This repository contains the COMPSs Obstacle Detection application file and the configuration files required to execute it.
The following software components are required in order to fully execute the application workflow described in this repository.
Download and install object detection (tkDNN) software, installation and execution instructions are available at the same repository. This component must be running at the edge resource running the COMPSs Obstacle Detection application when executing it.
Download and install COMPSs, installation and execution instructions are available at the repository in the link.
Download and install dataClay, installation and execution instructions are available at the repository. Refer to the deployment stage, as the build stage can be skipped as the dockers are already present at Dockerhub. This component must be running at the Cloud resource when executing the COMPSs Obstacle Detection application.
Download and install PyWren at the Cloud resource, installation and execution instructions are available at the repository from the link.
Download and install trajectory prediction at the Cloud resource, installation and execution instructions are available at the repository from the link.
Download and install collision detection at the Cloud resource, installation and execution instructions are available at the repository from the link. This component is not required in order to execute the COMPSs Obstacle Detection application.
The application has been developed and tested with Python 3.6.8. The following python packages must be installed as well:
python3 -m pip install dataclay pymap3d zmq struct requests pygeohash geolib
The tracker_CLASS submodule is pulled by:
git submodule init
git submodule update
To compile the tracker_CLASS component, go to the ./tracker_CLASS/ folder after initializing the submodule and use the following commands:
mkdir build
cd build
cmake ../
make
The previous steps will generate an executable with a name similar to track.cpython-36m-aarch64-linux-gnu.so. Use the following command to move the tracker executable next to the COMPSs Obstacle Detection application:
mv track*.so ../../track.so
The dataClay configuration file ./cfgfiles/client.properties must be updated. The HOST=${CLOUD_IP} variable should be replaced to point to the IP address of your cloud.
Moreover, the COMPSs configuration files project.xml and resources.xml are used to define the computing resources that will be used to execute the tasks composing this application, which can be found in ./config/. These files should be updated by replacing the IP address IP1 for the actual IP address of your computing resource, the username USER1 and the PATH1 for the current absolute path to where the app.py is located inside your computing resource. For each computing resource that is going to be used, you need to add the <ComputeNode> XML tag with the required information described above.
To launch the application use the following script:
runcompss --lang=python --project=config/project.xml --resources=config/resources.xml --cpu_affinity=disabled --python_interpreter=python3 --scheduler="es.bsc.compss.scheduler.rtheuristics.RTHeuristicsScheduler" --scheduler_config_file=config/sched.config app.py
NOTE: To remind again that the application relies on the tkDNN component being up and running on the edge resource, and dataClay and Pywren components being up and running on the Cloud resource. Follow the instructions to execute the different components in their respective repositories.