3D Robotics

3689641499?profile=originalAnother clever legal post from Diana Marina Cooper, a lawyer doing some smart writing on the frontiers of robot regulation. FWIW, Canada seems to have confused "autonomous" with "optionally autonomous", which is actually what most modern drones are (they have both fly-by-wire and autonomous modes). From RoboHub:

Chappie, the new robo-film on the block, takes place in Johannesburg, where the police force is made up of robots. In one of the early scenes, the main characters, Die Antwoord’s Ninja and Yolandi take part in a drug deal that is raided by robot cops. Hoping to avoid a similar fate in their next deal, Yolandi suggests that they find the robots’ remote so that they can switch them off like TV sets. Do the robots have a remote? And do Ninja and Yolandi find it? No spoilers here, but let’s take up the underlying question in the context of drone regulations…

What is Canada’s position on autonomous and automated operations? Does your drone have a ‘remote’? If it does, according to Transport Canada, it is not an autonomous drone. Transport Canada defines an autonomous drone as one that does “not allow pilot intervention in the management of the flight.” It is not enough for an autonomous drone to be capable of self-governance, rather it must not allow for any possibility of human intervention.

What about drones that have a remote but can complete automated tasks such as take-offs or landings or that can execute pre-defined waypoint operations? Transport Canada distinguishes these drones from autonomous drones by pointing to the fact that they require operator initiation or intervention.

Although there is no express prohibition, truly autonomous operations are outside of the scope of Canada’s current regulations. For the time being, if your drone doesn’t have a remote, you wouldn’t be able to operate. Our framework permits operations that involve automation, but it requires that an operator have the capability to intervene – or rather, as Ninja and Yolandi would hope – your flying robot must have a remote.

E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones

Comments

  • Thx Doug  for the links.

    But only the previous link to wiki worked and the other link didn't get the Clough pdf.

    As mentioned in my last post I don't have an issue with the fact TC want to have remote and I can't see a reason why anyone would want a non-RC UAV, at least not intentionally. So I'm happy to let TC do whatever they want...provided they have a RC! ;-)

    The discussion was more about the definition of what is/isn't autonomous. Then intelligence, to which I hope i offered some definition in the context of UAVs/Skynet. But I think we have covered that subject in enough detail now. Thanks for your input.

    Regards

    JB

  • Moderator
    Hi RD
    I agree with you complete, it's also something that I can and will live with as I am in Canada. If that's the rule, smile and follow the rule
  • Semantics and interpretations aside (definitions, though in print, do change in time) - the focus of the article I linked, by Bruce T. Clough, was not on what a thinking machine/robot/UAV is or is not but, on how do we measure autonomy?

    Give it a read and you will notice that the epitome of all measures ends up at humans and our ability to know, usually, when information is erroneous or misleading. I say usually because after reading a book on airliner disasters prior to 1950 (link), even with humans in control, things can go very wrong.

    TC has stated what non-autonomous is - if it has a remote control - and that is what they will measure by. They want the added safety factor, I suspect, that a remote control provides. Clough also mentions the work of Sheridan. The concepts he has addressed fit in many areas of our discussion regardless of the policy decisions by government agencies.

    -=Doug

  • lol Dwgsparky!

    I didn't mean to imply that you killed anything! It was meant to say "skills". :-)

    I can agree to the fact that people use the term in such a way, but it's hard to communicate any idea if we abandon previously agreed definition. I can also identify with your analogy with children as described further below, however TC should have specified it as "without any user control whilst in flight" instead of autonomous.

    On the definition of decision: a conclusion that results from consideration; further consideration means: thought, meditation or deliberation; which are all human traits, from something that is living.

    Let me expand on your child analogy: If you tell a child to not cross the street the child still has the power to decide, then based on that decision, act if it will or will not cross the street. There is no guarantee that the "program" will work as intended. If on the other hand you tell a drone not to cross the street, then it a) it's system is fully functional and it does not cross the road according to it's programming or b) it's system is faulty (GPS loss) and it crosses the road, not because it wants to, but because it can't tell if it did. The UAV could not however say "stuff the programming" and do as it pleased as the child could and see if there wasn't a reason to why the programmer told it couldn't. It can only do something different than expected because something failed to work as it should. Coincidentally that's when we start calling it names! (yes I made that joke three times now - it's getting old!) :-)

    The terminology gets even more muddy when we start saying the "UAV is flying itself" and  "she's an awesome looking machine" etc. By personifying it we are setting our own language trap. The truth is that the distinction, even in the word autonomy which has it as it's root, is the definition of self.

