[iCubGenova09] Increase the speed of the low level (WBD 500Hz) #20

GiulioRomualdi · 2022-08-23T12:33:00Z

This PR decreases the period of the WBD from 0.01 s to 0.002s. The PR also increases the rate of the hardware devices and the network wrapper.

This comment will first details the profiling analysis and then I will analyze each parameter that I modified

Time profiling

WBD profiling

Before entering into the details of the PR I would like to show you the time profiling result of WBD. The analysis has been performed by exploiting the timers I implemented in robotology/whole-body-estimators#142

The following table contains the average time (the average time window is equal to 1 minute). The labels publish computed forces etc. refer to the different sections in the main loop in robotology/whole-body-estimators@d49bdd1 The label all indicates the time spent to perform the entire loop.

name	time average (s)	number of deadline miss	time at deadline miss (s)
publish	0.0004850322333	0	0
compute forces	0.0003449385246	0	0
calibration	5.510801666667e-07	0	0
kinematics	3.934341346667e-05	0	0
remove offset	8.4732465e-06	0	0
contact points	2.486420513333e-05	0	0
read sensors	4.683991183333e-05	0	0
all	0.0009521654851667	1	0.002652847

Accordingly to the profiling, WBD is able to handle this frequency 🚀

Board messages

The boards stream sometimes the following warning. However, the robot seems working as expected.

[WARNING]  from BOARD 10.0.1.11 (right_leg-eb11-j0_2), src LOCAL, adr 0, time 1664s 41m 268u: (code 0x0000000d, par16 0x012e par64 0x00010093003e01e0) -> SYS: the TX phase of the control loop has last more than wanted. In par16: TX execution time [usec]. In par64: num of tx can2 and can1 frames and latest previous execution times of TX, RX, DO [usec] + .

yarprobotinterface.txt contains the log of the yarprobotinterface. Perhaps @traversaro or @S-Dafarra have some hints.

PR analysis

iCubGenova09/estimators/wholebodydynamics.xml
- <param name="devicePeriodInSeconds">0.002</param> increase the frequency of wbd from the default value 100Hz to 500Hz
iCubGenova09/hardware/FT/left_arm-eb1-j0_1-strain.xml
- I decreased the tfPeriod from 10 ms to 2ms (500Hz)
iCubGenova09/hardware/electronics/face-eb23-j2_5-eln.xml
- I decreased the TXrateOfRegularROPs from 3 (or 5) to 2 (500Hz here you can find some documentation)
iCubGenova09/hardware/inertials/head-inertial.xml
- I decreased the acquisitionRate from 10 to 5 (200Hz)
iCubGenova09/hardware/inertials/left_arm-eb1-IMU.xml
- I decreased the acquisitionRate from 50 to 5 (200Hz). I noticed that all the inertials attached to the FT were running at 20Hz before this PR
iCubGenova09/wrappers/FT/left_arm-FT_wrapper.xml
- I decreased the period from 10 ms to 2ms (500Hz)
iCubGenova09/wrappers/inertials/left_foot-IMU_wrapper.xml
- I decreased the period from 10 ms to 2ms (500Hz)
iCubGenova09/wrappers/motorControl/left_arm-mc_wrapper.xml
- I decreased the period from 0.02 s to 0.002s (500Hz). It is important to notice that 0.02 s is the default value of the controlBoard_nws_yarp device so before this PR the rate was 50Hz.

This substitutes #10

GiulioRomualdi · 2022-08-23T12:33:37Z

cc @S-Dafarra @isorrentino @DanielePucci

S-Dafarra

Thanks a lot for the PR!

The comments are related to a set of changes that in my opinion is not straightforward. If you happen to have some cross-reference that indicates why those changes are needed, it could be useful for future reference.

iCubGenova09/hardware/electronics/face-eb22-j0-eln.xml

iCubGenova09/hardware/inertials/left_arm-eb1-IMU.xml

S-Dafarra · 2022-08-23T12:50:53Z

Also, did you notice any error or warning in the logger regarding timeouts?

GiulioRomualdi · 2022-08-23T14:01:36Z

Also, did you notice any error or warning in the logger regarding timeouts?

