Some particle filter algorythms don't even use the odometry at all, when the robot moves, the part of the cloud where the robot moves naturally become a better prediction and the cloud move as a result.

Good to note the pros and cons of particle filters compared to a Kalman filter. Haven't watched the Kalman filter video yet, but for any budding engineer taking a dig through these, the trick is sometimes to leverage more than one filter. E.g., during initialization or multi-peak conditions (e.g. kidnapped robot states), the particle filter can help solve the problem, but then smartly switching to a Kalman filter, or using different time-scales for the two filters (e.g. long-period particle filters, and short-period Kalman filters), can help address the performance/capability tradeoffs.

And don't forget to take a look at extensions of these filters. E.g. Extended Kalman Filter (Locally linear approximating Kalman filter for non-linear systems).

Thanks for doing this. This was one of my research projects a few months ago and this video is a great refresher.

Good video.

I wish the voice was more human, but the visuals fit the description very well.

The robot has multiple guesses of where it is at all times

You have incredible content. Thanks for making it!

Hi, can you make videos on bayesian hierarchical model and hierarchical models?

Though the video is really good it was also highly repetitive

Nice video as always! Though if may say something, this one was a tad repetitive. Like the same concept was explained 3 or four times. Do way to drive the message home, I suppose, but the video could have done that work shorter.