Back to Posts

Designing an Unscented Kalman Filter

Posted in All Posts, Projects

Everything described in this post was developed as a partnership between myself and James Palmer HMC ‘18. He deserves half the credit for it all. The header photo is from this paper.

This past semester, my friend James Palmer and I been in Autonomous Robotic Navigation, a class covering the control, localization, and motion planning of mobile robots. We’ve been using the Pololu Romi robot to implement a number of robotic algorithms, including point tracking, odometry, particle filtering, and probabilistic road maps. Our code for this class is available on Github.

For our final project for the class, we designed and implemented an Unscented Kalman Filter (UKF).

We gave a presentation (slides) on the UKF, as designed by Uhlmann in the early 2000s. After implementing it, we gave a presentation (slides) on our implementation of it on the Pololu Romi.

Here’s a video of the UKF working to control a Pololu Romi robot!

Read Next

Working at Amazon Prime Air