Lighthouse 2.0 Two Base Station Flight Instability

Topics related to the Lighthouse positioning system, configuration and use
Post Reply
stebucur
Beginner
Posts: 1
Joined: Mon Aug 30, 2021 12:04 am

Lighthouse 2.0 Two Base Station Flight Instability

Post by stebucur »

Hi,

I have been using the Lighthouse Positioning System for a University Project and have had a few issues lately with the stability. When using a single Lighthouse 2.0 Base Station the accuracy and stability when flying Crazyflie 2.X drones appears to be as shown in the bitcraze videos. When adding a second base station however, I have encountered severe flight instability. I am aware the lighthouse beams occaisionally intersect and thus the drone has moments when the receiving indicators go red for both base stations however I believed that this would be negligible and the Crazyflie would be able to fly relatively well under its own onboard sensors until the momentary receiving loss ends.

I have double checked the channels of each base station, and all indicators appear to be green when checked within the Crazyflie Client before flying. Furthermore, the flight area contains no reflective items. I should mention I am using the latest firmware for the Crazyflies and decks as of this post alongside having a mix of 2.0 and 2.1 drones (However they both behave the same).

Video of behaviour can be seen here: https://www.youtube.com/watch?v=bNMVvXOC79s
kristoffer
Bitcraze
Posts: 630
Joined: Tue Jun 30, 2015 7:47 am

Re: Lighthouse 2.0 Two Base Station Flight Instability

Post by kristoffer »

Hi!

What you see is indeed a result of the interference between the base stations like you say.
What happens is that the information from the base stations is fed into to kalman estimator that tries to figure out the position of the Crazyflie. If the information from the two base stations (or other information sources) conflicts, the estimator will pretty much end up with an estimate that is the "average" position of the sources (that is a point in between). If you block one of the base stations, the estimator is left with only the other one and the estimate position will converge to the position suggested byt that base station. If you un-block the base station again the estimate slides back to the average position. This is essentially what happens when the base stations interfere, you typically loose the information from one of the base stations for a short time.

This would not be a problem if the information from both base stations pointed to the same location, but unfortunately this is not the case. There are two main sources for these errors.

1. Base station calibration data. Everything is not perfectly aligned in the base stations when manufactured in the factory and to handle this there is a model that compensates for the errors. The properties of each base station is measured in the factory and the model parameters are stored in the base stations (we call it the calibration data). We can read the calibration data but we are not sure exactly how the compensation model works (it is not public). This means that the are some errors in the angles we measure from the base stations. Improving the model is work in progress.

2. Base station geometry. The position and orientation of the base stations are calculated by using a function in OpenCV. The algoritm uses the angles we get from the base stations to calculate the pose so any errors in the angles (see above) will affect the estimated geometry. It is also possible that the algorithm is sensitive to noise/errors which does not help (needs further investigations).

I think both 1 and 2 can be improved but it requires work (and time). For 2 it would be nice to record angles from more than one position and create an algorithm that distributes the errors more evenly in space as opposed to the current solution that minimizes the error in the origin.

You can measure the error in any point by using the crossing beam algorithm and look at the lighthouse.delta log variable that shows the distance (sort of) between the solutions for the two base stations.

You can always try to re-estimate the geometry to see if you get better results.
Moj_Dev
Beginner
Posts: 10
Joined: Wed Jan 19, 2022 7:43 pm

Re: Lighthouse 2.0 Two Base Station Flight Instability

Post by Moj_Dev »

Hi stebucur,
I am interested to know did you fix your problem? Do you have still this flight instability with the latest firmware?
kristoffer
Bitcraze
Posts: 630
Joined: Tue Jun 30, 2015 7:47 am

Re: Lighthouse 2.0 Two Base Station Flight Instability

Post by kristoffer »

Hi Moj_Dev!

There has been some progress related to this, check out the latest blog post https://www.bitcraze.io/2022/01/improve ... stimation/

You could try it out and see if it makes any difference
Post Reply