runMultipleExperiments.py

''' This runfile starts the multi-agent system model
    for different mobilities and neighborhood sizes.
    The output will be two .out files. The first can 
    be used to print the average agents values for different
    mobilities and neighborhood sizes. The second can be 
    used to plot the density of agents trusting strangers.
'''

import pandas as pd
from trust.model import PDTModel
import numpy as np
import sys


N = 1000 #number of agents
n_min = 10 #minimal neighborhood size
n_max = 100 #maximum neighborhood size
n_stepsize = 10 #step size in which neighborhood size is changes
mob_rate_min = 0 #minimum mobility rate
mob_rate_max = 1 #maximum mobility rate
mob_rate_stepsize = 0.1 #step size in which mobility is changed


if (len(sys.argv) == 1):
    print("Please specify the name of Output file")
    sys.exit()

if (len(sys.argv) == 3 ):
    if sys.argv[2] =='MSAgent':
        print('MSAgent')
        model_args = {'AgentClass': 'MSAgent', 'mobility_rate': 0.2, 'number_of_agents': 1000, 'neighbourhood_size': 30}
    elif sys.argv[2] == 'WHAgent':
        print('WHAgent')
        model_args = {'AgentClass': 'WHAgent', 'mobility_rate': 0.2, 'number_of_agents': 1000, 'neighbourhood_size': 30}
    elif (sys.argv[2] == "RLAgent"):
        print("RLAgent")
        model_args = {'AgentClass': 'RLAgent', 'mobility_rate': 0.2, 'number_of_agents': 1000, 'neighbourhood_size': 30, 'learning_rate': 0.02, 'social_learning_rate': 0.5, 'discount_factor': 0.8, 'relative_reward': True}  
    elif (sys.argv[2] == "RLGossipAgent"):
        print("RLGossipAgent")
        model_args = {'AgentClass': 'RLGossipAgent', 'mobility_rate': 0.2, 'number_of_agents': 1000, 'neighbourhood_size': 30, 'learning_rate': 0.05, 'social_learning_rate': 0.5, 'discount_factor': 0.8, 'relative_reward': True, 'memory_size': 25}
    else:
        print("invalid agent type. choices are 'MSAgent', 'WHAgent', 'RLAgent' or 'GossipAgent'")
        sys.exit()
else:
    print('MSAgent')
    model_args = {'AgentClass': 'MSAgent', 'mobility_rate': 0.2, 'number_of_agents': 1000, 'neighbourhood_size': 30} 

#specify the number of epochs before and after strarting to record values
run_args = {'T_onset': 100, 'T_record': 100}

print("Model params: " + str(model_args))
print("Run params: " + str(run_args))


with open(str(sys.argv[1]) + ".out", 'w') as f:
    with open(str(sys.argv[1]) + "TrustDensity.out", 'w') as g:
        f.write(str(n_min) +" " + str(n_max) +" " + str(n_stepsize) +" " + str(mob_rate_min) +" " + str(mob_rate_max) + " " +str(mob_rate_stepsize) + "\n")
        g.write(str(n_min) +" " + str(n_max) +" " + str(n_stepsize) +" " + str(mob_rate_min) +" " + str(mob_rate_max) + " " +str(mob_rate_stepsize) + "\n")

        print("Number of agents: " + str(N))
        for n in np.arange(n_min,n_max + 0.001,n_stepsize):
            for mob_rate in np.arange(mob_rate_min,mob_rate_max + 0.0001,mob_rate_stepsize):
                print("Neighborhood size: " + str(n) + ", Mobility rate: " + str(mob_rate))

                model_args["mobility_rate"] = mob_rate
                model_args["neighbourhood_size"] = n
                model = PDTModel(**model_args)
                model.run_model(**run_args)
                df_m = model.datacollector.get_model_vars_dataframe()
                df_a = model.datacollector.get_agent_props_dataframe()

                f.write(str(df_m["Market_Size"].mean()) + " ")
                f.write(str(df_m["Trust_in_Strangers"].mean()) + " ")
                f.write(str(df_m["Signal_Reading"].mean()) + " ")
                f.write(str(df_m["Trust_Rate"].mean()) + " ")
                f.write(str(df_m["Cooperating_Agents"].mean()) + " ")
                f.write(str(df_m["Trust_in_Neighbors"].mean()) + " ")
                f.write(str(df_m["Trust_in_Newcomers"].mean()) + "\n")
                for value in df_a["Trust_in_Stranger_proportion"]:
                    g.write(str(value) + " ")
                g.write("\n")