CrowdBot Challenge


Moving robots in dense crowds is new. One reason this topic has not advanced as much as it could have is that it is difficult to evaluate technologies. There is no crowd easily available in labs to perform tests of the technologies under development. It is not easy to gather crowds to perform user-studies. All these factors slow progression of research on this topic. A good solution to alleviate these difficulties is to perform tests in simulation environments, but they are not available now. To accelerate dissemination our results in simulation environments and to motivate research on the topic, we decided to organize a challenge on robot navigation in crowded environments based on the CROWDBOT simulation platform. Of course, simulation results cannot directly transfer to real environments, but this will start a virtuous cycle: simulation environment will boost research on navigation in crowds; back, this research activity will require always more realistic simulation environment.


Our goal for the CrowdBot Challenge is ambitious: we want to organize around the simulation tools developed in the CrowdBot project an annual campaign and become the standard for evaluating robot navigation in populated environments. There is still a long way to go to reach this goal, and we need to think about how to meet this objective beyond the end of the CrowdBot project.



Useful links


The wiki is a documentation that allows to understand the challenge and to use it.


GitLab allows error reporting, access to source code, 3D models, scenarios and more.

Executables v0.4

Executables allow you to quickly install the challenge locally.


The deliverable 7.3 of CrowdBot, downloadable on our website, presents in detail the challenge.

*To request access to the Wiki and Gitlab, please fill out the following form.


This project was completed in December 2021. If you would like to discuss the project, please contact Julien Pettré by searching for his email address directly on Google.