kaboom

kaboom is a Python library for the KAI Agent-Based Organizational Optimization Model (KABOOM) described in a paper currently being submitted for publication. This agent-based model simulates the performance of teams of engineers solving a multivariate design problem, and incorporates the cognitive styles and solution-sharing interactions of agents. Cognitive styles of agents in KABOOM are based on the Kirton Adaption-Innovation Inventory which describes a range of problem-solving styles from adaptive (incremental change) to innovative (radical change).

Installation

Use the package manager pip to install kaboom from Github.

KABOOM has been tested with python 3.6.7. We recommend creating a new Anaconda environment before installing KABOOM. The code below assumes you have Anaconda.

conda create -n kaboomEnvironment python=3.6
source activate kaboomEnvironment
conda install pip
pip install git+http://github.com/THREDgroup/kaboom.git 

If you still have package issues, tell pip to install the specific versions of packages listed in kaboom/requirements.txt. In the example below, you'll need to replace "/path/to" in the command below with the path to requirements.txt on your computer.

pip install -r /path/to/kaboom/requirements.txt #

Running IDETC Studies

Run in your favorite python compiler:

import kaboom

# runs the first study, which analyzes the effect of a team's
#cognitive style composition on performance for 2 problems
kaboom.IDETC_studies.i_teamStyle.run()
# output figure is saved to kaboom/kaboom/IDETC_studies/results/ directory

#run the other experiments from IDETC paper:
kaboom.IDETC_studies.ii_subteamStyle.run()
kaboom.IDETC_studies.iii_strategicTeams.run()
kaboom.IDETC_studies.iv_problemDecomposition.run()
#results figures are saved to kaboom/kaboom/IDETC_studies/results/

Usage

Run a basic simulation

import kaboom

#create a parameters object:
parameters = kaboom.params.Params()

#modify simulation parameter values as desired
parameters.nDims = 14

#run the simulation with an abstract sinusoid objective function
team = kaboom.kaboom.teamWorkSharing(parameters)

#check team performance
#invert score *-1 so that higher score = better performance
print( team.getBestScore()*-1 )

#check other outcomes, such as the number of pairwise interactions:
print ( team.nMeetings )

Customized KAI score composition

Specify a set of KAI scores for agents on a team

import kaboom

#create a parameters object:
parameters = kaboom.params.Params()

#provide a list of KAI scores for each agent on the team
parameters.nAgents = 5
parameters.kaiList = [100,101,102,120,80]

#run the simulation using this custom team composition
team = kaboom.kaboom.teamWorkSharing(parameters)

#check that the KAI scores of each team member match the desired values:
print("KAI scores of each team member:")
print([a.kai.KAI for a in team.agents])

Running a Car Design problem

from kaboom.carMakers import runCarDesignProblem

#create a parameters object
#parameters (p.nAgents = 33, p.nTeams = 11, p.nDims = 56) are automatically set for this problem.
parameters = kaboom.params.Params()

#run the simulation with the car designer objective
team = runCarDesignProblem(parameters)

#check the performance of the team
#invert score *-1 so that higher score = better performance
print( team.getBestScore()*-1 )

Running a Beam Design problem

from kaboom.runBeamDesigners import runBeamDesignProblem
team = runBeamDesignProblem() #run beam design simulation
print( team.getBestScore()*-1 ) #performance: higher = better

Running Simulations in Parallel

#run simulations in parallel with multiprocessing
import multiprocessing
import itertools
from kaboom.kaboom import teamWorkProcess
from kaboom.params import Params

p = Params()
p.steps = 10
p.reps  = 4

allTeamObjects = []
if __name__ == '__main__':
      myPool = multiprocessing.Pool(processes = 4)
      allTeams = myPool.starmap(teamWorkProcess, zip(range(p.reps), itertools.repeat(p) ) )
      for team in allTeams:
          allTeamObjects.append(team)
      myPool.close()
      myPool.join()

print("team scores: ")
print([t.getBestScore() for t in allTeamObjects])


#now create your own experiment, varying a parameter and checking the performance

License

MIT