Reacher Sim

Simulation and Reinforcement Learning for Reacher Robot

System setup

Operating system requirements

• Mac
• Linux
• Windows

Mac-only setup

Install xcode command line tools.

xcode-select --install

If you already have the tools installed you'll get an error saying so, which you can ignore.

Windows-only setup

The PyBullet simulator requires Microsoft Visual C++ to compile on Windows. You can find the download link for the build tools here. Once you have that installed, run the program and select the option for "Desktop development with C++". Leave all the "optional" downloads checked and download the packages. It will be quite a large download.

Conda setup

Now install miniconda through this link. Once that's done, create a conda environment for the project by running the terminal commands below. If you're on Windows, you will need to do this in the Anaconda Prompt Terminal. Those on Linux and MacOS can run the commands in a regular terminal.

conda create --name reacher python=3.8
conda activate reacher
pip install ray arspb

Getting the code ready

If you followed along with the instructions in Hands-On 3, you should have already cloned this repo to your computer. Navigation to the reacher-lab folder and run the following commands to install the dependencies for the simulator.

cd RBT350-HandsOn3  
pip install -e .

Running simulation on computer

python reacher/reacher_manual_control.py

You should see the PyBullet GUI pop up and see Reacher following the joint positions set by the sliders.

Running with inverse kinematics

python3 reacher/reacher_manual_control.py --ik

Assuming you have implemented all the functions inside of reacher_kinematics.py according to their documentation, you can run the above command to enable Cartesian control of the robot. Be slow with the sliders as the leg has a very limited range for where it can be. You will see that sometimes the leg starts jerking because it is unable to find a suitable solution for the given XYZ coordinate.

Deploying to robot

Upload reacher lab firmware to Teensy

1. Open vscode and upload a stable version of your Hands-On 2 code. This is just to get the Teensy application to open in the background.

2. Click the "open hex file" button in the application window (left side) and choose the firmware.hex THAT'S IN THIS FOLDER (not from your Hands-On 2 code).

3. Click the green auto button if it's not already highlighted

4. Press the button on the Teensy to program it with the hex file

Run simulator and robot simultaneously

Run the python code:

python3 reacher/reacher_manual_control.py --run_on_robot

to do joint control of one leg and

python3 reacher/reacher_manual_control.py --run_on_robot --ik

to do Cartesian control of the leg.