collision-detection

Acknowledgements

This work has been supported by the EU H2020 project CLASS, contract #780622.

Project structure

collision-detection
|-- cd
|   |-- fileBasedObjectManager.py
|   |-- dataclayObjectManager.py
|   |-- __init__.py
|   `-- CD.py
|-- tp
|   |-- fileBasedObjectManager.py
|   |-- dataclayObjectManager.py
|   |-- __init__.py
|   `-- v3TP.py
|-- lithops_runner.py
|-- README.md
|-- test-file.py
|-- test-dataclay.py
|-- data
|   `-- data.txt

Testing

Update dataclay cfgfiles with your credentials

Copy dataclay stubs and cfgfiles to the root directory e.g.

ssh <user>@<dataclay host> "cd <DATACLAY_CLOUD_DIR>/;./GetStubs.sh
scp -r <user>@<dataclay host> "cd <DATACLAY_CLOUD_DIR>/stubs .

Update lithops config with optimal setup

Run collision detection with defaults:

python lithops_runner.py --operation cd

Or run trajectory prediction with defaults:

python lithops_runner.py --operation tp

The data sent to workers in chunks when default chunk size is 1. Use --chunk_size parameter to control the level of concurrency, e.g.

python lithops_runner.py --operation cd --chunk_size 3

Extend runtime docker image with map function module and dependencies using lithops runtime extend

lithops runtime extend $RUNTIME_NAME --filepath /home/class/collision-detection/centr_cd.py --function detect_collision_centralized

In order to avoid expensive serialization/deserialization on driver and workers layers --dickle, "dockerized pickle", parameter can be applied along with previously extended runtime docker image

python lithops_runner.py --operation cd --chunk_size 3 --dickle --runtime <EXTENDED_RUNTIME_IMAGE>

In case .lithops_config has rabbitmq section configured, to avoid interaction with object storage in driver and worker layers

python lithops_runner.py --operation cd --chunk_size 3 --dickle --runtime <EXTENDED_RUNTIME_IMAGE> --storageless