STRAIN2 bno055 orientation

[traversaro]

In the model generated using simmechanics, we do not have the frames corresponding to the bno055 sensors mounted on the new STRAIN2 electronics. As adding them to the CREO cad will take time (if it is worth at all) it would be already helpful if we could extract the transform between the sensor frame of the bno055 documented in Page 24 of ~https://ae-bst.resource.bosch.com/media/_tech/media/datasheets/BST_BNO055_DS000_14.pdf~ updated URL https://www.bosch-sensortec.com/media/boschsensortec/downloads/datasheets/bst-bno055-ds000.pdf and the FT sensor frame defined in http://wiki.icub.org/wiki/FT_sensor .

If the STRAIN2 electronics is already model in the iCub's CAD, I think this should be trivial.

[traversaro]

I extracted manually for now the transform. It is:

{}^{ft_sensor_frame} H_{bno055_sensor_frame} 
= 
-1.00   0.00    0.000     0.00750013
 0.00   1.00    0.000     0.00860000
 0.00   0.00    -1.00    -0.0150180
 0.0     0.0      0.0     1.0

(linear translation part in meters).

See the attached image for reference (note: creo uses millimeters for distance):

[traversaro]

Important document for getting the orientation of the bno055 in the STRAIN2: STRAIN-2_11996_B_ASS_TOP_ASS_BOT.pdf

[traversaro]

cc @hu-yue

[traversaro]

cc @fiorisi

[hu-yue]

cc @nunoguedelha here is the transformation between the IMU and the FT.

[LOC2Bot]

cc @nunoguedelha and @fiorisi as discussed F2F, we should regenerate the URDF with the sensor frames.

[fiorisi]

I had a look at the problem. As mentioned F2F the frames are not in the FT sensors used in the CAD model of iCub. More in details, in the CAD model are used the old FT sensor (1st generation) that don't have the new strain2 board. In order to generate the model we have to:

• [x] talk with mech. department to understand if we can replace the FT sensors in the iCub and Koroibot CAD models -> OK
• [x] replace the FT sensors in the iCub and Koroibot CAD models
• [x] check that the mass properties of the FT are consistent (i.e. the geometries in the electronic board model are all solid)
• [x] replace the FT sensors in the iCub 2 CAD models
• [x] redefine the Simplified Representation of the links that include the FT sensors (and check all the others)
• [x] add the IMU measurement frame in the FT sensor model (is not defined)
• [x] add the frames to be exported in the CAD model
• [x] regenerate the ShrinkWraps of the links that include the FT
• [x] check the mass of the new ShrinkWraps
• [x] replace the new links in the Creo Mechanism model
• [x] export the Creo Mechanism model to SimMechanics
• [x] update the .yaml file
• [ ] regenerate the URDF model

[fiorisi]

I checked with S. Dussoni, and the sensor should be located on the TOP face of the strain2 board (yellow pins in the figure).

To correctly define the orientation of the frame we need to locate the PIN 1 in the chip (in the manual is highlighted with the black dot).

From the electronic schematics we see that is the lower right pin.

This is confirmed by the small white dot on the chip.

The frame has been placed in the center of the chip (on the upper surface).

[fiorisi]

If the placement of the sensor frame is correct, the transform is different from the one obtained by @traversaro.

[nunoguedelha]

Hi @fiorisi I've crosschecked and find the same orientation as you did, as long as the board has the default configuration P1, among the possible values below (AXIS_REMAP_CONFIG, AXIS_REMAP_SIGN registers):

[traversaro]

Thanks @fiorisi and @nunoguedelha for the check. I am not sure were I did the error, probably it was when placing the frame in the CAD as the document linked in https://github.com/robotology/icub-model-generator/issues/91#issuecomment-391323851 contains the dot of the U10 board in the right position.

@hu-yue this was probably affecting the results of the analysis of the IMU euler angles output.

[fiorisi]

The FT are connected to the EMS boards. The frames in the CAD model will be named as follows (similar to the convention used for the gyro and accelerometers): SCSYSL/R BODY-PART FT-IMUELECTRICAL-NAME

The ELECTRICAL-NAME is documented here.

cc @nunoguedelha

[fiorisi]

Here the names of the frames in the CAD model: SCSYS_L_UPPER_LEG_FT-IMU_3B12 SCSYS_R_UPPER_LEG_FT-IMU_3B11 SCSYS_L_FOOT_FT-IMU_3B13 SCSYS_R_FOOT_FT-IMU_3B14 SCSYS_L_UPPER_ARM_FT-IMU_1B6 (not exported yet) SCSYS_R_UPPER_ARM_FT-IMU_2B6 (not exported yet)

[nunoguedelha]

The FT are connected to the EMS boards.

As a side note, it's a pitty the number of the EMS is not in the FT board electronic name, as it is the case for the other sensors. For instance :

• board MTB 10B2 is the MTB board #2 connected to the EMS eb10, and thus named: SCSYS_L_UPPER_LEG_MTB_ACC_10B2,
• Gyro directly placed on the EMS eb4, named: SCSYS_CHEST_EMS_GYRO_EB4.

[fiorisi]

@nunoguedelha, I actually wanted to add it but the name of the features in Creo is limited to 31 characters.

[fiorisi]

In the iCub Mechanism CAD model the FT sensor frames of the arms are defined in the link the contains the FT sensor itself.

As specified in the documentation of the simmechanics-to-urdf:

Note that for now the FT sensors sensor frame is required to be coincident with child link frame, due to URDF limitations.

Currently there is a frameReferenceLink parameter to remap the frame to the correct link, but while redefining the Mechanism model, we can also move the FT frames in the CAD model and update the simmechanics-to-urdf configuration file.

[traversaro]

Note that for now the FT sensors sensor frame is required to be coincident with child link frame, due to URDF limitations.

The actual limitation is in the SDF, in particular it is tracked by this issue: https://bitbucket.org/osrf/sdformat/issues/130/position-part-of-force_torque-sensor-pose .

[fiorisi]

The mass of the following links is different. I am trying to understand what happened. Maybe I will release a first version with the different masses. Please remember to double check that the simulations behavior is not different and to check that visually the inertia box is similar to the old ones (inside gazebo).

CAD LINK NAME	Mass difference c.a. [kg]	COMMENTS
SIM_SEA_2-5_L_ELBOW	0.06	Inertia tensor is also very different
SIM_SEA_2-5_R_ELBOW	0.06	Inertia tensor is also very different

The difference in mass is not relevant, the real problem is the inertia tensor.

[fiorisi]

PR opened: https://github.com/robotology/icub-model-generator/pull/111

cc @FabioBergonti

[traversaro]

This was fixed by https://github.com/robotology/icub-model-generator/pull/122, thanks to everyone that contributed.