ANKI COZMO LOCALIZATION

A GETTYSBURG COLLEGE AI 4th HOUR PROJECT (Spring 2022)

Our Goals:
– understand Anki Cozmo SDK and how to program with it.
– be able to run previous group’s code.
– implement MCL algorithm through sampling method.
– rotate Cozmo to its home state (angle).

Setting up the libraries:
It is important that you set enough time to set up these libraries and make sure you are familiar with them so that you minimize loss off time due to library-related issues deeper into the project. As of Spring 2022, the Cozmo SDK is only compatible with Python 3.7 (and prior versions). Make sure you install the right version that can work with your Cozmo. If you have a newer version of python that is not compatible, you will need to uninstall all the newer versions of python and install a version that works.

Software used:
– PyCharm: test and run python code (through Command Line).
– Anki Cozmo mobile app: necessary to use Anki Cozmo SDK.

Our source code:
Check out the source code for the project.

Outcome of project:
We managed to get Cozmo to take the panorama pictures and collect subsequent snapshot pictures to get sensor data to compare with those from the panorama. We managed to get the sampling algorithm into python but was not able to incorporate it into the code for a complete Monte Carlo Localization. As a result, despite being able to rotate, we didn’t manage to get the “home” angle to which Cozmo should return to.

Key points to improve upon:
– Incorporating the particle population into the sampling function.
– After Cozmo is confident on where its location is, accurately rotating cozmo to its initial “home” state/angle.
– Provide a simple way to show Cozmo’s confidence level on where it is located with respect to its environment.