I'm going to check it on the robot right now

GiulioRomualdi · 2022-08-24T10:16:22Z

@S-Dafarra @traversaro I updated the comment #20 (comment)

S-Dafarra · 2022-09-28T17:54:29Z

@pattacini could we ask for your support to make sure that these modifications are:

increasing the frequency of IMUs to 100Hz
all MAS interfaces for the FTs go at 500Hz
all control boards go at 500Hz
the modifications to the electronics folder are coherent with the above modifications. This is where we are more in doubt

DanielePucci · 2022-09-28T18:09:16Z

CC @traversaro

pattacini · 2022-09-28T18:14:38Z

Hi @S-Dafarra

I'll be checking that tomorrow with the fixers.

pattacini · 2022-09-29T06:27:53Z

iCubGenova09/hardware/inertials/head-inertial.xml

@@ -43,7 +43,7 @@
            </group>

            <group name="SETTINGS">
-                <param name="acquisitionRate">      10      </param>
+                <param name="acquisitionRate">      5      </param>


If I got it right, it seems that 10 is the fastest allowed rate as per https://github.com/robotology/icub-main/blob/master/src/libraries/icubmod/embObjInertials/embObjInertials.cpp#L527-L538.

Have you seen the following warning message popping up?

embObjInertials::sendConfig2MTBboards() has detected a wrong acquisition rate =...

Note that this device is of type embObjIMU, so I am not sure if any limit like that exists.

If I got it right, it seems that 10 is the fastest allowed rate as per https://github.com/robotology/icub-main/blob/master/src/libraries/icubmod/embObjInertials/embObjInertials.cpp#L527-L538.

Have you seen the following warning message popping up?

embObjInertials::sendConfig2MTBboards() has detected a wrong acquisition rate =...

The 100 Hz frequency (acquisitionRate = 10 ms) is the maximum frequency that we can squeeze from the IMU Bosch BNO055. And that is due to HW limitation, in particular to the way the chip replies over the used I2C bus. We thoroughly explored this limitation some time ago.

So, we should correct the value to be 10.

On the other hand, in here the intention is to

increasing the frequency of IMUs to 100Hz

Note that this device is of type embObjIMU, so I am not sure if any limit like that exists.

Regarding this, the function in question is then at https://github.com/robotology/icub-main/blob/26c74355c9d6463762352f41f8f1a1df593844ca/src/libraries/icubmod/embObjIMU/eo_imu_privData.cpp#L314-L340.

Note that this device is of type embObjIMU, so I am not sure if any limit like that exists.

@traversaro: thanks for it. I checked and device embObjIMU does not check vs incorrect values of acquisitionRate as the original device embObjInertials did. Maybe in the migration towards the new device something was missed.

But the FW indeed does the check and clips values in range [5, 200] ms. The lower possible value is 5 because in some cases (e.g., when the board asks via I2C only one or two values such accel and gyro) 5 ms can be OK. But again, 10 ms is the recommended lower value that we should use. Also because the fusion algorithm inside the BNO055 chip produces new values only every 10 ms.

marcoaccame · 2022-09-29T09:09:42Z

iCubGenova09/wrappers/inertials/head-imuFilter.xml

@@ -3,7 +3,7 @@


    <device xmlns:xi="http://www.w3.org/2001/XInclude" name="head-imuFilter" type="imuFilter">
-        <param name="period">       20                  </param>
+        <param name="period">       2                  </param>


The hardware section instructs boards to retrieve new IMU values every 10 ms.

So, the best period of the wrapper is 10 ms. To have 2 ms is not wrong, but it may stress the system more than we need.

This comment applies to all the wrappers inside folder inertials.

GiulioRomualdi · 2022-09-29T13:03:06Z

Thank you @marcoaccame! I'm going to apply your suggestions See e2201aa and 1ce2566

marcoaccame · 2022-09-30T07:10:25Z

Thank you @marcoaccame! I'm going to apply your suggestions See e2201aa and 1ce2566

OK, now the PR now seems fine by me.

GiulioRomualdi · 2022-09-30T08:33:39Z

I will squash all the commits and I will test it on the robot. If everything goes smooth we can merge it

- increasing the frequency of IMUs to 100Hz - all MAS interfaces for the FTs go at 500Hz - all control boards go at 500Hz

GiulioRomualdi self-assigned this Aug 23, 2022

GiulioRomualdi mentioned this pull request Aug 23, 2022

Increase the lowlevel feedback rate to 1khz #10

Closed

S-Dafarra self-requested a review August 23, 2022 12:41

S-Dafarra reviewed Aug 23, 2022

View reviewed changes

iCubGenova09/hardware/electronics/face-eb22-j0-eln.xml Show resolved Hide resolved

iCubGenova09/hardware/inertials/left_arm-eb1-IMU.xml Outdated Show resolved Hide resolved

GiulioRomualdi force-pushed the devel_iCubGenova09_high_freq branch from 763cff1 to 00b7c42 Compare August 23, 2022 14:04

