Parallelizing Motion Planning Using Lambda Serverless Computing

Details

• ICRA Paper: https://drive.google.com/file/d/1VHzHlAMv2MfTbqtcT6fy6qaqXoQTjVyG/view
• Video explaining the algorithm and results: https://www.youtube.com/watch?v=yqvL2lCwBCo

Usage instructions

1. This repository heavily borrows from previous work. Please follow setup and installation instructions from there for packages and libraries.
2. Copy the scripts/run_experiments.py file into your build folder and run it. python run_experiments.py --help should give configuration options.

Repository Structure

1. include/ contains most of the code and logic for this project in header files since the bulk of it is written using C++ templates.
2. Communication is handled by comm.hpp, packet.hpp, buffer.hpp, connection.hpp. These establish a connection to a centralized coordinator that invokes the lambda functions (or for the local case, forks processes).
3. include/demo contains the various testing scenarios. Scenarios have to implement a few methods like randomSample, isValid(state), isValid(start, goal) in order to be correctly used by the planner.
4. coordinator.hpp contains most of the code for the coordinator, and demo/local_lambda_fixed_graph.hpp contains most of the code for the lambdas.