Introducing AMoDeus.

AUTONOMOUS MOBILITY-ON-DEMAND SIMULATOR

AI Driving Olympics Logo | AMoDeus - Autonomous Mobility-on-Demand Simulation

Proud co-organizer of the Artificial Intelligence Driving Olympics beginning Oct 1, 2018.

Orchestrate your own fleet of robotic taxis!

AMoDeus is a simulation framework you can use analyze and synthesize autonomous mobility-on-demand systems.

Get started in seconds with AMoDeus's basic scenario, pre-configured with static travel demand. Go in-depth with benchmark algorithms and dedicated analysis tools providing insight into service levels and fleet efficiency.

Up and running? Monitor, understand and verify your results with the powerful built-in viewer.

AI Driving Olympics Logo | AMoDeus - Autonomous Mobility-on-Demand Simulation

Proud co-organizer of the Artificial Intelligence Driving Olympics beginning Oct 1, 2018.

What is the AMoD technology?

Autonomous Mobility-on-Demand is a novel technology in which self-driving vehicles transport customers on-request from their desired pick-up to their desired drop-off destination.

The technology promises to revolutionize transportation systems as it may offer levels of comfort as high a private transportation at the personal and environmental cost of public transportation.

AMoDeus is a tool designed to solve the AMoD design problem.

Car Icon (40px) | AMoDeus - Autonomous Mobility-on-Demand Simulation

What can AMoDeus do for you?

  • Rapid simulation results for your publication: verify your algorithms in a high-fidelity environment with ease.
  • Quantificative evaluation of potential for new cities: rapidly understand what potential autonomous mobility-on-demand has for your geography of interest without the need of lengthy software implementations.
  • Stand on the shoulders of programmers: we provide a viewer, automated analysis for AMoD systems, benchmark dispatchers and many other features in an open-source basis. Benefit from a well-tested and maintained codebase.
  • Contribute to the community: become a contributor and make your work available and visible to the global research community on our platform.

AMoDeus is Open-Source.

This makes it absolutely free to use and open for code contributions.

To get started, visit our GitHub repository and follow the instructions provided in README.md:

You will also need a specific transportation scenario. Either download one of the scenarios provided on this page or create your own based on other datasets.

AMoDeus uses various libraries including MATSim. It therefore opens up to you the full range of possibilities offered by MATSim without the need to deal with its full complexity at once.

For assistance, input, feedback please don't hesitate to contact us or send us a pull request on GitHub!

Getting Started

You may work on any operating system with a Java-compatible IDE. This repository demonstrates how to use the AMoDeus library in your project.

We provide installation guidelines as well as various demos:

  • Take control of the fleet by writing your own dispatcher for a fleet of robotic taxis and send them around in the city, see this demo.
  • Customize your analysis: The analysis of an autonomous mobility-on-demand system requires the processing of large amounts of data. We have simplified this, run your custom-analysis faster as shown in this demo.
  • Change the routing of the vehicles: you would like to take control of the path your taxis take in the city. Start here.
  • Move from unit-capacity taxis to taxis with many seats: write a shared dispatcher as shown here.

Source Code & Contribution

The AMoDeus source code is available on github in this repository, it is open-source and regularly updated and maintained by the AMoDeus team. We are very happy to add you as a contributor, please contact us if you are interested.

We have the following design principles that we enforce whenever possible, when writing contributions, please adhere to them:

  • Files are as short as possible and have less than 200 lines of code in any case.
  • Minimal redundancy.
  • For mathematical operations that require exact precision we use the tensor library.
  • Code should be documented and covered by tests.

Scenarios

For interested researchers, we present some AMoDeus scenarios available for testing and extension. The links lead to working directories containing all necessary files that can be used with the sequence ScenarioPreparer, ScenarioServer and ScenarioViewer as explained in the main GitHub repository.

  • San Francisco Taxi Scenario (ZIP, 5.1 MB): this scenario represents a static demand extracted from San Francisco taxi data available online. All trips that occurred on the first day of recordings, 17th of May, 2008, are included. The linkspeeds are adapted to match the travel times shown in the dataset. Original dataset from: M. Piorkowski, N. Sarafijanovic-Djukic and M. Grossglauser can be found here.
  • Berlin Scenario (ZIP, 69.7 MB): the existing MATSim Open Berlin Scenario presented by D. Ziemke and K. Nagel and available here was altered such that all car trips now need to be served by autonomous mobility-on-demand in an efficient way.
  • Santiago de Chile (ZIP, 51 MB): the scenario presented in this publication by B. Kickhöfer, D. Hosse, K. Turner and A. Tirachini available here was altered such that all public transportation and "colectivo" trips now have to berved efficiently with a fleet of robotic taxis.
  • Tel Aviv (ZIP, 28.8 MB): the scenario presented by G. Ben-Dor, B. Dmitrieva, M. Maciejewski, J. Bischoff, E. Ben-Elia, and I. Benenson presented here was taken and the entire transportation demand (both public and private transportation) needs to be served with a fleet of autonomous vehicles.

AI Driving Olympics

AI Driving Olympics Logo | AMoDeus - Autonomous Mobility-on-Demand SimulationAMoDeus is a co-organizer of the AIDO (AI Driving Olympics) event beginning October 1, 2018 at NIPS 2018, the conference on Neural Information Processing Systems in Montreal.

We are providing an interface to AMoDeus such that you can design and run fleet management algorithms based on artificial intelligence on our scenarios. Can the neural network beat the benchmarks? Who will win the competition? There are three tasks:

  1. Service quality challenge: for a fixed number of robotic taxis, what strategy provides the highest service quality while keeping fleet efficiency relatively high?
  2. Efficiency challenge: for a fixed number of robotic taxis, what strategy provides the most efficient fleet operation while keeping the service level acceptable?
  3. Fleet size challenge: the average waiting time of customers has to stay below a threshold, but what fleet can solve the task with the smallest number of vehicles?

The participants' challenge will be to solve the general problem. The final test will be on a surprise city scenario!

For more information, check out the AIDO website or get in touch with us!

Documentation

Car Icon (40px) | AMoDeus - Autonomous Mobility-on-Demand Simulation

Publications

If you use AMoDeus, we are happy if you cite us:

Claudio Ruch, Sebastian Hörl, and Emilio Frazzoli. Amodeus, a simulation-based testbed for autonomous mobility-on-demand systems. In: Proc. 21th IEEE Conf. Intelligent Transportation Systems, 2018.

The following research is based on AMoDeus software:

  • Hörl, Sebastian, et al. "Fleet control algorithms for automated mobility: A simulation assessment for Zurich." Transportation Research. Part C, Emerging Technologies (2018).

Help

If you need support running AMoDeus, please refer to our Frequently Asked Questions or contact us.

About Us

AMoDeus™ was developed by the Institute of Dynamic Systems and Control at ETH Zürich in collaboration with the Institute for Transport Planning and Systems at ETH Zürich.

Logo: IDSC | AMoDeus - Autonomous Mobility-on-Demand Simulation

Logo: IVT | AMoDeus - Autonomous Mobility-on-Demand Simulation

Contributors in alphabetical order:

Institute for Dynamic Systems and Control: Andrea Censi, Andreas Aumiller, Christian Fluri, Claudio Ruch, Emilio Frazzoli, Jan Hakenberg, Joel Gächter, Lukas Sieber, Marc Albert, Maximilien Picquet, Nicolo Ormezzano, Francesco Seccamonte, Samuel Stauber.

Institute for Transport Planning and Systems: Sebastian Hörl.

Contact

For feedback, inputs or to have an espresso break with us in Zürich, please contact Claudio Ruch:

Invalid Email