Motivated autonomous robot

[shurman-a]

Hallo!

I would like to inform you about my project of a motivated autonomous robot

1. A note on existing robots

The existing robots are neither intelligent enough nor autonomous, although many specialists say these robots are autonomous. Actually, they are only partly autonomous, because they perform only some sequences of motion and/or observations autonomously after receiving an order. In general, the activities of these robots are steered and controlled by people, and these robots do not react independently to situations in their surroundings. Known existing robots cannot judge which objects, situations and activities are important at present. Without the last property, robots can be neither intelligent nor really autonomous.

2. Project of a motivated full-autonomous robot

I developed the project of a motivated robot (denoted by aS) which has the following wished properties:

i. it works full-autonomous,
ii. it is high motivated to do his work,
iii. it knows how important is the execution of an activity at present,
iv. it can judge which objects, situations and activities are important at present,
v. it handles urgent and dangerous situations with appropriate high priority,
vi. it interrupts the current executed activity, when the execution of an other activity is more important at present,
vii. it can handle against its norm or rule, when this enables to handle against a dangerous situation and if this norm or rule is not so important,
viii. its activities may be very complex, as required for full autonomous work.

Full-autonomous robot means, that it itself decides, by means of its motivations and perceived situations, which activity it executes; of course, the robot performs a command, if its priority is high enough.
Said norms and rules(vii) protect additionally that the robot aS do not get up to mischief and that it behaves friendly towards a human.

The project of this autonomous robot aS is published by:

* WIPO (World Intellectual Property Organisation), in German: Determination und Kontrolle der Aktivitäten eines emotionalen Systems; Internat. Patent Application: PCT/DE2002/004172; Internat. Publication: WO 2004/044837 A1 , 27 May 2004,
* US Patent and Trademark Office, in English: Determination and control of activities of an emotional system, U.S.A. Patent No.: US 7,024,277 B2, 4 April 2006,
* UK Patent Office: UK Patent Application: GB 2409734 A; printed as Application No.: GB 0305707.2 , 06. 07. 2005.

The motivated full-autonomous robot aS contains:

* a set of needs of aS, e.g. the need to keep order in a room; such need is a domain in which the robot is active;
* rules and procedures which determine the current emotion intensities with respect to said needs;
* a set of object and situation models representing the surrounding of aS;
* a set of complex activity schemas, which the robot can execute and which use said situation models;
* a subsystem for recognition/perception of objects and situations;
* the steering and controlling subsystem which determines and controls the execution of said complex activities;
* the subsystem which executes and controls elementary motion and observation activities, which are used in said complex activity schemas.

The existing robots consist mainly of said subsystem for execution of elementary motion and observation activities.

3. What is understood by emotions and motivations of a robot?

Emotions and motivations of a robot are actually pseudo-emotions and pseudo-motivations, and they function different than human emotions and motivations. The emotions of a human are closely associated with the limbic system. The robot aS does not simulate the limbic system. Emotion intensities and motivations of a robot are calculated by simple procedures and rules.

Pseudo-emotions of the robot aS are related to emotions of a human as wings of an aeroplane to wings of a bird - these wings use the same principle of flying but they are very different.
More about the emotions of the robot      http://mywebpage.netscape.com/alschurm1

4. Does we need motivated full autonomous robots?

Robot aS considered in my invention can be used everywhere where autonomous routine work should be done; for example:
* taking care of ill or old people,
* making housemaid work,
* making routine service for clients,
* as personal assistant,
* making autonomous activities in a manufacturer, in a desert or on the moon,
* supervise a staff of robots (working e.g. in a manufacturer).

A robot making housemaid work is certainly a better solution than the so called "intelligent apartment" in which the refrigerator says which food should be bought in and the radiators guess which temperature the inhabitants would like to have. Such "intelligent apartment" (computer) can have only relative small number of internal models of objects and situations, and can execute only few activities.
The motivated full autonomous robot for housemaid works has internal models of all object and situations in the apartment, operates the devices and machines (e.g. washing machine), gets the meal, takes the dog for a walk (the silly real emotional full autonomous system dog will treat the robot as having higher status) and, in the future, goes shopping. Such a robot is indeed more comfortable, more useful and cheaper than the "intelligent apartment".
The existing robots are only the first step in manufacturing high-quality products. Full autonomous motivated robots will manufacture products in great part independent by operating and controlling other robots and machines, night and day.
More about industry robots    http://mywebpage.netscape.com/alschurm1/page5.html

Motivated full autonomous robots would be able to learn and would handle intelligently, if appropriate programs are implemented. A robot cannot actually handle intelligently without pseudo-emotions and pseudo-motivations.