S-Dafarra force-pushed the devel_iCubGenova09_high_freq branch from d6fcda1 to 6ec6b8a Compare August 25, 2022 15:08

S-Dafarra force-pushed the devel_iCubGenova09 branch 2 times, most recently from a65de00 to e3aec73 Compare September 1, 2022 13:18

S-Dafarra force-pushed the devel_iCubGenova09_high_freq branch from 6ec6b8a to bb8074e Compare September 28, 2022 16:54

GiulioRomualdi mentioned this pull request Sep 28, 2022

Increase the frequency of the walking-controller to 500Hz robotology/walking-controllers#130

Closed

S-Dafarra force-pushed the devel_iCubGenova09_high_freq branch from bb8074e to e2830ba Compare September 28, 2022 17:49

pattacini reviewed Sep 29, 2022

View reviewed changes

marcoaccame reviewed Sep 29, 2022

View reviewed changes

GiulioRomualdi force-pushed the devel_iCubGenova09_high_freq branch from 1ce2566 to 8b98ecd Compare September 30, 2022 08:43

[iCubGenova09] Increase the frequency of the low level

274612c

- increasing the frequency of IMUs to 100Hz - all MAS interfaces for the FTs go at 500Hz - all control boards go at 500Hz

GiulioRomualdi force-pushed the devel_iCubGenova09_high_freq branch from 8b98ecd to 274612c Compare September 30, 2022 12:44

S-Dafarra merged commit 6d58d6f into devel_iCubGenova09 Oct 4, 2022

GiulioRomualdi deleted the devel_iCubGenova09_high_freq branch February 14, 2023 18:21

GiulioRomualdi mentioned this pull request Feb 16, 2023

[ergoCubSN000] Increase the speed of the low level (WBD 500Hz) #29

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[iCubGenova09] Increase the speed of the low level (WBD 500Hz) #20

[iCubGenova09] Increase the speed of the low level (WBD 500Hz) #20

GiulioRomualdi commented Aug 23, 2022 •

edited

Loading

GiulioRomualdi commented Aug 23, 2022

S-Dafarra left a comment

S-Dafarra commented Aug 23, 2022

GiulioRomualdi commented Aug 23, 2022 •

edited

Loading

GiulioRomualdi commented Aug 24, 2022

S-Dafarra commented Sep 28, 2022

DanielePucci commented Sep 28, 2022

pattacini commented Sep 28, 2022

pattacini Sep 29, 2022

traversaro Sep 29, 2022

marcoaccame Sep 29, 2022

pattacini Sep 29, 2022

marcoaccame Sep 29, 2022 •

edited

Loading

marcoaccame Sep 29, 2022 •

edited

Loading

GiulioRomualdi commented Sep 29, 2022 •

edited

Loading

marcoaccame commented Sep 30, 2022

GiulioRomualdi commented Sep 30, 2022 •

edited

Loading

[iCubGenova09] Increase the speed of the low level (WBD 500Hz) #20

[iCubGenova09] Increase the speed of the low level (WBD 500Hz) #20

Conversation

GiulioRomualdi commented Aug 23, 2022 • edited Loading

Time profiling

WBD profiling

Board messages

PR analysis

GiulioRomualdi commented Aug 23, 2022

S-Dafarra left a comment

Choose a reason for hiding this comment

S-Dafarra commented Aug 23, 2022

GiulioRomualdi commented Aug 23, 2022 • edited Loading

GiulioRomualdi commented Aug 24, 2022

S-Dafarra commented Sep 28, 2022

DanielePucci commented Sep 28, 2022

pattacini commented Sep 28, 2022

pattacini Sep 29, 2022

Choose a reason for hiding this comment

traversaro Sep 29, 2022

Choose a reason for hiding this comment

marcoaccame Sep 29, 2022

Choose a reason for hiding this comment

pattacini Sep 29, 2022

Choose a reason for hiding this comment

marcoaccame Sep 29, 2022 • edited Loading

Choose a reason for hiding this comment

marcoaccame Sep 29, 2022 • edited Loading

Choose a reason for hiding this comment

GiulioRomualdi commented Sep 29, 2022 • edited Loading

marcoaccame commented Sep 30, 2022

GiulioRomualdi commented Sep 30, 2022 • edited Loading

GiulioRomualdi commented Aug 23, 2022 •

edited

Loading

GiulioRomualdi commented Aug 23, 2022 •

edited

Loading

marcoaccame Sep 29, 2022 •

edited

Loading

marcoaccame Sep 29, 2022 •

edited

Loading

GiulioRomualdi commented Sep 29, 2022 •

edited

Loading

GiulioRomualdi commented Sep 30, 2022 •

edited

Loading