In 1961, the First Robot Arm Punched In

I was one of two people in the "research department" at Unimation. I believe we were the first to connect a general purpose computer to an industrial robot. The first connection was a Data General Nova, connected to a 2-arm robot. That project was led by Pete Rogers. The second was a PDP-11/03 connected to a Unimate. William Perzley led the "research department". Curiously his name is never mentioned. Also, no one ever mentioned Maurice Dunne, the chief engineer.

The 2-arm robot built speed regulators for automatic transmissions.

The Unimate worked on the problem of training a robot in a static position but playing out the training in time. The robot was trained to weld a car "in a static position" but later to weld the car as it went by on a conveyor line. I worked on the mathematics and computer programs to do the time/distance translation.

I also worked with Vic Schienman and did modifications to the VAL language for robot control. Vic was an interesting guy. He did the first demo of his tabletop robot on my desk (I sent him the film shortly before he died.) During that demo the arm motor failed. He used our machine shop to completely build a motor from scratch and then re-ran the demo.

I was the "technical liason" to SRI (Stanford Research Institute). There Bill Perzley and I learned about Denavit-Hartenberg matrices from Richard "Lou" Paul.

We also did a lot of work on "out of position seam welding" for Kawasaki Heavy Industries. In many cases seam welding is done by rotating the part so it lies flat. Only master welders do things like vertical seams. Kawasaki built ships like oil tankers and they can't be rotated flat.

Joe Engelberger was an interesting guy. He would go to various industries and then come back to Danbury. He kept explaining different things he saw and asked me to figure out how to get the robot to do them. For instance, he saw a company making bricks. The bricks have to be stacked a certain way to cool properly. We talked about ways to get the robot to do the brick stacking problem. One difficulty was the robot used a plated wire memory which was quite small, not large enough to handle the number of bricks in a pallet.