This tutorial will guide you to the use of Motor Interfaces to control the iCub movements in the joint space relying on the following modalities:
We will show you how to control the elbow joint of the iCub arm in three different contexts:
- With a single call delivering the final position set-point to the low level layer.
- With a controller thread that continuously computes the velocity commands to reach for a target.
- With a planner thread streaming suitable position references used by the low level layer.
In the end, we will obtain the joint moving as below.
Launch the applications available in ./app/scripts using yarpmanager: first the "system" part and then the "app" part.
The three modules open up the following three RPC ports:
/position/velocity/positiondirect
Each RPC provides the following services:
goto initiate the movement.encto read the encoder feedback.
Use case:
$ yarp rpc /position
>> goYou might want to play with our yarpmotorgui, which lets you interact with the joint control of the robot.
To do so, type:
$ yarpmotorgui --robot icubSimThe high-level user code provides a one-shot command in terms of final position set-point to achieve along with the time required to attain it (or, equivalently, the corresponding speed). Then, the low-level control is responsible for generating a suitable hard-coded trajectory (typically minimum-jerk profiles) connecting the current joint angle with the desired set-point as well as for controlling the motor to let the joint reach for the target. Thus, the control takes place exclusively at the low-level, whereas the high-level only sends the command to initiate the movement.
The high-level user code is responsible for continuosly providing joint velocity to be used to solve a given task defined upstream (e.g. reach a position). The low-level layer is only dealing with the actual control of the motor, instantly integrating the received velocity into a position set-point. Therefore, part of the control complexity is shifted to higher level.
The high-level user code is responsible for continuosly providing position references to be used to solve a given task defined upstream. The low-level layer is only dealing with the actual control of the motor. This modality allows for doing position planning at higher level, since we may not need to read back the encoders to perform control upstream as in the case of velocity control. As a result, we can achieve faster movements (compared with closed-loop approach) with the freedom to plan generic joint trajectories (compared with profiles hard-coded in the firmware).
