1953 – Timothy Turtle – Jack Kubanoff (American)
1953 – Timothy Turtle by Jack H. Kubanoff, Radio and Television News, April 1953
The text to the article is below. For the moment, the pages themselves can be found here.
The article itself has a good description of its functionality. Some of its functionality reminds me of Berkeley's Squee i.e. flashing lamps controlling the mode of operation.
A ROBOT ELECTRONIC TURTLE
By JACK H. KUBANOFF
Mechanical and electrical construction data on a novel toy for the children—which will provide hours of fun.
"TIMOTHY" is an electronic turtle with an ability to perform in a manner comparable to his living counterpart. He can see; he can interpret information; he can follow instructions; he can think; search for food, memorize, and is able to learn. He may be relied upon to retain what he learns for an indefinite period of time.
"Tim" was originally constructed as a toy for my daughter but finally emerged as a full-fledged problem in cybernetics (a study of the control circuitry of the human brain and nervous system as applied to electronic machines. These machines, more commonly known as computers, are devices which are rapidly replacing the human element required in the operation of various mechanical devices.)
In order to obtain scientific information, simplification, wherever possible, always yields results which are both tangible and upon which predictions can be made. "Timothy" is a light sensitive animal and understands only such information as can be transmitted to him by light. This handicap in limiting his scope of operation is overshadowed by the simplicity in circuitry that it affords.
His fundamental operation was that of following a light wherever it was moved about the laboratory. For this purpose, a form of light discriminator was incorporated which will be fully described later. To this function was added the search for a form of sustenance. This gave "Timothy" a real purpose in life and elevated his status is that of a computing machine. At present, "Tim" is quite simple in circuitry but his ability to perform complex functions is remarkable. "Tim" can duplicate many of the functions of the human brain and nervous system and he reacts in much the same manner as his animal equivalent when he is confused or fatigued.
In their proper sequence, "Timothy" will do the following:
1. He will search for food (initially he is hungry). In this mission, he will determine the light word (in this case, a light flashing at a certain rate) for food and follow it to his nest. If any other information is transmitted to him, such as instructions for playing, etc., he will turn away from that source of information and continue in the pursuit of sustenance. Changing the word for food (by changing the flashing rate and duration) only causes some slight hesitation as the repetition of the word provokes "Timothy" into learning the new word for food. The removal of any light he has been memorizing for a period of time will cause him to rely upon his memory for the position of the light. Upon arrival at his nest, he will obtain sufficient sustenance for about one-half hour of playing time.
2. "Timothy" will now follow a light for play anywhere he may he required to travel, using his memory to remember the position of the light if it is hidden from him and disregarding words which indicate food sources. The absence of the "play" light will cause him to stop until he again intercepts the words for play.
Thus "Timothy" can work or play but his most unique characteristic is his ability to learn and make vital decisions concerning his own welfare. "Timothy" has two memories, a vital memory in which he learns a new word and commits it to a memory that retains it indefinitely and a statistical memory, less intense, and through which he retains information regarding the position from which he last received information. Both require repetition and differ only in the length of time they are retained for reference. These memories have their counterparts in real life—the statistical memory, in which we obtain information and store it for future reference but whose retentivity deteriorates from failure to use the information, and the longer vital memory which is sympathetic in nature and is retained for the duration of our lives. It might be interesting to note too that "Timothy's" memories like their human counterparts are not instantaneous in action, but require constant repetition before the information is completely stored. In the same way, words for "Timothy" must be repeated several times before he can commit them fully to memory. "Timothy's" vital memory may be compared to the word "food," being repeated in various dialects. At first, he will find the new dialect strange. but because it contains some similarity to the other words for food he has learned, he will learn the new word and from that time he will respond to the new dialect. "Timothy" can retain in his vital memory only one word at a time. As a result all previous words are rejected in preference to the newest word he has learned. He obviously can become
Fig. 1. Complete circuit diagram covering the control circuit for the robot turtle.
Fig. 3. Construction details for fabricating the turtle's head from a coffee can.
Fig. 2. Front wheel modification of wagon.
confused by changing the dialect more rapidly than he is able to commit it to memory. At this point, of course "Timothy" has the symptoms of a very intense neurosis aggravated by his desire for food and his inability to read the signs which indicate the direction he must go to obtain food. "Timothy's" ability to learn is closely akin to what psychologists term a "conditioned reflex." This can be explained simply.
