47 replies to this topic

[Lew Payne]

Paul - This evening I stumbled upon some rather remarkable claims for the DCM implementation.  Here are select excerpts:

Quote

What I am saying is that DCM is a non-linear approach based on the rotation group that provides a better solution to the problem of estimating aircraft attitude than Kalman filters does. It has superior performance and does not require a lot of computation. [...] With DCM, there are are no unknown model parameters to be computed, it is based entirely of the kinematics of rigid-body rotation. [...] DCM is fully functional the moment you turn it on. [...] Paul and I have an elegant, computationally efficient, implementation of DCM that is done entirely with multiplication and addition. There no divisions, and certainly no matrix inversions. [...]  I chose to use only the estimation portion at this time, because it is more than adequate for what I want to do with it. It is only the estimation portion that is running in Louis's demo. [...] So it is my opinion that Mahony is really on to something, and that his approach to estimation and control for UAVs will become the "gold-standard".

The above really does make it sound like the implementation is a "breakthrough" by which all others will be measured (and switch to).  Yet, it seems to me, from reading about your personal experience with the algorithm, that you would tend to disagree with the above.  As one of the developers of said algorithm, I was hoping you could clear up some of my confusion.  I'm simply trying to determine if the new DCM algorithm is superior to a typical non-linear Kalman approach, or if it was simply over-hyped.  Your help in clearing this up would be appreciated.

[dankers]

Not my place to comment at all and I certainly have no clue, but the last I read was Bill was going to switch to Paul's EKF on the UAV DB, this was around one month ago. Also, I know that Paul is very very busy with work so if he takes a while to reply this would be why. A very interesting discussion and question Lew, I find all this stuff fascinating but so far over my head it need to be measured in kilometers .  

Lewcifer, on 08 June 2010 - 05:18 AM, said:

Paul - This evening I stumbled upon some rather remarkable claims for the DCM implementation.  Here are select excerpts:

The above really does make it sound like the implementation is a "breakthrough" by which all others will be measured (and switch to).  Yet, it seems to me, from reading about your personal experience with the algorithm, that you would tend to disagree with the above.  As one of the developers of said algorithm, I was hoping you could clear up some of my confusion.  I'm simply trying to determine if the new DCM algorithm is superior to a typical non-linear Kalman approach, or if it was simply over-hyped.  Your help in clearing this up would be appreciated.

[murmur]

atahk, on 02 June 2010 - 03:17 AM, said:

does your work use a honeywell imu?? The data output rate is not so much related to considerations for EKF computation, but more closely tied to your overall control architecture.

As a matter of fact, it is a Honeywell IMU.

And I think I was mistaken when I said the KF we use at work is a linear one. It's an EKF; it appears that the Nav guys are just not in the habit of calling it that for some reason.

[Lew Payne]

dankers, on 08 June 2010 - 07:00 AM, said:

Not my place to comment at all and I certainly have no clue, but the last I read was Bill was going to switch to Paul's EKF on the UAV DB, this was around one month ago. Also, I know that Paul is very very busy with work so if he takes a while to reply this would be why. A very interesting discussion and question Lew, I find all this stuff fascinating but so far over my head it need to be measured in kilometers .

Given that a new version of MatrixNav was just announced (which incorporates magnetometer and a few other things into the DCM algorithm), I'm still very curious to get the Paul's comments regarding this prior post of mine.  It doesn't seem like Bill is switching to Paul's EKF.  As a result, I'd really like to hear "from the man" himself.  Speculation just tends to muddy the waters.

[dankers]

My comment came from Bill's post on DIYDrones, it was not just speculation. Sorry you felt this was muddying the waters, I was trying to help.

Paul will likely not reply here, I could be wrong but don't expect it.

[Lew Payne]

dankers, on 16 June 2010 - 07:51 AM, said:

My comment came from Bill's post on DIYDrones, it was not just speculation. Sorry you felt this was muddying the waters, I was trying to help.  Paul will likely not reply here, I could be wrong but don't expect it.

Thank you, Dankers.  It sounded, by the way the original response was phrased, as if it was just a guess.  No offense meant... just trying to determine which way the DCM project is going.  I don't want to spend time on it if someone else has already determined that it doesn't "live up to its hype."  No doubt, you can sympathize with that.

[dankers]

I'd like to know as well which way DCM is heading as well, just for curiosity, it seems to change a lot. If you search Bill's posts on DIYClones you should see the post I was referring too, my comment was based on this post and Bill's words, not a guess but passing on what Bill wrote. If he planned to switch to Paul's EKF in the UAV DB it gave the impression to me that DCM was being dropped obviously. You could ask Bill directly about this?
  
There have been a few discussions behind the scenes about all this a while back, as I keep saying the technicalities are above my head but there are a lot of people I trust that make some good points. You have to listen to the podcast we did again way back in Feb. We were going to implement what we jokingly called the "Full Mahoney", this would be to extend DCM for Mags, this was long before Bill started on this. We dropped it as not as suitable as Pauls EKF, re-read this entire thread again, a lot of Paul's points are covered already. There is also the work of Seb Magwick on MARG, it would be worth talking to him as well if you can get him to be less candid that might be enlightening.  

None of these people read the thread here I would guess, Paul is flat out with professional work and won't have any time till August, even then he will be taking it easy.

  

Lew Payne, on 17 June 2010 - 06:40 AM, said:

Thank you, Dankers.  It sounded, by the way the original response was phrased, as if it was just a guess.  No offense meant... just trying to determine which way the DCM project is going.  I don't want to spend time on it if someone else has already determined that it doesn't "live up to its hype."  No doubt, you can sympathize with that.

[Lew Payne]

dankers, on 17 June 2010 - 08:25 AM, said:

There is also the work of Seb Magwick on MARG, it would be worth talking to him as well if you can get him to be less candid that might be enlightening.

I'm familiar with the work of Sebastian Madgwick, and I'm actually interested in testing his algorithm.  What I want to do, specifically, is figure out how to implement GPS latency compensation into his algorithm... the way a few others have done with the sigma-delta Kalman filter.  The cheap GPS units we use for these projects don't offer point source output mode, nor do they offer a latency computation output.  As a result, the measurements that we are correcting against are always way off, especially when traveling fast (100mph).  See my discussion here about it.

I am now overflowing with technical literature, to the point where it's difficult to find any one thing I read in the past (in addition to my memory not being as good as it was when I was younger).  As a result, I am slowly starting to put everything in one place, where I can comment in particular articles as to what caught my attention within them.  Here's my technical blog, which lists only a few of the articles I have read (and which are not all fully commented by me yet).  Hope that helps... please feel free to comment on, or add, articles yourself to it.