Due Wednesday Mar 12, 11pm

Worth: 9% of your final grade

In the first part of this lab you’ll attach a manipulator arm to your robot. This arm has three rotating joints and a gripper, all actuated by Robotis AX12 “Dynamixel” robot servos (doc). These receive battery power via the LLP and are controlled by a serial communications interface also on the LLP. Monitor commands are provided (and their Java and Scheme correspondents) to move the arm and to query its state.

In the second part of the lab you’ll write high level code on the HLP to move the gripper to commanded locations in world frame, using the robot’s turn-in-place capability to add a fourth rotating degree-of-freedom (DoF) to the arm workspace. In the third part you will program the robot to drive towards the location of a known object on the ground, deploy the arm, and grasp the object.

Robot Assembly

  1. Get your arm assembly, 4 servos and cables, “lollipop” manipulation object, mast bracket, 4 4-40x5/8" screws, 4 4-40x3/8" screws, and 16 M2x8mm screws.
  2. Disconnect the battery, IR sensors, bump switches, drive motors and encoders. Remove the 4 4-40x3/8" screws holding the top plate to the vertical aluminum standoffs and remove the top plate.
  3. Remove the 4 4-40x3/8" screws holding the PandaBoard to the top plate from underneath. Set the electronics stack aside.
    top plate disassembly
  4. Attach the servos to the underside of the top plate using the M2x8 screws. Make sure the screws are snug. The servos are not all the same. They are numbered (on the back near the electrical connectors) 0,1,2,3. Number 0 goes in the robot’s right front, 1 in the left front, 2 in the right rear, and 3 in the left rear.
    servos
    servos right
    servos left
  5. Attach a cable to servo 1 through a nearby hole in the top plate. Then “daisy chain” the remaining servos together. The two connectors on the back of each servo are interchangeable. Finally, manually rotate the output sprocket on each servo to line up the white marks.
    servos front
  6. Position the arm flat on the table with the gripper closed. Slip the belts over the servo sprockets. Try not to let the servos rotate too much from the white-mark alignments. Hold the arm in place with two 4-40x5/8" (long) screws through the top plate in the front inner corners of the shoulder brackets.
    attach arm
  7. Feed a 4-40x3/8" screw up through the PandaBoard mounting hole towards the outside of servo 1. You’ll probably need to temporarily loosen the servo.
    servo 1 screw
  8. Temporarily remove the PandaBoard red/black power cable so that you can set the electronics stack upside down on the Orangutan board LCD. Flip the top plate/arm assembly over and re-attach it to the electronics stack using 4 4-40x3/8" screws. Start with the one you already inserted near servo 1, and use the small screwdriver for that one. Be careful not to flex the PandaBoard while it is attached with only one or two screws. It’s ok of the arm moves now.
    panda reassembly
  9. Now turn your attention back to the bottom plate. Remove the two outer 4-40x5/8" (long) screws holding the rear wheel brackets.
    mast bracket 1
  10. Position the mast bracket with the notched outer tabs above the same holes from which you just removed screws in the bottom plate, and then reassemble those screws through the tabs.
    mast bracket 2
  11. Remove the two front (but not the two rear) aluminum standoffs that were holding the top plate. They will not be needed now because the arm shoulder brackets will take their place.
    disassemble standoffs
  12. Place the top plate back onto the bottom plate. Make sure the cable between servos 2 and 3 (in the rear) is not pinched. Carefully feed the wheel motor and encoder wires back up between the arm belts and through the square hole in the front of the top plate. Reattach all wires to the PandaBoard (but don’t connect the battery yet). Remember to double check that the red/black wires for the left and right motors are not crossed. Also remember when reconnecting the IR sensors that the black wires go towards the front.
    top plate reassembly
  13. Use 8 4-40x3/8" screws (four of them from the front standoffs you removed) to attach the top plate to the rear standoffs and the mast bracket.
    mast bracket 3
  14. Use 2 4-40x5/8" (long) screws to attach the underside of the arm shoulder brackets to the bottom plate. shoulder bracket screws
  15. Plug the arm servo daisy chain into the connector on the (robot’s) left side of the Orangutan board. arm plug
  16. Have the course staff check your work before reconnecting the battery.