When two events occur simultaneously over a period of time, it is normal for us to associate the two events with each other, i.e., noise-crowds, heat-fire, etc. Thus, we know by the noise that a crowd is present or by the heat that a fire is present. It is also possible to synthetically produce these sensations. As an example: ringing a bell before dinner is served. After sufficient repetition of bell ringing before dinner, ringing of the bell causes the digestive juices to flow copiously even when dinner is not a consequence. This method of learning is time consuming and as a result "Timothy" can be hopelessly confused. When we can make the word for play appear to him like the word for food, he becomes frustrated; as a result, his actions in frustration can be predicted and given a mathematical significance. We can tell in this manner exactly how "Timothy" will react to various stimuli.
One of "Timothy's" outstanding features is the versatility of his construction. He was constructed of parts that can be found in junk boxes, from old toys, or from the odd parts available in an experimenter's workroom.
"Timothy's" body is a metal cart that was purchased for our first youngster and set aside for the next. It is of the variety that sells for about one dollar. The handle was discarded and the front cut away to within 1 inch of the front wheel bolts. This was done with a metal shears so that it appeared as shown in Figs. 4A and 4B.
The front wheels were removed keeping the original lock nut, bolts, and washers. The tynes for supporting the handle were cut away as shown in Fig. 2B. A hole 6/32" in diameter was drilled about 1/4" from the front edge and the resulting tongue-shaped piece of metal was bent back in a radius so that the hole in the tongue was directly over the hole through which the bolt must pass to secure the front wheels to the cart. A small flexible wire was then soldered to the cart body near the head and permitted to make contact with the floor. The cart body will be a common electrical ground throughout the construction and this serves as a return for "Timothy's" tongue!
Construction of "Timothy's" head requires the following parts: two lenses 1" focal length; two RCA Type 927 photoelectric cells; a 1 foot length of Belden 8423 three-conductor shielded cable; and one empty coffee can.
1. Scribe two lines through the center of the coffee can bottom perpendicular to each other.
2. Select one of these lines and mark off 1/4" on either side of the line.
3. On either side of the remaining line, mark off 5/16".
4. Return to the line marked off as in 2 above and using a can opener, cut from each quarter inch mark about the circumference of the can to the same mark on the other side of the can as shown in Fig. 3A. The ends of the 1/2" strip remain secured to the sides of the can.
5. Cut along the 5/16" marks on either side of the second line; this piece will then come out. Cut 1/8" strips along each long edge of this strip. Then, cut the remaining 3/8" strip in half along its short dimension.
6. Pull the remaining flaps forward so that the inside of the can is more accessible.
Top chassis view of robot with the turtle's "shell" removed to show construction.
7. Solder to the ends of the 3/8" strips. the 1/8" strips and mount the phototubes as indicated in diagram, Fig. 3D.
Drill the openings for the lenses and cement them into place in the center of the flaps still attached to the can. Bend the can into an oval shape as shown in Fig. 3E, and bend the phototube flaps so that the edges fit within the oval can 1 3/8" below the lid flaps. Solder wires to the phototube terminals as indicated in Fig. 3G and set the phototube support and eye flaps at a thirty degree angle as indicated in Fig. 3F.
Pieces of tin from another can may be used to reinforce these structures in order to make them hold their relative positions in the next step.
Mix a preparation consisting of one cup of flour and one bottle of mucilage to 1 1/2 quarts of water. Slit newspapers into 1 1/2 inch wide strips and pass them through the well-mixed preparation. Now, lay them about the head of the turtle, covering the can completely except for leaving the back open. By varying the size of these strips and alternating the direction of each strip so that they interlock as they dry, a head for "Timothy" is easily shaped. The final paper mache head should be put away to dry for at least twenty-four hours. This mixture should be set aside for "Timothy's" shell.
When the head is completely dry, drill a hole at the back for mounting on the support indicated in Fig. 2C. Before attaching the head to the cart body, you may glue a small piece of metal screen to the face and drill a hole through the head for the wire of the tactile tongue. Solder a wire to 'Timothy's" tongue, passing it through the head and alongside the cable. "Timothy's" tongue must he insulated from the body! Attach this wire to the grid of V3, Fig. 1. The head may next he painted any suitable color. The eyes will function best if the area immediately surrounding them is painted white.