    The dictionary defines self as:

    n. pl. selves (sĕlvz)
    1. The total, essential, or particular being of a person; the individual: "An actor's instrument is the self" (Joan Juliet Buck).
    2. The essential qualities distinguishing one person from another; individuality: "He would walk a little first along the southern walls, shed his European self, fully enter this world"(Howard Kaplan).
    3. One's consciousness of one's own being or identity; the ego: "For some of us, the self's natural doubts are given in mesmerizing amplification by way of critics' negativeassessments of our writing" (Joyce Carol Oates).
    4. One's own interests, welfare, or advantage: thinking of self alone.
    Further interesting reading on self: http://en.wikipedia.org/wiki/Self

    So unless the UAV gets a "self" it also can't be autonomous. 

    BTW on the subject of RC less UAV's. I can't think of a situation where this would be desirable, unless it's a "one way trip" like a missile or something. Some sort of control and feedback to a human user should always be available.

  • Moderator

    Hi JB 

    Normally I try NOT to kill people with automation!! but I'm not perfect!!

    I usually take the view that language can be difficult and that the best result is obtained by applying the concept of "What is reasonable, practical and understood." a concept used in many safety and law discussions and trials. 

    The difference between automatic (programmed in advance) and autonomous( makes its own decisions) is a good example but first you need to define what is a a decision.? If a machine or UAV has only one purpose in life and responds to various outside influences or sensors then it is automatic but it can also be autonomous as it does not need you or me to perform its task.it is taking its own decision based on what someone has told it to do in certain circumstances.

    This is no different to you telling a child not to cross a road if there is a car coming (you are pre-programming the child). The kid can still make the wrong decision just as the UAV can make the wrong decision sometimes because it trusts one sensor more that the other. A child may not see a truck but he might hear it. its still there and will still kill him. an autopilot may read the GPS and determine that the craft is falling, The altimeter may disagree and decide the UAV is level, if it trusts the altimeter and it is falling it will crash. it has made its own decision and so it is generally refered to as autonomous. 

    TC are trying to use this wording to separate the different type of UAV from a RISK perception. they are not trying to define words in the dictionary. The majority of normal, intelligent persons understand the words in the correct context as an autonomous UAV is one with an autopilot but NO RC. so the human is not in control and cannot intervene if it is required. it does not matter whether the failure to intervene is because the RC equipment is not there or the flying skills are not there. If they cannot take effective manual control of the vehicle then it is automomous because I can turn an automatic system OFF.  

    Apache2 Ubuntu Default Page: It works
  • Dwgsparky

    I do not doubt your automation kills or your experience in using them. The problem is more so the definition of the term "autonomous" not only AI (artificial intelligence). In the context of the discussion here about UAV's, or the Canadian position about the use of drones with RC but not those that are "autonomous", it would seem that they too do not understand the difference between automatic or auto, and the actual definition of autonomous.

    If they would have looked the word up in their dictionary they wouldn't be making such baffling statements, because technically "autonomous" UAVs or drones simply don't exist, and it's disputable if they every will, going by the actual definition of the word. But UAV jargon and shop talk have watered down the meaning of the term and it is now understood as meaning automatic.

    In essence it's the difference of us telling it what to do, which should be defined as "automatic", even if that occurs in advance through programming, or it having the ability to decide for "itself" without a programmer ever telling it to do such a thing, which would then be "autonomous". Hence my attempt in humour to explain how we sometimes name our UAVs all sorts of things when they don't do what we want - ie they do not do what we programmed them to do and like kids sometimes are therefore showing traits of rebellion! ;-)

    -

    R.D.Starwalt

    OK now we want to talk about what intelligence is and is not? (not on subject but interesting nonetheless) ;-)

    Not all things that are "smart" are intelligent (like my smart watch isn't!). Once again we're using the terms loosely, but that's the nature of language. At times there are concepts that are only understood by reading between the lines. At times this is due to either the lack of knowledge in what terms to use to describe it, or simply because the idea is so new or "un-describable" it has no words yet! Love is a classic example.

