I noticed that those files contained a [VELOCITY_CONTROL] group that will stop working in gazebo-yarp-plugins 4 once the velocityControlImplementationType parameter, introduced in gazebo-yarp-plugins v3.5.1 https://github.com/robotology/gazebo-yarp-plugins/blob/master/CHANGELOG.md#351---2020-10-05 , will be made compulsory instead of automatically use the default value direct_velocity_pid. If those files are not mantained, why we don't just tag a release that contains them, and then remove them from the master branch of the repo?

While working with HDE, and try to test with human models, I noticed I do not have the default human models stated in xml files.
Therefore, HDE implicitly has dependency to this repo. I suggest, we move this repo to robotology, and I add it to superbuild for HDE installation.

CC @traversaro @DanielePucci @lrapetti @Yeshasvitvs @claudia-lat

Currently, the urdf models present in this repository are used primarily in Rviz for visualization. The possibility to use the human urdf model for dynamic simulations is limited for the time being as the low level gain tuning for the joint motor control was quite complex to achieve with in the infrastructure of yarp. Thanks to @Giulero who pointed out Matlab Simscape Multibody ability to import urdf files. It is a very straight forward process and the urdf model is imported to a simulink model as shown below:

With the human model imported to simulink, it is easier to analyze the dynamics and find the proper PID values for the low level control for all the joints in the model.

@diegoferigo @lrapetti @claudia-lat @traversaro @DanielePucci

Following #2 it is better to have a proper readme file with pictures to provide basic details on what this repository contains.

I cloned the "human-gazebo" repo, and tried to spawn the urdfs using

Gazebo showed a white box instead of a human body. Could you please tell me how to fix this issue?

The model falls from the sky when I spawn it and then bounces back and the breaks.
How should I fix this?
As mentioned, I have changed the values of masses and inertia of fake links to 0.0001 and 0.1 respectively. Should I change it to something else?
Also, what is the purpose of fake links?

In some recent tests related to human-dynamics-estimation, we suspected there may be some inconsistencies in the model geometry and inertial parameters. So, we decided to add a model test script that will assist in assessing the human models saved in this repo. A matlab script model_test.m is available at feature/model-test.

@prashanthr05 @lrapetti @claudia-lat @traversaro @kouroshD @DanielePucci

Will you please let me know how the joints can be controlled on this URDF model in ROS?

Is there a workaround to enable this control of the joints?

There is a bug in the auto-generation of the human urdf models. The LeftUpperLeg link visual radius value is negative for all the models:

https://github.com/robotology/human-gazebo/blob/master/humanSubject01/humanSubject01_48dof.urdf#L663

@claudia-lat please fix this bug in the urdf autogeneration script

CC @lrapetti @kouroshD

Hello,

For each "Subject", you specify in the urdf this :

<!--To open this model with Gazebo replace the null masses, inertias and dimensions of the 'fake links f1 and f2'-->

I'm sorry if it's already specified somewhere but please, can you tell me why theses fake links? and what are the correct values to put? If you have an urdf filled with the correct values ​​and you agree to transmit it to me, you would be of great help to me :)

Regards

In the master branch, 8 new urdfs models have been uploaded and old model have ben deleted.
However, those urdfs currently don't have the configuration for yarp and gazebo, for this reason I have created a new branch gazebo-model with the old model and its configuration files.

Hi all,

This is a question.
I am using the URDF model to visualize the real time movement of the model in RViz. I have attached IMU sensors on the joints (mainly upper joints 2 sensors on each hand and one attached to the pelvis, in total 5) and streams fused Euler angle data.
I have a joint state publisher that subscribes to the data and I send it to the respective joints. This can rotate the joints. However I am planning to use a camera to capture the human motion and visualize the motion in the URDF model.

Therefore, I am capturing the motion capture of human using a stereo camera and I am using openpose to extract information 3D points of a person and can visualize it in RViz. Do you think there is any way to use the visual Point data to replicate the same motion in the URDF?

