Main ideas for this demostration and steps to follow

[rsantos88]

We are interested of create a demostration which we can see Teo manipulating an object with two arms. It can be something with handles or something that it's necessary to grasp with two hands at the same time. The objective is to check the correct functionality of the robot, manipulating with two arms. It's not necessary to use path-planning. We can grab waypoints (see this old issue for that) like teo-self-presentation and make something simple and nice to see.

UPDATED:

• At first, the robot complete a path with the differents waypoints grabbed to avoid the obstacles (table) and reach the object to be grasped (tray).
• After this, the robot performed some movements with two arms moving the tray in the three cartesian axes.

Next to do:

• [x] improve the movement of the tray (https://github.com/roboticslab-uc3m/teo-bimanipulation/issues/3)
• [x] be able to grip the tray correctly with the current hands
• [x] close the control loop using the torque force sensors

[rsantos88]

One possible idea of a demo is:

1. Teo grasps and object which is necessary to use the two hands to hold it. It can be a tray for example. The object will be known and it cannot be change. It's necessary to position the object with respect to the robot always in the same distance and location.
2. The robot will elevate the tray. Once the robot holds it in the air, you will can place an object over the tray.
3. At this point, I thought that Teo can do something like:
   • Say the weight of the object using the jr3 sensors
   • Try to set the object in the center, balancing the tray and using jr3 to know the forces in each hand. In this case, the object should be a sliding object.

[rsantos88]

Talking with @smcdiaz, we've reached to a conclusion of a possible demo. First approximation is:

1. Teo takes a tray with the two hands holding it. Once it's in the air, Teo moves the tray, doing movements with both arms in the air. He will do some displacement of the tray to ensure proper operation of the arms. This movement will be record using waypoints.
2. The second goal is to use the JR3 sensors. Once the arms are raised, apply a force in one hand and move it in the direction of the applied force. This force and it direction will be measure with the JR3 sensor. This can be do with the other hand too. For example, you move the right hand, and the left hand moves in the same direction with the same velocity.

[rsantos88]

Pending of buy a tray to start and get to work with this demo

[rsantos88]

The tray has been bought. Some handles have been added to the tray adapted to the current hands of Teo.

[rsantos88]

Blocked by https://github.com/roboticslab-uc3m/teo-hardware-issues/issues/22

[rsantos88]

Unblocked thanks to https://github.com/roboticslab-uc3m/teo-hardware-issues/issues/22

[rsantos88]

Blocked again by https://github.com/roboticslab-uc3m/teo-hardware-issues/issues/14#issuecomment-382696473

[rsantos88]

Unblocked thanks to https://github.com/roboticslab-uc3m/teo-hardware-issues/issues/14#issuecomment-392781382 but blocked anyway by https://github.com/roboticslab-uc3m/teo-hardware-issues/issues/26

[rsantos88]

Next to do:

• [x] improve the movement of the tray (https://github.com/roboticslab-uc3m/teo-bimanipulation/issues/3)
• [x] close the control loop using the torque force sensors

[rsantos88]

• [x] close the control loop using the torque force sensors

In order to close the control loop, I've done some advances. I've understood the performance of the JR3 sensor. I have linked the output of the sensor with the movement of the arm in Cartesian positions, so when we push the hand in one direction of the three axes, the arm will follow that direction in the Cartesian axes, using position mode.

--> See video ! :smiley:

[rsantos88]

I've edited a small video that shows the advances that have been made for this demo so far :smiley:

[rsantos88]

I have already made several advances, which were on standby due to various changes that have had to be made in the robot. Changes at the software level, such as the review and unify joint limits , the change of the direction of rotations of various joints in order to unify all in the same sense with respect to their axes of coordinates, the change and unify TEO kinematic model issue, and the test of different operating modes in position (Make position direct mode usable) with the aim of achieving the best movement in cartesian coordinates. Also, there have been several problems at the hardware level that have stopped the tests, such as the break of the right forearm or problems with values obtained by some absolute encoders. Finally, and after having solved these problems, here I can show a video of the application working:

Some notes:

• Closed the control loop using the torque force sensors. This point has been finished :)
• After trying different modes of operation of the drivers in position (subject investigated by @PeterBowman ), the best method that works for me is the old PTMode, separated in the branch: https://github.com/roboticslab-uc3m/yarp-devices/tree/old-pt
• The tray has had to rotate down, because the upper surface is curved. The lower part is flatter, although it can be seen that in the center there is a logo, preventing them from rotating to their true center (this can be seen in the video when both balls appear separated in the center of the tray)
• In relation to the point of replace the tray grips (sponges) with a harder material for a correct transmission of force, I haven't had any serious problems. Just note that sometimes, when the robot moves tray, the JR3 sensor, which had been calibrated at the beginning of the execution, later takes a non-zero value at rest, leading to an erroneous measurement and diverting the center of gravity of the tray. This problem seems to be related to the pressure that the robot exerts on its sensors with the tray while is moving it. The simplest solution has been to release part of the sponge that surrounded the handles of the tray, making less the lateral pressure on the JR3 sensors.