Kinova-arm

Code created for various manipulation tasks using the 6 Degree of Freedom Kinova Gen3. Maintained by the Özay group at the University of Michigan.

This code is based upon the work done by Kwesi Rutledge in the repository https://github.com/kwesiRutledge/OzayGroupExploration.git To pull changes from this repository, you can add it as remote location with

git remote add upstream https://github.com/kwesiRutledge/OzayGroupExploration.git

Then when you want to pull changes:

git pull upstream main

If unwanted files are added when merging with upstream, then you will have to remove them. This workflow is flawed, but will work for now.

Getting Started

There are two possible methods for getting started with kinova-arm.

1. Docker
2. Local Install

The Docker method for getting started appears to be more robust (won't break as often), but it is not necessary if you are working on the lab laptop.

Docker

In this section we will discuss how to: build the Drake-Kinova Docker Image and run a container with it.

In order to control the 6 Degree of Freedom Kinova Gen3 in the Özay group, Drake is used along with the Kinova Kortex API and kinova_drake (a library built by Vince Kurtz).

If you are interested in getting set up with all of the software that you need to control our robot, do the following:

1. Pull this git repository.
2. From the repository's main directory, run a shell script to create the docker image: ./shell-scripts/create-drake-v2-docker-image.sh.
3. Start the docker container: ./shell-scripts/run-drake-v2-docker-container.sh

Local Install

1. Create the project folder (e.g., ~/kinova)

2. Create a virtual environment for the project. (e.g., python -m venv kinova-venv)

3. Into your project folder clone both kinova-arm and kinova-drake.

   1. The currently recommended version of kinova-drake: https://github.com/ozay-group/kinova-arm
   2. The currently recommended version of kinova-arm: https://github.com/kwesiRutledge/kinova_drake

4. Locally Install kinova-arm

   1. Make sure your virtual environment is activated.

   2. Enter the kinova-arm directory.

   3. Run the following commands:

      pip install --upgrade pip
      pip install -r install/requirements.txt
      pip install -e .

   4. This will install kinova-arm into the virtual environment. (This tells your computer where kinova-arm is whenever you run import kinova_arm.)

   5. Example for how to import kinova-arm is shown here (note that this file does not work without installing kinova_drake first):

      kinova-arm/scripts/debug/velocity_control1/velocity_command_test.py

      Line 34 in 0435f84

      from kinova_arm.controllers.command_sequence_controller2 import (

5. Locally Install kinova_drake

   1. Make sure your virtual environment is activated.

   2. Enter the kinova-arm directory.

   3. Run the following commands:

      pip install -r requirements.txt
      pip install -e .

   4. This will install kinova_drake into the environment. (This tells your computer where kinova_drake is whenever you run import kinova_drake.)

   5. An example of how to import this is listed above. (4e)

6. Install kortex_api

   1. Install kortex_api 2.6.0 which is compatible with state-of-art kinova_drake (instead of 3.2.0). .whl file can be found in the link below https://github.com/Kinovarobotics/kortex.git

      JFrog

   2. The 2.6.0 version of kortex_api includes dependency on protobuf==3.5.1, which is not compatible with python 3.10+. Hence, force install protobuf==3.19.4 after installing kortex_api 2.6.0

7. Install Intel RealSense SDK 2.0

   1. Follow the instruction to install at the link below:
      https://github.com/IntelRealSense/librealsense

Developing Code for Kinova

Make sure that the container named drake-container is running, use an editor like VS Code to begin developing.

In VS Code, you can attach your application to the running container, giving you access to all of the libraries installed in the container after you built it.

FAQ

pip install -r requirements.txt doesn't work and I'm installing on Mac OS X with an M-series chip and some parts fail.

We are unsure about why this happens. To complete installation without some of the vision libraries that are causing the issue, comment out the following lines in requirements.txt:

open3d          # Point Cloud
pyrealsense2    # RealSense
dt_apriltags    # Apriltags

Additional Info

• Hardware Instructions