5. Can the motivated full autonomous robot be made now?

The development of humanoid robots and robots playing football is so advanced, that it is only a stone`s throw to motivated full autonomous robots.
Actually, we can produce a motivated full autonomous robot from a humanoid or football robot, which can perform the elementary motion and observation activities, after some technical modifications, by implementing the method presented in my patent, in 2 - 3 years dependent on the number of specialists involved.
It is easier to make a motivated full autonomous industry robot than an other full autonomous robot, because (a) it must not look like a human and (b) the situations in its surrounding are foreseeable and its number is not so great, compared e.g. with the number of situations which a housemaid robot must store.

The existing robots in industry are only the beginning of using industry robots. In the near future, many production processes will be carried out by robots and other machines, where the motivated full autonomous robots will operate and control these machines and other robots. These robots will produce cheap products night and day. More about industry robots   http://mywebpage.netscape.com/alschurm1/page5.html

With friendly greetings

Schurmann/shurman-a

[Admin]

Hmmm I know what you are getting at, but I would argue that in the end, robots will only do what you program them to do. Does not matter how advanced your program is, how good the AI is, or even if it incorporates adaptation algorithms - it will still do only what you program it to do. You could of course half-hazardly write a program with a bunch of errors and algorithms not planned out very well, and when the robot runs around it would do things that surprise you - but it will still do exactly what you program it to do if not what you planned it to do . . .

I actually tried what you talk about. My strategy was to use as many different sensors as possible (eight total, as limited by the ADC's) and integrate them with fuzzy logic (http://www.societyofrobots.com/programming_fuzzy_logic.shtml). But my program was too complex, I ran out of memory on the PIC, and quit with a functional but not yet smart robot.

its posted here if you are curious:
http://www.andrew.cmu.edu/user/jsp2/robotics/fuzzy/index.html

Anyway keep me updated with what you learn, I am really interested 

[shurman-a]

You are of course right. I am not interested in so called developmental robots which use said half-hazardly written programs. First of all we should use reliable algorithms and all posibilities which are in motivated autonomous robots.

shurman-a

[Gopher]

I don't really see the advantage of such autonomous robots; it's far more complex and expensive than a suite of simple robots to perform each particular task, with a simple supervisor program that is able to guide the operation of those robots with simple commands.

Don't get me wrong, though, a truly sophisticated humanoid robot is definately something I would love to have myself, but I don't see a time ever coming when there's a housekeeper bot in every home, the current approach of making the existing tools smarter and suplimenting them with a other specialized robots is just more practical.

Don't let that deter you, though, because I still really want one myself, and I have no doubt others do too!

[shurman-a]

Hallo Gopher,

You write:
I don't really see the advantage of such autonomous robots; it's far more complex and expensive than a suite of simple robots to perform each particular task, with a simple supervisor program that is able to guide the operation of those robots with simple commands.
First: This supervisor program must be implemented in a robot or a computer which must send and receive many many detailed information to and from the supervised simple robots.
Second: Such supervisor program cannot be simple, if the system should work properly and autonomously - it must be intelligent enough to control and operate the simple robots; thus you must develope a supervisor/control program which would be similar to my steering/controlling method in order to operate and control the simple robots intelligently.
Third: Your proposed system of simple robots is neither cheap nor efficient; e.g. it would not be able to operate a washing machine (or you propose a special robot for operating a washing machine?).


 a truly sophisticated humanoid robot is definately something I would love to have myself, but I don't see a time ever coming when there's a housekeeper bot in every home, the current approach of making the existing tools smarter and suplimenting them with a other specialized robots is just more practical.

