Skip to content

Latest commit

 

History

History
96 lines (58 loc) · 4.72 KB

usermanual.md

File metadata and controls

96 lines (58 loc) · 4.72 KB
Raw

User Manual

Test examples

Directory structure

The directories test/example_files* each contain the following folders:

  • raw: this folder contains a number of photographs taken around the periphery of a steel welding target. For each image, it also contains the corresponding robot pose (in the world frame) at the time the photograph was taken, in .npy format. A human readable poses.txt file is included, where each row corresponds to a flattened augmented rotation matrix of 12 elements (9 rotation values, 3 position values). The camera.yaml file contains the camera intrinsics.

  • mesh: This folder contains the output of the 3d reconstruction module, which is the transformed mesh with respect to the robot's coordinate system (world). mesh is the standard output of the photogrammetry pipeline (omitted here), while transformed_mesh is the output of the transformation pipeline. This folder contains .mtl, .obj and .png files.

  • welding_trajectory: This folder contains the results of the weld seam detection module. simplified_mesh.obj is the result of simplifying th emesh to a maximum number of 400 triangles, and is useful for path planning, as the few number of elements reduces the computational time for arriving at a solution. welding_paths.npy contains the detected welding seams in the form of a numpy data structure, while trajectories_mesh.obj contains the 3-D representation of the latter.

A total of three examples have been compiled. A detailed description of these can be found here.

Cloning the repository

git clone --recursive [email protected]:ikh-innovation/roboweldar-rose-ap.git

Since the submodules in this repository were added using SSH, you will need to have SSH access enabled on your Github account to fully clone this repository.

Running the component instance

Setting up your dependencies

First, make sure you have installed the docker Nvidia driver:

cd docker
sudo chmod a+x docker_cuda_install_script.sh
./docker_cuda_install_script.sh

Then, we set up the services by starting our docker containers. Run

sudo chmod a+x entrypoint.sh
./entrypoint.sh

This will create folders in the docker dir, which allow you to peek into each container's state. This is helpful because you may want to manually inspect files, or copy them over to your host manually. Now that the services are up and running, we can move on to interacting with them.

From the repository root, run

cd src
python3.7 -m virtualenv .venv
source .venv/bin/activate
pip install -r roboweldar_networking/interfaces/requirements.txt

Now you should have the required packages installed in the Python venv on your host, so that you can proceed with running the example.

Running the 3-D reconstruction module

In the same terminal, run

python run_3d_reconstruction.py

This script uses the roboweldar-coordinator REST API to transmit robot pose and photo data residing in the raw data directory to the roboweldar-3d-reconstruction service, which will fetch these files from the roboweldar-coordinator filesystem and start computations on them to produce a 3D reconstructed model. The result will be placed in the newly created (created on runtime) docker/roboweldar-networking/mesh/ directory.

To fetch it via the API, use the following function in the roboweldar-networking module:

get_mesh_files(host, httpPort, path_to_dir, mesh_files)

Alternatively, you can manually navigate to the docker/roboweldar-networking/mesh/ and inspect the files.

Running the weld seam detection module

In the same terminal, run

python run_weld_seam_detection.py

This script uses the roboweldar-coordinator REST API to transmit a message to the roboweldar-weld-seam-detection service, which will fetch the previously reconstructed mesh file from the roboweldar-coordinator filesystem and start computations to produce welding trajectories(as npy and mesh datatypes). The result will be placed in the newly created (created on runtime) docker/roboweldar-networking/welding_trajectory/ directory.

To fetch it via the API, use the following functions in the roboweldar-networking module:

get_mesh_files(host, httpPort, path_to_dir, mesh_files)
get_trajectory_file(host, httpPort, path_to_dir, trajectory_file_name)

Alternatively, you can manually navigate to the docker/roboweldar-networking/welding_trajectory/ and inspect the files.