes simplification might be disabled.

Why do I get ERROR (403) while using OnShape API ?

If you want to export the robots yourself, you need to have the rights to access to the assembly. Thus, if you want to re-run the import for the examples in this repository, you will have to make your own copy of the document. Fortunately, this is quite simple because it is just about clicking on that button on top left:

Then, update the url parameter in config.json of the robot to match the new document created where you own the robot design.

Why the colors are looking weird in pyBullet ?

Actually, pyBullet does only keep one color per link currently in the viewer. This is a known limitation that does not affect your physical simulations. (https://github.com/bulletphysics/bullet3/issues/2650)

For instance, the quadruped robot above will look like this: