Photograph of the Phase IV arm, removed from the table. Phase IV built upon the aluminum arm that was developed in Phase IIIc. The addition of a smooth operating roll axis wrist for pouring operations was the most notable addition to the arm. Control electronics were greatly improved to provide more image scanning and processing power. The software was again completely rewritten, this time in C. A Mattel PowerGlove interface was to provide a means of controlling the arm, but the software was not completed for this capability.

Vital Statistics

Additional Photos

Phase IV arm base, showing the base motor with the winching drum, the shoulder motor, linear encoder optical interrupter, threaded rods for attaching to the base bearings, and the underside of the wiring panel.	Phase IV arm, showing the shoulder, wrist motor, limit switch optical interrupter, linear encoder optical interrupter, and wiring panel.	Phase IV arm as viewed from the rear. Note the wrist control mechanism with the small DC motor and bicycle brake cable. The concept was to eliminate additional weight at the extremes of the arm. The wrist was the first experiment in implementing a cable-driven scheme where the motors could be located elsewhere. I had originally hoped to build a 3 or 4 fingered articulated gripper using this drive scheme, and even had gone as far as buying a dozen motors and bicycle cables.
The Phase IV gripper and wrist. The gripper was based on the one constructed for Phase IIIc, but it sported several useful additions. Perhaps most important was the addition of a feedback potentiometer so that the control PC could read the position of the gripper finger. The photoreflector sensor from Phase IIIc was removed in favor of a pair of tactile switches. Conductive foam was replaced with a more dense (albeit ugly) foam. The magnetic sensor from earlier phases was also installed for detection of magnetized objects.	Detail of the two bearings with mounts used to support the new wrist roll motion (intended for pouring applications). You can clearly see the feedback potentiometer for the wrist.	Closeup of the connection between the wrist motor winch drive and the bicycle brake cable. The line on the right is high-test deep sea fishing line. The clamp holding the bicycle cable is constructed out of aluminum angle channel.
A shot of the wrist in action. The gripper is holding a 16oz soda bottle, which just barely fits in the grippers when fully open. It is easy to see the gripper feedback pot, wrist springs, and gripper limit detectors in this photo.	The Phase IVc interface chassis. Originally this chassis was about 6 inches tall and contained a 286 PC motherboard, power supply, ISA I/O cards, and a front panel with 40x2 LCD and 12 key keypad. However, due to reliability problems with the 286 hardware, I removed it in favor of a more reliable but certainly less "cool" system. The big red button on the front was the emergency kill switch.	The ISA I/O card used in Phase III and IV of the project. In Phases IVa and IVb, this I/O card resided in the interface chassis, connected to a 286 PC. In Phase IVc, the card resided in a 486 PC and connected to the chassis on the left.
The control electronics PC board. The left side of the board contains 18 ice cube type relays for motor control. The middle top section of the board contains the analog multiplex circuitry and ADC for the image scanner and feedback potentiometers. The bottom middle contains the digital demultiplex circuitry for controlling the analog multiplex circuits. The two D-style connectors on the right connected to the arm. This PC board was too big (about 16" by 10") and too messy, but it served the purpose. Several parts had been removed by the time this photo was taken, such as the transistors used to power the relays, some ICs, etc.	The underside of the control electronics board. It was laid out with no real schematics to speak of, with tape and donuts directly on the PC board.

Software Screenshots

One frame of a FLI animation from the RACS 4.x software introduction screen. This was rendered on a friend's copy of DOS 3DStudio. I think the 16-frame animation rendering took a ridiculous amount of time, even on a fairly fast PC at the time.
RACS 4.30 GUI, showing a few windows open on the left, and all windows open on the right. The background image is a frog, which has nothing to do with the project, but it looked cool at the time. The software supported a joystick for real-time control, and was to eventually support the powerglove as well (I could obtain X, Y, Z info from the glove, but I never used it for anything). Scanner functions, task execution, and diagnostics each had their own windows.
Closeups of the scanner window, showing a raw image before normalling and grayscale reduction on the left. The image on the right shows the normalled image with a small amount of reduction. The image was generated by laying a number of empty soda cans on the scanner to spell out "MIT," which is where the state science fair was held.
More images detailing additional image processing. Further grayscale reduction is illustrated on the left, while final 1-bit output is shown on the right.