Once the motivated autonomous robot is produced, it would be cheap and housekeeper would use it. I think however, that it is more easy to build the motivated autonomous robot  as an industry robot (see http://mywebpage.netscape.com/alschurm1/page5.html

shurman-a

[Admin]

Im gonna have to agree with gopher on this . . .

Remember those microwaves that came out 2-3 years ago which you can check your email with for 10 times the price?

Thats why. No one wants to pay a lot extra for features they dont need.

As for manufacturing, its been shown time and time again that for mass production robots are just too expensive and complex. Machines dedicated to just one or two simple tasks outperform robust robots in cost, speed, and quality when it comes to mass manufacture. Robots only perform better than the dumber machines during very short production runs, or production lines that need to be changed often in time.

[shurman-a]

The proposal of Gropher "some simple robots and a supervisor computer/robot" is the so called "intelligent flat/house". It is not cheaper than a motivated autonomous household robot but more expensive.
It is very important, that the operations/simple actions of these simple robots are very restricted. They cannot prepare meals, they are even not able to make tee and serve it to a person being in bed. It is impossible to have a specialized robot for each activity which a housemaid does. So the "intelligent flat" cannot be a solution for making household activities.

Admin: As for manufacturing, its been shown time and time again that for mass production robots are just too expensive and complex. Machines dedicated to just one or two simple tasks outperform robust robots in cost, speed, and quality when it comes to mass manufacture. Robots only perform better than the dumber machines during very short production runs, or production lines that need to be changed often in time.

This contradicts the reality; example: cars are mainly manufactured by robots, and therefore cars are relatively cheap.
In opposite to existing robots, the motivated autonomous robot is able to adaptation to a new machine, device, room or new activity. Thus, if a manufacturing line is modified then the motivated autonomous robot can be adapted to these modifications.

But with the autonomous robots new manufacturing possibilities are open; examples: (i) coal mining 2000 m underground done by a staff of motivated autonomous robots, simpler robots and machines (existing robots cannot do it), (ii) a manufacturer which processes chemical substances could be operated and controlled by a staff of full-autonomous robots, simple robots and other machines, (iii) a staff of autonomous robots, not autonomous robots and other machines could mine solid methane located 1000 m underwater, take it into containers and transport it to ships/gas containers oversee. Without full-autonomous robots this cannot be done.

Schurmann/shurman-a

[Admin]

Actually, only the assembly process of cars are by robot arms. Everything else, such as manufacture of all parts, are done 100% by dumb highly limited machines 

And the arms are required for as I said, "production lines that need to be changed often in time." Car companies change the designs annually.

There is this book on robotic manufacturing automation called Affordable Automation that you may be interested in. I learned a lot myself from it . . . May be worth a look.

[Gopher]

shurman-a said:

First: This supervisor program must be implemented in a robot or a computer which must send and receive many many detailed information to and from the supervised simple robots.
Second: Such supervisor program cannot be simple, if the system should work properly and autonomously - it must be intelligent enough to control and operate the simple robots; thus you must develope a supervisor/control program which would be similar to my steering/controlling method in order to operate and control the simple robots intelligently.

I don't agree with the characterization that the supervisor needs to "control and operate" the simple bots; it makes it sound like the other bots are being teleoperated by the supervisor, and that the supervisor needs to understand their jobs. In fact the server needs only to give the most basic information to the bots. It doesn't have to "control" the vacuum, just say "You can go vacuum now" and maybe "Stop and go recharge." Neither of these decisions requires much intelligence or reasoning on the supervisor.

shurman-a said:
Third: Your proposed system of simple robots is neither cheap nor efficient; e.g. it would not be able to operate a washing machine (or you propose a special robot for operating a washing machine?).