For the electrical construction, type 12AU7 or subminiature 6BF7 vacuum tubes may be used. All diodes are germanium 1N34's or 1N52's. All tube biases are returned to the negative terminal of the heater supply. The positive terminal of the heater supply is grounded.
The grids of V1 and V2 are connected the clapper of relay RL4. The phototubes are connected to the normally – closed contacts of RL4. The cathodes of V4, V5, V6, V7, and V8 are connected to the normally-open terminals of RL4. Remember that this relay is made up of two relays! One a double-pole, double-throw and the other a single-pole, single-throw relay. The grids are applied to the d.p.d.t. relay and the cathodes to the s.p.s.t. relay. The resistors bridging CR? and CR, (germanium diodes) are 1 megohm, 1/2 watt units.
"Timothy's" sight circuits are shown in Fig. 5A. The two photocells shown are located in "Timothy's" head in such a manner that either of his eyes may intercept light signals. The original system was that of balancing light in each eye, but a more positive and reliable system is the one shown in Fig. 5B. Here, three relays are utilized. RL1 is the "right" relay, RL2 the "left" relay, and RL3 the motor relay. RL3 is connected in series with the power supply so that when a phototube draws current and causes V1 and/or V2 to conduct, the current passing through either RL1 or RL2 must also pass through RL3. Thus the interception of light immediately causes "Timothy" to move forward. Note that the arrangement of the relays is such that RL, and RI, in combination operate the reversible motor for right and left or neutral. Equal light in both eyes will energize both relays and "Timothy" will turn in neither direction but will move forward. The arrangement of the photocells is such that light can be perceived in both eyes for only a very small angle and, as a result, the operation of this system is one that is primarily binary in nature. As "Timothy" approaches a light source, and greater accuracy in his searching is necessary, the images on his photocell
Fig. 4. Wagon and rear wheel modifications.
retinas travel toward the outer edges of the conductive surface where the slightest movement of "Timothy's" head might cause complete loss of the signal on one side, making necessary an adjustment in his sight position to a very close tolerance. Short focal-length lenses have been used to increase the light intensity in each eye as well as to present a clear-cut image on his retina which defines his exact position with respect to the light source in each eye.
The photocells are mounted as indicated in Fig. 7 with their long axes horizontal and fixed in such a position that the light reaching each photocell from straight ahead barely reaches the edges of the cell on close light sources. This binary system ("off-
Fig. 5. (A) Fundamental sight circuit and (B) motor circuits for the robot turtle.
Fig. 6. (A) Electrical circuit for "Timothy's" nest. (B) Mechanical details for constructing nest. (C) Alternate circuit for providing the necessary — 6 volts bias when the motors operating "Timothy" are a.c. driven.
Fig. 7, Maximum and minimum interception range of "Timothy's" phototube "eyes."
on") in each eye, obviously increases "Timothy's" ability to direct his motion more accurately and far more simply than a discriminator action in the normal form since this discriminator action is in the form of a curve with sharp discontinuities in each eye near the point where either photocell conducts.
"Timothy" has two main functions —eat and play. V3 of Fig. 1 controls these functions. C1 charges to about —110 v. through "Timothy's" tactile tongue when it contacts the charged wire in the nest. C1 then discharges slowly through R2 and R4, this time of discharge being determined by the setting of R4. This discharge period is the "play" period. When C1 reaches a potential of —2 volts, V3 conducts and RL4 switches "Timothy" from "play" to "search" for his food. RL4 operates the ground return for the cathodes of V4, V5, V6, V7, and V8. All tubes used in these circuits are dual triodes. All the diodes used are germanium semi-conductors. A pulsing light transmits "food" information to "Timothy." The duration of the light's off and on periods are variable.