Preparing the Code

Follow similar instructions as for lab 2 to update your svn checkout and copy the lab 3 framework into robotics/gN/l3.

Then rebuild and flash the LLP monitor code using the same procedure as for lab 2.

Finally, rebuild the HLP OHMM jarfile using the same procedure as for lab 2.

About the Arm

The monitor program includes a set of ax* commands to interact directly with the AX12 servos (documented in robotics/llp/monitor/ohmm/ax12.h). It also includes a set of a* commands to interact with the servos in the context of the manipulator arm (documented in robotics/llp/monitor/ohmm/arm.h). It is strongly recommended to use only these higher-level arm commands except for debugging and calibration purposes.

The Java OHMM library and the Scheme OHMMShell also provide access to all these commands.

The notes for Lecture 9 give details of the arm kinematics, including the link lengths. They also present an analytic IK approach for this arm.

Arm Testing and Calibration

  1. Bring up OHMMShell or use a serial terminal program like minicom to talk directly to the OHMM monitor program running on the LLP. Then run the following commands to enable the arm and send it to the calibration pose:

    OHMMShell
    > (ae #t)
    > (ac)
    
    minicom
    > ae 1
    > ac
  2. The calibration pose is intended to have the arm pointing straight forward with the gripper closed. This pose should be reached by setting each servo to the goal position 512 in AX12 counts (the AX12 servos have a range of motion of 0 to 300 degrees, with 0 degrees corresponding to 0 counts and 300 degrees corresponding to 1023 counts, so 512 counts corresponds to 150 degrees, with the servo at the midpoint of its range of motion). However, various sources of error will cause each arm to reach a slightly different pose when all servos are commanded to 512.

    Very carefully using commands like

    OHMMShell
    > (axsg I C)
    
    minicom
    > axsg I C

    where I is a servo index (0 = shoulder, 1 = elbow, 2 = wrist, 3 = gripper) and C is the goal counts (start with 512 and work in increments of about +/- 10 counts at a time), try to adjust your arm so that it appears to be in the correct straight-forward calibration pose.

    You may also judiciously tension the gripper to close with a little bit of force during this process. Try adding and subtracting 10 counts from the gripper servo 3 first to understand which direction closes and which direction opens the gripper. Then open the gripper to a small gap, followed by closing it in small increments until the fingers are just touching. Continue closing by 10 or 20 further counts to tension the gripper a bit. (Note: the gripper state actually is set by the difference between servos 2 and 3, so you may also need to fiddle with servo 2 to maintain the calibration pose.)

  3. Once you are satisfied with the calibration, “freeze” the current targets of each servo as the new calibration:

    OHMMShell
    > (afac)
    > (agac #f)
    
    minicom
    > afac
    > agac s

    Write down the four calibration values, you will need them if you later re-flash the LLP. The arm should not move when you do this, but if you then test the calibration by running

    OHMMShell
    > (ah)
    > (ac)
    
    minicom
    > ah
    > ac

    the arm should move to the home pose when you run ah and then back to the calibration pose when you run ac. The calibration values will be stored on the LLP in nonvolatile EEPROM so so that you should not have to re-calibrate unless you re-flash the LLP (in which case the EEPROM will be erased).

    If you ever want to reflash the LLP, you can first read out the arm calibration data stored in the EEPROM by running

    OHMMShell
    > (agac #f)
    
    minicom
    > agac s

    Then write it down, do the reflash, and then reset the calibration values to EEPROM with

    OHMMShell
    > (asac C C C C)
    
    minicom
    > asac C C C C

Inverse Kinematics

Using the inverse kinematics method of your choice, implement a program that runs on the HLP (this is a requirement), homes the arm, and then responds to keypresses as follows, with step distance initially:

w — move the gripper + mm in the world-frame direction
s — move the gripper - mm in the world-frame direction
a — move the gripper + mm in the world-frame direction
d — move the gripper - mm in the world-frame direction
r — move the gripper + mm in the world-frame direction
f — move the gripper - mm in the world-frame direction
q
z

Assume that the robot is at world frame pose initially.

