Tensegrity

Tensegrity-related code (using ROS Noetic)

Dependencies:

GTSAM

git clone https://github.com/borglab/gtsam
cd gtsam
git checkout 4.2.0 # Don't use a newer one
mkdir build
# use ccmake .. for more options. installing in local dir to avoid permissions issues
cd build
make -j8 -DCMAKE_PREFIX_INSTALL=../install -DGTSAM_USE_SYSTEM_EIGEN=ON .. 
make install
export GTSAM_DIR=$(pwd)/install/lib/cmake/GTSAM # <- add to .bashrc

TORCH Download libtorch according to OS / Cuda: https://pytorch.org

cd <Path/to/Torch/zip>
unzip <torch.zip>
export Torch_DIR=$(pwd)/libtorch # <-- Add it to bashrc!

Other dependencies:

• pyrealsense2 - For using a realsense camera for RGB-D observations
  • Need to install using pip (NOT using conda -- python makes no sense)

Compile

Clone this repository inside src:

cd <HOME>/catkin_ws/src/
git clone https://github.com/PRX-Kinodynamic/tensegrity.git
cd ..
catkin_make
export TENSEGRITY_ROS=$(pwd)/ # <-- Add it to bashrc!

Prerequisites

Install torch:

cd PATH/TO/LIBTORCH/INSTALL # cd into where to install libtorch
wget https://download.pytorch.org/libtorch/cu128/libtorch-VERSION.zip # Go to https://pytorch.org/ to get the latest libtorch version (c++)
unzip libtorch-*
echo "export Torch_DIR=/PATH/TO/LIBTORCH/INSTALL/libtorch/" >> ~/.bashrc # Add env variable to bashrc. Might need to source

Tests

Running tests for all packages:

catkin_make               # Compile and generate msgs
catkin_make run_tests     # Run tests 
catkin_test_results       # Check for failures

##Running State Estimation In separate tabs:

roscore
rosrun perception image_publisher.py
roslaunch interface camera_info_publisher.launch camera_info_file:=/common/home/pm708/Desktop/catkin_edgar/src/tensegrity/estimation/config/camera_info.yaml

Launch Files

• Publish realsense RGB-D images and camera info in ros topics
  • roslaunch perception camera_publisher.launch
• Manual calibration of color masks:
  • roslaunch perception tensegrity_endcap_v2.launch
  • rqt # plugins -> Configuration -> DynamicReconfigure . Configure each mask Low - High
• Run pointcloud clustering
  • roslaunch interface pointcloud_from_camera.launch
    • Assumes correct color filters are in file: perception/config/color_filters.yaml
• Run the state estimation of Tensegrity
  • roslaunch interface tensegrity_state_estimation.launch