DaymudeLab / EvoSOPS

A genetic algorithm for discovering diverse and high performing distributed algorithms for collective behavior in self-organizing particle systems

SOPS-SwarmGA

This is a research project in Swarm robotics using Genetic Algorithms as the backbone, to search for local controllers to produce global collective behaviors. The base model used is the SOPS model developed @SOPS Lab, ASU.
This project is conducted under supervision of Prof. Joshua Daymude*

Basic RUST setup

Make sure you have rust installed* To install latest rust compiler run command:
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
Then to compile and run the simulation run command:
cargo run

Basic Repository Structure

The main.rs file houses the execution loop to run a full scale Genetic Algorithm or a Standalone evaluation of a genome. The RUST modules are organized based on the 2 main components of the code-base the SOPS environment and Genetic Algorithm housed respectively in the SOPSCore and GACore folder.
The modules have a main module file which contain all the code shared within the respective module files. Separate files are maintained for the Separation behavior which is implemented as a extension of the Aggregation behavior.
Please use a similar file format for making any extensions of new behaviors*
The utils module contains all the complementary helper functions required for main experiment tasks. There is another folder misc_scripts folder which contains python scripts used to generate all the auxilary data and graphs for/from the experiments.

Running the experiments

Pass options to run the experiments
cargo run --package swarm_aggregation_ga --bin swarm_aggregation_ga -- --behavior <BEHAVIOR> --exp <EXPERIMENT_TYPE> --ks <PARTICLE_SIZES> ...(other options)

Printing out the options: cargo run --package swarm_aggregation_ga --bin swarm_aggregation_ga -- --help

Options:

-b, --behavior <BEHAVIOR>
          Behaviour to run

          Possible values:
          - agg: Aggregation
          - sep: Separation

  -e, --exp <EXPERIMENT_TYPE>
          Type of Experiment to run

          Possible values:
          - ga: Full scale genetic algorithm run
          - gm: Stand-alone single genome run
          - th: Stand-alone single theory given solution run

  -g, --gen <MAX_GENERATIONS>
          Maximum no. of generations to run the Genetic Algorithm Experiment
          
          [default: 0]

  -p, --pop <POPULATION>
          No. of genomes in the population for a Genetic Algorithm Experiment
          
          [default: 0]

      --gran <GRANULARITY>
          Genome representation granularity
          
          [default: 20]

  -m, --mut <MUTATION_RATE>
          Mutation rate per gene
          
          [default: 0.08]

      --eli <ELITIST_COUNT>
          Maximum no. of elite genomes to preserve in a generation
          
          [default: 0]

  -k, --ks <PARTICLE_SIZES>
          Particle Sizes to run on (eg. use multiple -k<String"(<u64>,<u64>)"> arguments to specify multiple sizes)

  -s, --seed <SEEDS>
          Seed values to run experiments with, for reproducable trials (eg. use multiple -s<u64> arguments to specify seeded trials)

      --path <FILE>
          File to read genome value from

  -v, --verbose
          Specify if execution description is written to the output

      --snaps
          Specify if snapshots of the experiment run are written to the output (only valid for stand-alone experiments)

  -h, --help
          Print help (see a summary with '-h')

  -V, --version
          Print version