Robotics StackExchange | Archived questions

AMCL odom_alpha1 etc settings

Hi

Whilst I am working on trying to fix my odometry, I am finding these settings in AMCL hard to understand. I have pulled the below from the code so that we see the default, min and max settings for each one:

"odom_alpha1", double_t, 0, "Specifies the expected noise in odometry's rotation estimate from the rotational component of the robot's motion.", .2, 0, 10 
"odom_alpha2", double_t, 0, "Specifies the expected noise in odometry's rotation estimate from the translational component of the robot's motion.", .2, 0, 10 
"odom_alpha3", double_t, 0, "Specifies the expected noise in odometry's translation estimate from the translational component of the robot's motion.", .2, 0, 10 
"odom_alpha4", double_t, 0, "Specifies the expected noise in odometry's translation  estimate from the rotational component of the robot's motion.", .2, 0, 10

I note that odomalpha5 is not relevant to me as I have a differential style robot (<param name="odommodel_type" value="diff"/>).

I note I am sticking to odommodeltype diff, not diff-corrected - as I dont have any reference frame for suitable values for the diff-corrected model.

So two questions really:

1) What does each setting actually mean in more natural language ? I mean odomalpha1 I think I get - if the odometry says we are rotating, how acurate is the actual rotation in relation to what the odometry tells us - am I correct? Similarly odomalpha_3 - if the odometry says we are moving forwards or backwards, how acurate is the actual distance traveled in relation to what the odometry tells us?

2) What are the scale of the units in each setting, ie is the robot odom says it has gone forwards 1m, does an odom_alpha3 setting of 0.2 mean we should expect it to have gone somewhere between 0.8m and 1.2m forwards?

Thanks very much Kevin

Asked by KevWal on 2023-07-28 02:09:28 UTC

Comments

Answers

Read Probablistic Robotics: https://docs.ufpr.br/~danielsantos/ProbabilisticRobotics.pdf

Its not explained in detail because its commonly understood as the bible of probablistic robotics techniques which has this stuff explained in detail.

Asked by stevemacenski on 2023-07-28 17:24:23 UTC

Comments

Thankyou, great to see that it is avaliable, but it is a 500 page highly mathmatical book, and a search for alpha_1 didnt find any hits :)

I think it is probably better that I stick to 4 high numbers (alpha1 to alpha4) representing my poor odometry and spend my effort on understanding other things...

Thanks for the link through, it is appetiated.

Asked by KevWal on 2023-07-30 04:38:26 UTC

You should read it. Its considered the basic description of this technology which essentially all roboticists have read.

Asked by stevemacenski on 2023-07-31 13:47:15 UTC