Robots have a long history from the (not so) simple mechanical creations of antiquity up to today's fully autonomous models. The earliest ones could only carry out scripted actions, modern ones adapt to their environments. Where was this line crossed? What is the earliest example of an intelligent digital robot?

The criteria:

  • Must make decisions
    The system must do more than blindly follow prescripted actions.
  • Digital control system
    The control logic must be fully digital (though not necessarily binary).
  • Must be intelligent
    Must use "advanced" methods of decision-making. This should be something more complex than simply-connected logic gates, lookup tables or cellular automata1.
  • Must be fully autonomous
    Remote controlled vehicles, waldos, fly-by-wire, and the like are excluded. Rough guidance (e.g. "go to such and such waypoint" or "flip over each widget passing by on the assembly line") is permitted but the robot should do more than blindly perform an action. The control logic can be external to the robot (e.g. on a cable-connected minicomputer) but should take only very limited human input.

  • Have at least two sensors
    An array of sensors (e.g. a CCD camera) counts as multiple sensors.

  • Must control at least two independent actuators
    Simply turning something on and off like a thermostat doesn't count. A robot with two independently controlled wheel motors does count.
  • Must be "robot-like"
    It must be capable of moving through its environment in some fashion. This includes autonomous vehicles, industrial robot arms, etc. It does not include thermostats, PLCs on an assembly line, etc.
  • Not a simulator
    It must do something productive: e.g. move through its environment, identify and pick something up, etc. Full-motion flight sims and their ilk are excluded.
  • Not a weapon
    No missiles and the like (because I said so).

I suspect the answer is something not much more interesting than a thermostat but who knows. What was the earliest intelligent digital robot?

1 Cellular automata might be permitted if you can make a good case for it.

To potential down-voters: If you feel this question is e.g. too broad, please explain why in the comments below.

closed as primarily opinion-based by a CVn, Chris Stratton, NobodyNada, DrSheldon, Wilson Nov 27 '18 at 16:08

Many good questions generate some degree of opinion based on expert experience, but answers to this question will tend to be almost entirely based on opinions, rather than facts, references, or specific expertise. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 8
    I'm pretty sure there isn't any intelligent robot around even today. – tofro Nov 18 '18 at 14:32
  • 1
    Except the definition is rather unclear, vague and contradicting. a) A script containing a switch according to some source is still scripted b) what's wrong with analogue c) advanced is marketing blurb d) - e) Vague, is a switch with three setings also an array? f) what makes two actors more inteligent than one? g) vague, after all, even a thermostat moves in its environment h) Come on, a full motion flight simulator is one of the most complex robot systems. i) unneccessary, but works for me :)) – Raffzahn Nov 18 '18 at 23:23
  • 1
    "Meaningful answers or too-broad a question?" Well, yeah, too broad may also be a good description - too road and at the same time too restricted. Also, a LUT doesn't have to be very big to create extreme complex systems. (On a side note, I still like to use LUTs to cover complex logic in a few lines of source and even less asembly instructions. They are an awesome tool to manage incredible complex combinations). – Raffzahn Nov 19 '18 at 0:41
  • 1
    FPGA and TTL are different issues, as FPGA are themself already a collection of LUTs. And While of course any LUT can be replaced by a number of gates, an fairly simple 8x8 LUT can already require a multitude of basic 74xx gates (measured in transistors) for replacements. I think it is even proven for smaler ones than 8x8. LUTs are as well in hard as in software an incredible tool to reduce implementation effort (less RAM or gates and computing/signaling time) – Raffzahn Nov 19 '18 at 1:10
  • 1
    @AlexHajnal may be now but in the past when they start it it got just single actuator that positioned a weight in front of wanted movement and the globe rotate while the weight falls gravitationally towards it ... And it was not a find really as I was called as a consultant during their article review then ... – Spektre Nov 21 '18 at 9:19

I think the first robot to meet your criteria is Shakey the robot, developed between 1966 and 1972 at the Stanford Research Institute:

  • it could make decisions (it included a planner, and could break down complex commands into individual steps on its own);
  • its control logic was digital, running initially on a SDS-940 computer, later replaced by a PDP-10;
  • as mentioned above, it used a planner; it was also an early adopter of route planning, and led to the development of the A* search algorithm;
  • it was autonomous: once given its high-level instructions, it would construct a plan and execute it autonomously, building its own model of its environment;
  • it had multiple sensors, including at least a camera, range finder, and bump sensors;
  • it could navigate, so it must have had at least two independently-controlled motors;
  • it could move through its environment;
  • it was not a simulator;
  • it was not a weapon.

Another interesting pair of robots of similar vintage are Freddy and Freddy II, developed from 1969 to 1976 at Edinburgh University; in particular, Freddy II had an actuator arm which it could use to assemble models from component parts.

  • If the control logic ran on a PDP-10, then I don't assume the PDP-10 moved around with the robot, so wouldn't this exclude Shakey as "not fully autonomous" as required by the question? (Not that I think that this is a very good criterion for "first", after all waiting until the controlling computer is small enough that it can move around is incidental to developing such a robot). Nor sure about the size of the SDS-940, but I don't think the "camera control unit" in the picture is the complete SDS-940. – dirkt Nov 18 '18 at 17:08
  • 2
    @dirkt I took “autonomous” here to mean that the robot didn’t need any human input other than its initial instructions (as opposed to the “remote controlled vehicles, waldos, fly-by-wire” mentioned in the question). Neither Shakey nor Freddy would qualify if they had to be “small” self-sufficient units — they both had controlling computers separate from the robot’s body. – Stephen Kitt Nov 18 '18 at 17:36
  • 2
    @dirkt When I said "fully autonomous" I was referring to decision-making, not mobility. Things like fixed-base robot arms are permitted. – Alex Hajnal Nov 18 '18 at 22:49
  • 1
    @StephenKitt Spot on – Alex Hajnal Nov 18 '18 at 22:59
  • Also, "fully autonomous" could be interpreted to cover a much wider degree...no reliance on external power, able to maintain itself... I guess that would be Tesla's "Atomic Robo" - which is, unfortunately, entirely fictional. (atomic-robo.com) – Klaws Feb 13 at 11:29

The word "intelligent" makes this a philosophical question. Here's one possible answer: scientists mapped out the brain of a worm and in 2014 they recreated it digitally and put it into a fully autonomous robot as part of the OpenWorm project. They didn't even have to teach it or give it any instructions.


Stephen's answer is already a good good one, at least when viewed by today's nomenclature. As usual, there's a grey area before such a clear point, one somewhere between the first, completely analogue versions of Elmer and Elsie or Wiener's Palomilla, and Shakey, the modern style robot.

Some good candidates may be "La Tortue Cybernetique" of 1951 by Paul-Alain or Zemanek's (*1) "Vienna Turtle" of 1954, as both control their machinery with relays, and thus binary decisions. Both (and Wiener as well) are influenced by W. Grey Walter's Elmer and Elsie turtles, but changed the implementation away from completely analogue circuitry.

Long story short, it may be a good idea to lighten up this grey area to find a satisfying answer.

*1 - Zemanek is a quite important figure in computing history, not least through his lead of the Mailüfterl project.

Not the answer you're looking for? Browse other questions tagged or ask your own question.