Running the model

Single run

To run a single simulation, we first need to import the necessary python packages.

1. Import libraries

import jsonimport pandas as pdfrom src.SimulationEngine import SimulationEnginefrom preprocessing.BikeGeneration import BikeGeneration

• json : to read the configuration json file
• pandas : to read the stations and users csv
• SimulationEngine : to call and execute the simulation
• BikeGeneration: to generate the bikes initial location and metadata

2. Import data and generate bikes

The second step is to indicate the path of the configuration, stations, and users files, and import them into memory. The BikeGeneration function is used to generate bikes and their initial location. Using this function, bikes are generated randomly at stations, with a probability proportional to their capacity.

The parameters of the BikeGeneration function are: fleet size (indicated at the NUM_BIKES parameter), the system mode (0: station-based, 1: dockless, 2: autonomous, indicated at the MODE parameter) and the station file path. The returned object is a pandas.DataFrame* with the stations information and the number of bikes assigned to each station.

config_path = "data/config.json"stations_path = "data/stations.csv"users_path = "data/users.csv"
with open(config_path) as f:    config = json.load(f)
stations_data = BikeGeneration(config["NUM_BIKES"], config["MODE"], stations_path)users_data = pd.read_csv(users_path)

3. Run simulation

The last step is to simply create the SimulationEngine class and the execute the run function, indicating maximum simulated time (in seconds) to be executed. In this example, one hour (3600 seconds) of simulated time would be executed.

city = SimulationEngine(config, stations_data, users_data)city.run(until=3600)

Complete example

Here is the complete code example:

import jsonimport pandas as pdfrom src.SimulationEngine import SimulationEnginefrom preprocessing.BikeGeneration import BikeGeneration
config_path = "config.json"stations_path = "stations.csv"users_path = "users.csv"
with open(config_path) as f:    config = json.load(f)
stations_data = BikeGeneration(config["NUM_BIKES"], config["MODE"], stations_path)users_data = pd.read_csv(users_path)
city = SimulationEngine(config, stations_data, users_data)city.run(until=3600)

Batch experiments

To run a batch of simulations with different configuration files, we can use the single-run example as a template and repeat the call to the simulation for each desired configuration. For example, given two configuration files configA.json and configB.json, a simple for loop can help to sequentially execute a simulation for each configuration file.

import jsonimport pandas as pdfrom src.SimulationEngine import SimulationEnginefrom preprocessing.BikeGeneration import BikeGeneration
configuration_paths = ["configA.json", "configB.json"]stations_path = "stations.csv"users_path = "users.csv"
for config_path in configuration_paths:
    with open(config_path) as f:        config = json.load(f)
    stations_data = BikeGeneration(config["NUM_BIKES"], config["MODE"], stations_path)    users_data = pd.read_csv(users_path)
    city = SimulationEngine(config, stations_data, users_data)    city.run(until=3600)