I had been bitten by the bug during the construction of Walter, it seemed, and was now fascinated with the idea of building a free-roaming robot unencumbered by any sort of tether. There was little point in trying to refurbish WALTER; structural damage notwithstanding, all the electrical components were rated for 117 volts AC. My next creation had to be battery powered. And so I began to amass an impressive collection of DC motors, relays, and other diverse components while sorting out the design in my head. The end result was CRAWLER I, intended to be my junior-year science fair project. (This eagerly anticipated event was unfortunately canceled due to faculty indifference.)
I had also decided to build a tracked vehicle for improved maneuverability. Two 24-volt DC gearmotors from an aircraft surplus catalog were mounted on an 18- by 13-inch plywood base, driving left and right tracks fashioned from 1.5-inch rubber timing belts turned inside out. Control was again provided by relays, but the motors each had centrifugal speed-limiting switches that could be adjusted to achieve straight-line travel. By adding an override circuit on the stationary side of the slip rings that fed the centrifugal governor, it was possible to momentarily boost the motor rpm to maximum. Skid steering was achieved by providing differential speed commands in this fashion or stopping one motor altogether. The vehicle could also turn in place by reversing one track.
The tough part in building an autonomous vehicle, of course, lies in how to control its motion, made even tougher still in an era that predated microprocessors and low-cost sensors. I had in mind a platform that would drive around until it encountered an object, then alter course in an intelligent fashion. I also wanted it to automatically recharge the onboard lead-acid motorcycle batteries when they ran low. Like most engineers, I tackled the tougher issue first: automatic recharging. I settled on a beacon homing scheme and elected to use an ordinary light bulb as the source. (It would take me some time, and several follow-on robots, to shake this mind set.)
Providing for truly autonomous operation meant adding some type of collision avoidance sensor and implementing a scheme of intelligent reaction. Tactile sensors made from guitar strings were subsequently installed on the four corners of the platform to support this task. Intelligent response was another matter; single-chip microcontrollers were not yet even a figment of anyone’s imagination in those days. My Hollywood-inspired image of a computer centered around a lot of flashing lights and punched cards. I had already wired dozens of very impressive indicator lamps in parallel with the relay coils of the CRAWLER’s logic and control circuitry (for diagnostic purposes, of course). Operating the CRAWLER with the four-channel radio control developed on WALTER had quickly become boring, so it seemed the appropriate thing to do was build a punched-card reader.
The robot’s environment could be simplistically described by four state variables associated with the tactile sensors situated at each of the four corners of the platform. By comparing these sensor input states to a 4-bit address field punched into each card, the correct response to any particular scenario could be read from the output section of the one card with an address code matching the specified input conditions. The robot would simply stop whenever input conditions changed state and cycle the cards until finding a match. The preprogrammed response (i.e., drive and steering commands) to the new conditions would be punched into the 4-bit output field of the correct card.
I was really excited about the prospect of building this card reader and made pretty fair progress using modified index cards with eight photocells to detect ¼-inch holes made by a standard office hole punch. An actual 3.5- by 8-inch card is shown below; the top row of up to four holes represented the inputs, while the bottom row controlled the outputs. The individual illumination sources for the eight opposing photocells were 3-volt pilot lamps, wired in series to ensure the entire string would extinguish to prevent faulty readings if any single bulb burned out. The lamps were powered by the 12-volt battery at half their rated filament voltage to ensure extended life, and the reduced light output prevented premature activation of the photodetectors through the thin index-card paper. But the mechanics associated with reliably recycling the stack of cards (once all had been read) proved too much for my limited shop facilities, so I resorted to using a 12-inch circular disk of poster paper traced from a 33-rpm record album.
This approach greatly simplified matters. The address and output fields were aligned along the radial axis of the disk with 16 possible states, with the most significant bit towards the outer periphery. The disk would rotate at 6 rpm while the photocells looked for a hole pattern corresponding to the sensor input states. When a match was found, the disk drive motor was disabled and the output field would be read, thus determining the desired control relay states for left and right track drive and direction. The output holes were punched in radial columns offset exactly 78.75 degrees from their associated input columns to allow sufficient room for the two photocell arrays. The circular card was secured to a rubber-covered drive capstan with a ¼-inch wingnut and washer.
All the added features (particularly the 12-inch disk reader) necessitated a complete repackaging of the CRAWLER’s mechanical layout, so I elected to scrap the plywood prototype altogether and build an aluminum chassis. The result was CRAWLER II, basically the same size, but with the electronics implemented in a layered approach.