During the function when "Timothy" searches for food, his circuitry changes in such a manner that he is most sensitive to pulsating light. Thus, V1 and V2 are capacitively coupled by the action of RL4 which opens the d.c. path in "Timothy's" light-sensitive nervous system. V5 and V6 also add to "Timothy's" preference for obtaining food. If the duration of the light is longer than "Timothy" last remembered it to be, V5 and V6 will turn him away from the light source. When V4 first conducts "Timothy" travels to the left, describing a complete circle, until he intercepts a light source. He will analyze the light in V4, V5, and V6 in the following manner:
Negative pulses appearing at the plates of V1 and V2 are applied through C2 or C3 or both to diode CR1. This charges condenser C4 which causes V4 to cease conducting, whereupon "Timothy" ceases his searching and will travel toward the pulsing light source, continually testing its duration. If the light is picked up by his right eye, V1 receives a pulse through C5 and the same pulse will be applied to the grid of V5. If the duration of the pulse is long enough to raise the voltage across the integration network consisting of R6 and C8 to the point where V5 will conduct, then V5 will cause "Timothy" to veer to the left and away from the light source. Obviously then, if a steady beam is suddenly intercepted by "Timothy," he will turn away from it. Now, R5 and R6 are adjusted by a small reversible motor which is operated by V7 and V8 through the double-coil relay RL5, RL6. This relay is so constructed that if the current through both coils is the same, the contacts remain open. If the current through one coil is greater by a sufficient amount than that through the other, the clappers will move in unison, making contacts as shown. Note that the grid of V7 is connected to V5, and both have a common cut-off bias. If the duration of the positive pulse is greater than the setting of V5, V7 will conduct, increasing R5 and R6. This increases the time necessary for charging C7 and C8 to the point where V5 or V6 will again conduct. Meanwhile, if the pulses are of shorter duration, the pulsing of V8 causes the motor memory to decrease the value of R5 and R6 thus making the time necessary to cause V5 to conduct shorter. When the duration is the same, the amount that V8 reduces resistors R5 and R6 is the same amount as V7 increases them! C13 and C14 are optional memory circuits. When signals appear at the plates of V1 and V2, CR2 conducts and charges C13 or CR3 conducts and charges C14. R13 and R14 are current-limiting resistors which permit the necessary current to discharge C13 or C14 and close their respective relays. This system affords a quick method of "Timothy's" learning and a fairly long retentive period or memory. C11 and C12 associated with V7 and V8 perform in essentially the same way.
Thus "Timothy" actually measures light duration in a manner which will allow him to retain the information without power consumption. This indicates that he remembers the duration of the light last transmitted to him even though he is activated on rare occasions with long periods of deactivation. These deactivation periods have no limit in time! He can follow the light indicating a food source even though it be modified many times between the first interception and the final arrival at the nest. "Timothy's" condenser is charged, when he arrives at his nest, by a tactile tongue which he extends in search for food. At this point he again switches from pulse light sensitivity to continuous light and is ready to follow a continuous beam of light for the duration of the charge in his "condenser stomach."
(Continued on page 150)
Under chassis view of the wagon used as the body of the electronic turtle.
"Timothy" – (Continued from page 38)
The motors can be powered either from a 6 volt battery mounted inside the body, or from the 6.3 volt secondary winding of a filament transformer. In the latter case, an extension cord must be provided which is long enough to reach the turtle from the nearest outlet. This transformer can be seen in the photographs, mounted on the upper platform. When the transformer is used, a bias supply capable of providing a negative 6 volts must be provided. The circuit of Fig. 6C shows how this can be easily accomplished with a minimum of components.
The "B+" voltage is provided by a single 67 1/2 volt "B" battery, mounted inside the turtle's body.
Motor and Control Circuits
There are two motors used in "Timothy's" mechanical motion circuits. These are a power drive motor and a turning motor. They are of the type described in drawing number 7100 of the Hansen Mfg. Co., Princeton, Indiana. It is a six volt, 1/2000 hp., 5400 rpm, 16.2 ft. lbs./minute unit. With the gear-reduction unit, the speed is approximately 18 rpm. This is a type C gear-reduction unit. The turn motor is fashioned as follows; a piece of aluminum is fabricated according to Fig. 8 and is mounted to fit the drive shaft of the motor. Again the dimensions are not critical except for the lever arm which is 3". A set screw (6/32) is used in the shank to secure the lever to the gear box of the six-volt motor. The lever is then secured by a 6/32 screw and bolt to a torque arm which extends from the point at which the motor is mounted to the hole in the front wheel assembly where it is attached by a second 6/32 bolt and lock washer, as indicated in Figs. 2A and 2B and Fig.
Fig. 8. Mechanical details of the turn, memory, and drive motors and associated parts.
Fig. 9. "Timothy's" operation includes a search for food (dotted path) and typical play path (shown in solid line on diagram).
