QArm Lab Procedure

Lead Through
Setup
1. Launch Quanser Interactive Labs and load the QArm workspace

2. Launch MATLAB and browse to the working directory for Lab 3 – Lead Through

Lead Through (Learn)

1. Open LeadThrough_Learn.slx. This model uses a welding operation to represent a lead through a
learning process. You will first complete the mapping of the keyboard I/O to control the movement
of the robotic arm. Then you need to move the arm through a pre-set operation path and store all
the joint angle data.

Figure 1: LeadThrough_Learn Simulink model.

2. Open the subsystem called Keyboard I/O. This subsystem, shown in Figure 2 contains an incomplete
section called Map keys to end-effector speed. You will need to connect the keyboard inputs to the
corresponding command inputs to the QArm.

Figure 2: Incomplete Forward Kinematics subsystem.

1
3. Double click on the Host Keyboard block and take note of the keys that the block outputs, and in
what specific order. Close the block.

Figure 3: Orientation of the QArm Workspace

4. Note that the demux breaks the 7x1 vector output from the Host Keyboard block into 7 individual
signals. Connect the first output of the demux (labelled space for you) to the first output of this
subsystem (labelled Learn?). This will ensure that while you hold the space key on your keyboard,
the Learn? signal will remain high, indicating that the manipulator record the trajectories.

5. View the manipulator from the top in a way that the manipulator’s end-effector points south in the
home position, as shown in figure 3. The up and down arrow keys should drive the end-effector
correspondingly in the North and South directions, which in turn correspond to a positive or negative
x-direction. Add a gain block to the two channels labelled up and down. Set the value of these gains
to +1 or −1 depending on the correct directions.

Hint: The down arrow key would move the manipulator in the South direction, which is +X.

6. Repeat the process above for the demux outputs labelled right and left to drive the end-effector
along the y axis.

7. Repeat the process above for the demux outputs labelled Q and A to drive the end-effector along
the z axis (+𝑧 is out of the view towards you and −𝑧 is into the view).

8. Add a sum or add block to add the outputs of the gain pairs from steps 5 and wire that to the input
on the mux labelled v_x. Do the same for the gain pairs from step 6 and 7 to the mux inputs labelled
v_y and v_z, respectively.

9. Place a gain block between the output of the sum or add blocks and the mux you wired it to in step
8. This will help you set the correct magnitude and handle units for command conversion. We want
the maximum speed of the end-effector to be 5 cm/s or 0.05 m/s. Set the gains to 0.05. Thus, as
you tap the up or down arrow keys, the output of the corresponding add block equals −1, 0 or 1,
and the v_x input sees −0.05, 0 or 0.05 m/s. Do the same for the v_y and v_z channels.

10. After the three speed signals are grouped through the mux block, you need to convert the speed to
a position increment. The constant qc_get_step_size is the simulation time step. You need to multiply
the speed output of the mux with time step constant block to get the position increment. Use a
product block. Wire the output of the product block to the checkKinematics block.

11. Return to the root level and double click on the subsystem called Blackbox. This subsystem records
the joint angles and position data to the workspace. Add a To Workspace block and wire the phi
(rad) [4] input to it. Double click on the block and change the variable name to phi_trajectory. Under
the Save format dropdown menu, select Structure. Under Sample time section, replace −1 with to
qc_get_step_size.

12. Repeat step 11 with another To Workspace block connected to the p (m) [3] input. Use the variable
name p_trajectory and set the other settings as step 11.

2
13. At the root level, you want the Blackbox to record data when the Learn? signal is high. Which signal
do you connect that accomplishes this?

14. Copy the Forward Kinematics MATLAB function to this model.

15. Prior to running the model, open the model’s Configuration Parameters and verify that they are
configured as follows:

1. Solver type: Fixed-step

2. Solver: ode4 (Runge-Kutta)

3. Fixed-step size (fundamental sample time): 500 Hz

16. View the manipulator from the top as shown in Figure 4. You can hold the left-click mouse button
and drag to change the camera view in Quanser Interactive Labs.

Figure 4 Top view of the manipulator

17. Run the model using the green Play button under the Simulation Tab of your model. Once
started, the model will command 0 rad angles to all four of the manipulator’s joints.

18. At the root level in your Simulink model, use Q and A to move the manipulator away or towards the
table (along z axis). Use the four arrow keys to move the end-effector in the corresponding direction
specified by the arrow keys. Do so with the Space key depressed (not recording).

19. Move the end-effector to [0.25, 0.25, 0.10] (approximately). Monitor the end-effector’s location
through the display in the Simulink model.

20. The welding procedure is to start from the default [0.25, 0.25. 0.10] to A [0.25, 0.25, 0.01] to B
[0.25, 0.50. 0.01] to C [−0.25, 0.50, 0.01] to D [−0.25, 0.25, 0.01] then to A, and back to default.. This
emulates holding the end-effector by hand and driving it along the desired path, while it learns the
joint positions. Practice several times to get familiar with the keyboard input and its corresponding
direction.

3
 21. When ready, repeat the procedure while holding down the space key as all the joint positions and
the end-effector locations will be stored in a Blackbox. If you make a mistake, stop the model and
restart to erase stored data. Do not release the space key during the welding procedure.

22. After one complete cycle, stop the Simulink model and close it.

Lead Through (Follow)

1. Open LeadThrough_Follow.slx. This model will take the trajectory data as inputs to recreate the
welding process.

Figure 5 LeadThrough_Follow Simulink model

2. Open the Playback subsystem, where you will read the position and angles recorded by the
Blackbox of the LeadThrough_Learn.slx. Add a From Workspace block, and double click to open it.
Set the Data: field to out.p_trajectory and uncheck the Interpolate data option. Replace 0
in the Sample time field with qc_get_step_size. Set the Form output after final data value by field to
Cyclic Repetition to repeat trajectories. Connect it to the output port labelled position trajectory.

3. Repeat step 2 to read the workspace data out.phi_trajectory with another From Workspace
block.

4. At the root level, connect the output position trajectory to the first input of the Tracking scope and
connect the output phi trajectory to the first input of the plant Phi Cmd (rad) [4].

5. Copy the qarmForwardKinematics MATLAB function you completed in the previous lab to this model.

6. Connect the output p4 of the qarmForwardKinematics function to the second input of the Tracking
scope.

7. Prior to running the model, open the model’s Configuration Parameters and verify that they are
configured as follows:

1. Solver type: Fixed-step

2. Solver: ode4 (Runge-Kutta)

3. Fixed-step size (fundamental sample time): 500 Hz

4
8. Run the model using the green Play button under the Simulation Tab of your model. Use the
tracking scope to comment on how well the manipulator tracks the recorded trajectory and the
repeatability performance metric. Take a screenshot of a full welding cycle in the scope.

9. Stop the model, close it, then close MATLAB and Quanser Interactive Labs.