To enable direction motions use turn-in-place drives to add a “yaw” DoF to the arm kinematics. You may want to use the dos monitor command, part of the solution code for the drive module documented in robotics/ohmm-sw-site/llp/monitor/ohmm/drive.h as driveOrientationServo(float). This is available at the java level as OHMMDrive.driveOrientationServo(float) and in scheme as (dos <float>). The dos command is not queued like df and dt. You can check whether the robot has already reached the desired orientation by repetitively running dos with the same target angle; the return will be nonzero (true) until the angle has been reached to within a tolerance band.

Use the wrist joint to keep the gripper horizontal (parallel to the ground, as it is in arm home pose) at all times.

If the target location is not reachable, your program must emit a message. However, it is up to you to decide how to move the arm in the case of unreachable targets. For example, you could make a best-effort motion to reach towards the target. However you do it, you must maintain the following property: any sequence of keypresses followed by the reverse sequence, e.g. “wwrrffss”, must return the arm to the home pose. This should hold even if the arm target is unreachable at any time during the sequence.

You may find that reading a single keypress from the console is not easy in Java; in fact, technically, it’s not possible to do in a portable way using the current standard Java API. However, here we will accept a solution which runs on Unix; for that see the provided code in ohmm.ConsoleNonblocking. Alternatively, we will accept solutions which open a graphics window and then respond to KeyEvents there.

Grasping

Retrieving objects on the ground is a common task in mobile manipulation. In this part, you will implement a basic method to approach an object, grasp it, and carry it. The manipulation “lollipop” object is tall enough so that the gripper can engage it when lowered to within a few centimeters of ground clearance.

  1. Using your horizontal-gripper IK solution from above, write a program to produce a yaw-free grasp sequence starting from home pose:

    1. open the gripper
    2. lower the arm with a interpolation so that the gripper is about 5cm from the ground
    3. move the arm forward with a robot-frame interpolation by about 2.5cm
    4. close the gripper
    5. return to home pose.

    The robot chassis should remain in place at all times. Note that the interpolations must approximate straight-line motions in space. It will not work to go directly to the joint angles at the ends of the interpolations; you need to generate intermediate waypoints along the interpolation lines at some rate and use inverse kinematics to convert each of those waypoints into intermediate joint angles.

  2. Calculate the robot frame offset distance from the robot frame origin where an object on the ground should be grippable at step d. Practice picking up the object specifically placed at robot frame coordinates using the whole sequence of motions above. Tune the sequence as you see fit to maximize grip success.

  3. Extend your program so that it can take the coordinates of the object on the ground as command line arguments in world frame millimeters (if no command line arguments are given, then the program should revert to keyboard control inverse kinematics as above). For example, if your Java class to solve this part is called Grasp, you should be able to invoke it like this

    > ./run-class Grasp 30 40

    for a goal at mm in world frame. Assuming the robot starts at pose in world frame, have your program

    1. home the arm
    2. turn towards the object
    3. drive straight until it is within a radius (the above-determined offset distance) from the object
    4. stop
    5. engage your gripping sequence to acquire the object
    6. return home to the world frame pose (keeping the arm in its home pose)

Note: you may need to re-calibrate the robot wheel-to-wheel baseline distance to improve the accuracy of turn-in-place motions. You may optionally take a third command line parameter giving the robot baseline distance in mm (thus allowing it to be adjusted at the beginning of each run).

Grading

You will be asked to demonstrate your code for the course staff in lab on the due date for this assignment (listed at the top of this page); 30% of your grade for the lab will be based on the observed behavior. Mainly want to see that your code works and is as bug-free as possible.

The remaining 70% of your grade will be based on your code, which you will hand in following the general handin instructions by the due date and time listed at the top of this page. We will consider the code completeness, lack of bugs, architecture and organization, documentation, syntactic style, and efficiency, in that order of priority. You must also clearly document, both in your README and in code comments, the contributions of each group member.