8. Limit switches ("switchettes") are optional and are of the normally-closed variety — single-pole, single-throw.
They are connected in series with the ground return from each relay contact as shown in Fig. 8C. They are mounted so that the lever may traverse forty-five degrees from center of the neutral position.
This motor is mounted on a 2" x 2" bracket, three of which are required. One will be used for the propulsion motor and the other two in the memory circuits. This bracket will be secured to the underside of the chassis (cart body). A grooved drive wheel to carry a rubber or spring-drive belt is connected to the gear reduction shaft and a second wheel of similar dimensions is mounted on the rear axle. The drive wheel used in the original turtle is 1 3/4" in diameter with a sleeve 3/4" in diameter and 6/32 lock screws. The wheel is 1/4" wide with a 1/16" radius grooved into the circumference to accept the drive belt. The drive belt is of the type used in motion picture projector's or radio dials and may be purchased from any reliable drive belt manufacturer. The belt used in "Timothy" was manufactured by S. S. White Dental Mfg. Co., 10 E. 40th St., New York, N. Y. The third or memory motor is mounted on the second bracket and is coupled to the dual potentiometer R5-R6 as indicated in Fig. 8E. It is mounted on the third bracket fabricated as mentioned before.
"Timothy" will retire to his nest to fill his condenser stomach and he does so through the use of a #1 photoflood lamp. This lamp is flashed off and on at variable intervals. The operation of the circuit is as follows. A selenium rectifier (100 ma.) charges a condenser through the solenoid of a relay. The relay closes due to this current and remains closed until the condenser reaches a potential where less than 1 milliampere will flow. At this point, the condenser is switched across the solenoid and discharges through the solenoid to a point where again the difference in potential across the solenoid terminals will permit less than one milliampere to flow at which time the relay again opens and repeats the cycle. The cycle may be varied by potentiometers R2 and R4, which change the charging and discharging time for the condenser. The circuit is shown in Fig. 6A. The relay used in the nest circuit is the same type as RL1 in Fig. 1. "Timothy's" nest is of simple construction. A 1' x 1' fiberboard roof is supported by 1" x 1" x 10" wooden pillars. A cigar box above the roof houses all the electronic equipment associated with the nest. A wire is secured by staples to the pillars in such a manner as to be just high enough above the floor to contact "Timothy's" tongue. The other end of this wire is attached to the rectifier supplying the switching action for the light as shown in Fig. 6B. A fiberboard base 3' x 3' is coated with aluminum foil and the other side of the a.c. line is attached to this floor through a 500,000 ohm resistor. Thus, as "Timothy" enters his nest, the ground wire of his body makes contact with the floor of the nest and finally the tactile tongue touching the suspended wire fills his condenser stomach to the brim. A 500,000 ohm, 1/2 watt resistor should be connected in series with the rectifier and charging wire to avoid a serious electrical shock. It is advisable that the nest he carefully guarded from the children as the exposed terminals might shock them even slightly and this is an undesirable condition. For daily play, "Timothy's" work-play control tube is disabled (V3) and he will follow the children about the room or outdoors without requiring a nest or experiencing a single pang of hunger!
"Timothy's" shell is a shaped chicken wire form and paper maché covering which is painted and secured to "Timothy" by means of a central supporting pillar of wood or metal. Actually, "Timothy's" support for his shell is a discarded i.f. can. This offers a broad shell support while the two screws which normally secure the can to a metal chassis offer the requisite rigidity. The layout of "Timothy's" circuits is optional and are constructed as simply as possible. The vacuum tubes as well as the turn relays in the turtle were mounted in the cart body. The other functional relays were mounted on the platform above the cart body. The platform consists of 1/4" plywood fashioned to the perimeter contour of the cart body.
"Timothy" is an initial experiment in self-contained control systems. He can be improved upon by increasing his powers of perception. By allowing him to have variations of tactile sense, his ability to recognize and remember shapes, temperatures and even taste is a distinct area of improvement. These senses will open a completely new system of closed control. The interaction and combination of these controls in a system as simple as that indicated in "Timothy's" original control circuitry makes a self-controlled electronic robot a distinct possibility and an intriguing problem in cybernetics.
Acknowledgement is gratefully made of the help of Edward Jaczack and Fred Tulley for their frequent consultations on the mechanical devices incorporated in "Timothy."
See other early Cybernetic Creatures and Models here.