Please let me know if you need more information

Thanks

Currently the 48dofs model have joint constraints based on anthropometrics tables.
Given the human model

I would expect those constraints to be symmetric between left and right limbs. The symmetry, though, should involve the constraints to be flipped (min_A=-max_B, max_A=-min_B) for those joints around x and z axis accordingly to the frame definition. However, looking at the models, this is not true. In fact, the constraints are defined as follow:

Arms

jLeftC7Shoulder_rotx -5.0000 45.0000
jRIghtC7Shoulder_rotx -5.0000 45.0000

jLeftShoulder_rotx -90.0002 134.9997
jRIghtShoulder_rotx -134.9997 90.0002

jLeftShoulder_roty -90.0002 90.0002
jRIghtShoulder_roty -90.0002 90.0002

jLeftShoulder_rotz -179.9998 45.0000
jRIghtShoulder_rotz -45.0000 179.9998

jLeftElbow_roty -90.0002 85.0000
jRIghtElbow_roty -90.0002 85.0000

jLeftElbow_rotz -145.0001 0
jRIghtElbow_rotz 0 145.0001

jLeftWrist_rotx -60.0001 50.0000
jRIghtWrist_rotx -50.0000 60.0001

jLeftWrist_rotz -20.0000 30.0000
jRIghtWrist_rotz -30.0000 20.0000

Legs

jLeftHip_rotx -30.0000 45.0000
jRIghtHip_rotx -30.0000 45.0000

jLeftHip_roty -15.0000 120.0003
jRIghtHip_roty -15.0000 120.0003

jLeftHip_rotz -45.0000 45.0000
jRIghtHip_rotz -45.0000 45.0000

jLeftKnee_roty 0 134.9997
jRIghtKnee_roty 0 134.9997

jLeftKnee_rotz -30.0000 40.0000
jRIghtKnee_rotz -40.0000 30.0000

jLeftAnkle_rotx -35.0000 45.0000
jRIghtAnkle_rotx -35.0000 45.0000

jLeftAnkle_roty -30.0000 50.0000
jRIghtAnkle_roty -30.0000 50.0000

jLeftAnkle_rotz -50.0000 25.0000
jRIghtAnkle_rotz -25.0000 50.0000

If i try to import one of the urdf file in Gazebo, only the Pelvis link spawn.
In line 9 there is the following comment:
To open this model with Gazebo replace the null masses, inertias and dimensions of the 'fake links f1 and f2
I have changed all the values and now all the links spawn correctly. Despite that, the robot is unstable (See Video).

In some recent tests I discovered the file humanSubject08_66dof.urdf is actually a 48 DoF file. This originated from the commit eebc204.

CC @lrapetti @claudia-lat

Analogous to what was done in #25, we should port the same change to all the models

How can we tune a humanoid joint in gazebo if we are applying the joint effort to it on the gazebo.

Issue #11 (comment) revealed a bug in all the models stored in this repo. Models have to be regenerated by following these instructions https://github.com/dic-iit/human-model-generator#how-to-generate-models.

Hi,

I am trying to use the human_Subject01_66dof_colored. It was mentioned in the urdf file to replace the null masses, inertia and dimensions of the fake links f1 and f2 for using the model in Gazebo.

I used the same masses, inertia for the actual link on the fake links. Initially it was all collapsing. Then I added gazebo plugin of gazebo_ros_control and associated library in the urdf (at the end before the tag ). I also gave the z value to be equal to the height of hip from ground (z = 0.91) and the model stood up fine.
But when I tried to make the moveit package along with the plugin it didnt work. So I made a copy of the urdf, removed the gazebo plugin loaded it in moveit and made left and right arm groups and generated its moveit config.

Next, I spawned the robot with gazebo_ros_control plugin in gazebo and then when I was trying to launch the move_group files from the generated moveit_config it says the robot model is different.

Currently I have added transmission tags with each joint, also have added position controller through a yaml file and still the robot collapses and I am being unable to figure out to have a urdf being spawned in gazebo and for which I can make a moveit_package that I can control afterwards through python code.

Guidance requested please

Thank you

Best Regards
Hasan

At the moment this repo contains only the models generated with 66DoFs. It would be useful to have also the same model with 48DoFs.

@claudia-lat @Yeshasvitvs

Following the investigation during risk assessment sprint, thanks to @lrapetti insights, we noticed the base vs human configuration mismatch is due to the human joint limits in 48 DOFs. The idea is to remove those limits similar to 66DOF models.
cc @Yeshasvitvs