Pupper Robot: Control & Simulation

This repo has been replaced by: https://github.com/stanfordroboticsclub/stanfordquadruped

Overview

This repository contains Python code to run Pupper, a Raspberry Pi-based quadruped robot. In addition to the robot code, this repository also contains a wrapper to simulate the robot in MuJoCo or PyBullet using the same code that runs on the robot.

Video of the robot following a QR code: https://www.youtube.com/watch?v=iyuJq_Pn7TM

Installing and Running Code on the Raspberry Pi

Materials

• Raspberry Pi 4
• SD Card (32GB recommended)
• Raspberry Pi 4 power supply (USB-C, 5V, >=3A)
• Ethernet cable

Steps

• Install Raspbian Buster Lite onto the Pi
  • Download https://www.raspberrypi.org/downloads/raspbian/
  • Use BalenaEtcher to flash the OS to the SD card
• Set up the Raspberry Pi
  • Before even ejecting the SD Card, follow the instructions on this repo to put the self-installing setup script on the Pi: https://github.com/stanfordroboticsclub/RPI-Setup
  • Complete the “Actually Doing It”, “Getting Internet Access”, and “Getting Started With the Pi” sections
• Test that the Pi works and connects to the internet
• (Optional) Install the PREEMPT-RT kernel onto the Pi
  • Download the kernel patch https://github.com/lemariva/RT-Tools-RPi/tree/master/preempt-rt/kernel_4_19_59-rt23-v7l%2B
  • Follow these instructions starting from “Transfer the Kernel” https://lemariva.com/blog/2019/09/raspberry-pi-4b-preempt-rt-kernel-419y-performance-test
  • Test by running in the shell:

    uname -r

  • We haven't yet characterized the benefits of using the preempt-rt kernel on the Pi so skipping this step might still give fine performance.
• Install pigpio
  • Run

    sudo apt-get install python3-distutils

  • Follow the instructions available at http://abyz.me.uk/rpi/pigpio/download.html to install a GPIO library for the Pi that adds software PWM capability.
• Get the Pupper code
  • Clone the Pupper repository https://github.com/Nate711/PupperPythonSim/
  • Install requirements:

    sudo bash install_packages_robot.sh

• Get the Pupper controller code
  • Clone the controller repo: https://github.com/stanfordroboticsclub/PupperCommand
  • Follow the instructions in the README

    Running the Robot

• SSH into the robot

  ssh pi@10.0.0.xx

  where xx is the local address you chose for the Pi

• Once connected to the Pi, go into read-write mode

  rw

• Now go into this repo's PupperPythonSim directory and run the robot code!

  cd PupperPythonSim
  sudo pigpiod
  python3 run_robot.py

  If you already have the pigpio daemon running, you can ignore when sudo pigpiod says it can't initialize pigpiod.

• You can interrupt and stop the program by pressing Control-C.
• To turn off the servo motors, run

  sudo pkill pigpiod

Installation for PyBullet Simulation

The PyBullet simulator is free for academic use and requires no license whatsoever, but in my experience PyBullet is much slower than MuJoCo and is less clear about how to tune the contact parameters.

0. Clone this repository

   git clone https://github.com/Nate711/PupperSimulation.git

1. Install pybullet (in a python 3.7 environment)

   pip install pybullet

   Run PyBullet Simulation

2. Run simulation:

   python3 simulate_pybullet.py

Installation for MuJoCo Simulation

MuJoCo has been faster than PyBullet in my experience, but requires that you request and activate a license. The process to acquire a license can take several days.

0. Clone this repository

   git clone https://github.com/Nate711/PupperSimulation.git

1. Acquire a license for MuJoCo at http://mujoco.org/. You can get a free trial of the professional license for a month, or with a student account, a free year.

2. Follow the instructions at https://github.com/openai/mujoco-py to correctly install MuJoCo.

If you have trouble installing MuJoCo on macOS because gcc can't find certain header files, like "limits.h" or "stdio.h", then try completing the installation wizard that pops up when you run:

sudo open /Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg 

And also do:

brew update
brew install gcc@8
brew link --overwrite gcc

3. Install the python requirements:

   sudo bash install_packages_sim.sh

Run MuJoCo Simulation

(Sorry, the MuJoCo sim has not been updated in a while so does not currently work. You're welcome to update it to work with our other code though!)

1. Run

   python3 simulate.py

2. The MuJoCo simulator should then pop up in a new window with various interactive options. Press space to stop or start the simulation.