README

Code Organization

• .devcontainer: Docker related files.
• src: All the source code.
  • dogmeetsworld.py: The main script for running an end-to-end motion planning test in an environment.
  • obstacles.py: Contains classes and functions for generating an obstacle environment.
  • path_vis.py: Contains the PathVisualizer class PathVisualizer used to visualize a global, multi-segment path of the robot.
  • search/: Global path planning code for the quadruped trunk.
  • example-*: Example code of specific components, mainly for sub-system testing purposes.
  • quadruped_drake/: The main reference code used, which includes TOWR. The only modifications made to this reference code base are listed below.
    • planners/towr_of_powr.py: Mid-level trunk planner used to loop through the segments of the global path and start/end states to TOWR. This extends the originally existing functionality of the TowrTrunkPlanner class.
    • towr/include/towr/terrain/Grid.h: Contains the Grid class used to calculate terrain height and derivative values.

Running our project

Startup Docker Container

Be sure you have docker installed locally on your computer. Then you can start it as shown below, or by using the devcontainer functionality of VSCode.

First build the docker image:

cd .devcontainer
docker compose build sim

Startup a docker container of the image:

cd .devcontainer

# start up the container
docker compose up --detach

# open a shell in the container
docker exec -it rbe-550-sim-1 bash

# navigate to the directory mapped to the project directory
cd ws

# do things in the container...

# stop the container once done
docker compose stop

All other instructions are assumed to be in the docker container with, unless otherwise stated.

Build TOWR

Move into the quadruped_drake directory from the ws directory.

cd src/quadruped_drake/

Setup the build profile for conan (C++ package manager):

conan profile detect --force

Install dependencies through Conan (this creates a build directory):

conan install . --output-folder=build --build=missing

Configure the build:

cd build
cmake .. -DCMAKE_TOOLCHAIN_FILE="conan_toolchain.cmake" -DCMAKE_BUILD_TYPE=Release

Compile C++ code (includes TOWR and custom LCM bindings for interface with drake):

make

Run the main Python Script

In any webbrowser open the address localhost:7000. Nothing will appear yet, but it will be used for environment visualization so be ready to refresh it.

From the ws directory, move into the src directory and run the dogmeetsworld.py python script:

cd src
python3 dogmeetsworld.py

The script will initially show output related to TOWR performing optimization computations. This will happen for several seconds until it stops with the following text, at which point the webbrowser page should be refreshed.

Full multi-trajectory plan complete!
INFO:drake:Meshcat listening for connections at http://localhost:7000

The rough terrain environment should appear with a visualization of the planned, global trunk path, as well as the robot attempting to follow it. The robot will initially collapse due to torque limits, followed by a visualization of the planned trunk as a transparent black box, and foot placements as black balls.

Other python scripts in this directory can be run in a similar way.