    The dictionary defines intelligence as:

    1. capacity for learning, reasoning, understanding, and similar forms of mental activity; aptitude in grasping truths, relationships, facts,meanings, etc.
    2. manifestation of a high mental capacity:He writes with intelligence and wit.
    3. the faculty of understanding.
    4. knowledge of an event, circumstance, etc., received or imparted;news; information.
    5. the gathering or distribution of information, especially secret information.

    In the case of the popular idea introduced by the Terminator movies, "Skynet" is somewhat real in the form of the internet. The internet ticks many of the intelligence boxes as well, provided it is maintained and influenced by man. Things going "viral" support the idea that the internet is more than just a machine; it's alive by the amount of "life" we give it, even if it is just all that time our wives spend on facebook! 

    My reference to "Skynet" was related to the idea that the internet was intelligent (but not quite so bent on kiiling humanity -  I hope!). The reason I say this is that it's more of a hybrid beast where both man and machine can transfer and process information, learn, understand, form relationships between the inanimate and animate etc. Unlike a UAV the difference here is that the internet, only with human input however, can demonstrate the same complex behavior as a "autonomous group thing", the borders of which can be very blurred at times. But maybe that conclusion is wrong, and it only looks that way because we are confused by it's complexity!

    As a consequence the answer is yes; intelligence is required to make UAVs "smart", but it would be sourced from various human and non-human entities that "co-exist" on the web aka Skynet. UAV's just need to get connected.

    Regards

    JB

  • Moderator

    Lots of reading there, reply later

  • Good distinction regarding AI but, would not some level of intelligence be required and isn't that the Skynet reference?

    THIS article discusses UAVs in particular (the link should pull down the pdf).

    A quote on Wiki from the above article (scroll down to the Robotics section)...

    Automatic means that a system will do exactly as programmed, it has no choice. Autonomous means that a system has a choice to make free of outside influence, i.e., an autonomous system has free will.

    An automatic system could act autonomous until the assigned task is complete or not complete. It would then give an accounting of why it did what it did.

    When I make a threatening sound or gesture to one of my cats, they decide (based on instinct and familiarity with me) if they will ignore me or run away.

    When I start cursing my multirotor, it is unaware I am upset with it's behavior. It has no audio detection circuitry and no way of processing the signal if it did...at this time.

    At a minimum this discussion needs boundaries, specifications, and agreed upon definitions.

    Rules, rules, rules...

    -=Doug

  • Moderator

    Hi JB 

    My profession is as a automation engineer and I have been here long enough to have seen the moon landings and everything since then so I understand control systems very well. but your description of autonomy is really defined as Artificial Intelligence. AI. There are many machines in our everyday life that are autonomous ( engine management systems, Home heating systems, industrial production machines, car engine assembly lines. they are ALL autonomous but they do not have AI.. 

    The difference is that ultimately the human has control, he can switch it off or in our case pull the battery plug. If you look at what a drone with AI would do compared with an autonomous drone it is easy to see the difference. 

    A drone with AI could receive a call from a customer and then decide what route to take to survey the project,fly the route and return to base and process the images. I am not part of the process.  

    An autonomous drone can do the same job BUT I decide when to connect the battery , what route to take and when to land. I do the processing. I SUPERVISE it and have the ultimate control. If I see a major problem I can decide to dump it in the river. I have that responsibility. It is still an autonomous vehicle.

    As a Canadian I believe that this is what TC are trying to achieve. Freedom to fly but with a operator in final control with a RC unit even if it is a back up to the flight controller. 

    Truthfully I think that they are doing a good job, at least they are doing something. Do a google search for anything remotely UAV and you see a link to Transport Canada on the first line, explaining the rules and the dangers. Great. 

    The BIG issue is the number of people who have no knowledge, no insurance, no skills, no training and no awareness of the dangers of operating these vehicles. They need educating and training before someone gets killed by a UAV falling form the sky. 

  • As I have said for over 15 years...  IT ≠ IQ

    Our tech isn't even a smart as a cat - yet - and probably will not be for quite a while.

    A cat's ability to 'sense and avoid' is amazing and part of their basic programming - instinct.

    Our goal with this tech should/will be to make safety part of it's instinct; a part of the ROS.

    Asimov's laws are a good place to start; ingraining that as part of the code would help the cause.

    -=Doug

This reply was deleted.