While iCub plays air hockey, it is required to keep the gaze as stationary as possible in order to minimize the effects of the ego-motion in the camera images.
The standard approach that recruits IK and head stabilization via inertial feedback tends to be slow if compared with the velocities at stake. To get around this problem, one could consider solving IK offline and thus running at full speed being limited only by other low-level factors such as the performance of the PID controllers.
The problem has been analyzed and tackled in the following studies.
cmake -S . -B build
cmake --build build/ --target installThe code has been tested in Gitpod 🌐
Find out more on YARP-enabled Gitpod workspaces 🔎
Just run yarpserver, launch first the system app
and then the study-related applications.
👨🏻💻 Study 1
This study aims to sample the joint trajectories spanned by the torso, the arm and the head as found by IK while the hand moves along the allowed path.
The samples are conveniently stored in a table. An example is given in table.tsv,
where the columns have the following meaning:
| 1 | 2...4 | 5...11 | 12...17 |
|---|---|---|---|
| end-effector y-coordinate | torso joints (reversed order) | arm joints | head joints |
We can also plot the content of the table by doing
cd plotting
./plot.sh <path-to-the-table-file>and thus getting something like the following graph:
| Example table |
|---|
 |
The module can be configured via command-line options up to a certain extent.
👨🏻💻 Study 2
We replay here the target joints saved within the table in the previous study. The main idea 💡 is to be able to go as fast as possible having spared the burden of doing IK online.
The output trajectories are interpolated using splines. To this end, we make use of
ttk592/spline as a dependency.
The desired location of the hand along the allowed path can be provided by sending
a number within [-1, 1] to the YARP port /study-air-hockey/target.
| Example output |
|---|
 |