simobility is a light-weight mobility simulation framework. Best for quick prototyping

simobility is a human-friendly Python framework that helps scientists and engineers to prototype and compare fleet optimization algorithms (autonomous and human-driven vehicles). It provides a set of building blocks that can be used to design different simulation scenarious, run simulations and calculate metrics. It is easy to plug in custom demand models, customer behavior models, fleet types, spatio-temporal models (for example, use OSRM for routing vehicles and machine learning models trained on historical data to predict ETA).

Create an environment for experiments with machine learning algorithms for decision-making problems in mobility services and compare them to classical solutions.

Some examples:

• Deep Reinforcement Learning with Applications in Transportation

• T. Oda and C. Joe-Wong, "Movi: A model-free approach to dynamic fleet management". 2018

• A. Alabbasi, A. Ghosh, and V. Aggarwal, "DeepPool: Distributed model-free algorithm for ride-sharing using deep reinforcement learning", IEEETrans. Intelligent Transportation Systems (to appear). 2019

• C Wang,Y Hou, M Barth, "Data-Driven Multi-step Demand Prediction for Ride-hailing Services Using Convolutional Neural Network". 2019

• J. Ke, F. Xiao, H. Yang, and J. Ye. Optimizing online matching for ride-sourcing services with multi-agent deep reinforcement learning. 2019

pip install simobility

Thanks to all who contributed to the concept/code:

• Steffen Häußler
• Stephen Privitera
• Sultan Imanhodjaev
• Yábir Benchakhtir

Grid world simulation

Simple simulation

Taxi service

Log example

Benchmark simulations with LinearRouter and GreedyMatcher. Simulations will run slower with OSRMRouter because OSRM cannot process requests as fast as the linear router.

Processor: 2,3 GHz Dual-Core Intel Core i5; Memory: 8 GB 2133 MHz LPDDR3

A heuristic that allows estimating a maximum number of booking a fleet of N vehicles can handle: assume that an avarage trip duration is 15 minute, than 1 vehicle can not more then handle 4 booking per hour and the upper limit for 1000 vehicles is 4000 bookings per hour.

{
    "avg_paid_utilization": 63.98,
    "avg_utilization": 96.87,
    "avg_waiting_time": 292.92,
    "created": 3998,
    "dropoffs": 589,
    "empty_distance": 640.37,
    "empty_distance_pcnt": 33.67,
    "fleet_paid_utilization": 63.98,
    "fleet_utilization": 96.87,
    "num_vehicles": 50,
    "pickup_rate": 15.48,
    "pickups": 619,
    "total_distance": 1902.04,
}

The are multiple ways to collect simulation log - use CSV or InMemory log handler or implement your own handler: loggers

Read CSV logs with pandas:

import pandas as pd

data = pd.read_csv(
    "simulation_output.csv",
    sep=";",
    converters={"details": lambda v: eval(v)},
)

details = data.details.apply(pd.Series)

wget http://download.geofabrik.de/north-america/us/new-york-latest.osm.pbf
docker run -t -v "${PWD}:/data" osrm/osrm-backend osrm-extract -p /opt/car.lua /data/new-york-latest.osm.pbf
docker run -t -v "${PWD}:/data" osrm/osrm-backend osrm-partition /data/new-york-latest.osrm
docker run -t -v "${PWD}:/data" osrm/osrm-backend osrm-customize /data/new-york-latest.osrm
docker run -d -t -i -p 5010:5000 -v "${PWD}:/data" osrm/osrm-backend osrm-routed --algorithm mld /data/new-york-latest.osrm