No, I propose a washing machine that can be operated through simple wireless instructions. This is not a terribly complicated thing to do, and would have a relatively small impact on the price of an already-digital washing machine. The only remaining tasks would be loading and unloading the machine, which could be handled by a fairly simple "hamper-bot" which has just enough sense to navigate from the bedroom/bathroom (wherever you put it) to the washing machine, and appropriate manipulators to allow it to load the wash, as well as swap clothes from the washer to the dryer. It could be sold as a high-end accessory for the washer/dryer, or even bundled standard.  The only tricky part of this entire process (and by tricky I mean "not easily reducable to a very simple mechanical steps") is folding/hanging clothes. A bot would have to be pretty intelligent to pull out individual garments from a pile, select the appropriate folding or hangar, and put them away in the right place. With a bit of ingenuity, though, I'm sure a relatively simple bot could be designed for even this task; again, it would likely be a high-end accessory sold to those not only too lazy to fold their own clothes, but unwilling to carry them to a laundromat (and if you can afford a 'bot to automate your household chores, you could certainly afford a laundry service!)

I agree there might be a place for these supervisor bots in industrial settings. However, household economics and industrial economics are two very different things; I don't think that household chores are complicated, demanding, or critical enough to justify the complexity of an autonomous humanoid robot. Unlike a factory, 24/7 operation is not necessary or desireable; the range of unexpected situations is larger, but their importance is far lower. If you get home and the dishes aren't completely clean because the 'bot responsible was perplexed by some unexpected connundrum it is unable to deal with, this does not cost anyone money, and is unlikely to hurt you in any significant way. You just straighten it out, remove the obstacle or whatever, and it goes back to what it was doing. Since it takes far less than 24 hours a day to keep a house in perfect order, it will be caught back up in no time. With a set of simple, single-purpose household bots (many of which don't even need to be mobile), then even if the dishwasher bot is stumped or even breaks down completely, then only that task is interrupted; the vaccum bot will still vacuum. The best part is that the separate bots allow for scalability of the level of automation, where your system is an "all-or-nothing" system that requires a huge investment - I imagine a household robot with the level of automation you describe being comparable in price to a luxury car. A total suite of bots to provide the same level of automation might be just as expensive, but it can be bought a piece at a time, the same way we upgrade our appliances.

For a factory/industrial setting, every idle minute can cost real money, and an autonomous factory supervisor who works efficiently 24 hours a day without lunch breaks or vacations could be worth hundreds of thousands of dollars to the factory - which might be just a years' salary for a group of humans to perform the same job. I don't think we're even there yet, but we are getting there. But note that, if the supervisor's job is just to oversee a group of specialized robots, does it really need to be a humanoid either? Surely even in this case, a central computer controller that is just capable of remotely viewing the factory and communicating with the bots wirelessly could do the job as well? If a problem requires action, simple drones can handle it. Say a part slid on a conveyor and is blocking the assembly line. The bots on the line are specialized, and not able to deal with this simple situation. The supervisor gets a signal from one of the bots that there's a problem - "the next part has not arrived" - uses it's cameras and identifies the problem. A command is generated and sent out to a little forklift-bot, which scurries to the scene, grips the object and moves it back to the center of the belt. This bot needs only to be able to navigate to a specified location and move objects. The supervisor could even guide this process using targeting lasers, eliminating the need for these floor drones to be independently capable of sophisticated visual processing (paint the obstruction, so the bot can see and identify it, then paint the place it should put the object) A body, humanoid or otherwise, would increase the cost and complexity, and decrease the reliability, of the supervisor, without seeming to provide much added benefit, Seems better to just wire the entire facility with a variety of sensors and cameras, so that in effect the entire factory structure is the supervisor bots' "body." In effect, though, this reduces the supervisor to just another specialized robot; it doesn't do much of anything, it simply collects information and turns it into instructions, which the other specialized 'bots know how to execute. I may be wrong, but this seems to be within the range of an expert system, and not one really requiring what I think of as an "autonomous" robot at all?

Your examples of applications for autonomous robots all have one very important difference from a typical factory: they're all in settings where it is expensive, dangerous, or outright impossible for human workers to go. This is not true of the typical factory, and so the extra cost is not justified.

A final note: I use the word "robot" very inclusively, to describe any machine which operates itself, however simple and limited it's function; I point this out because Admin seems to use a stricter standard,  not including most of these simple mechanisms as "robots."

[Admin]

I use the word "robot" very inclusively, to describe any machine which operates itself, however simple and limited it's function; I point this out because Admin seems to use a stricter standard,  not including most of these simple mechanisms as "robots."

Its generally accepted that no two people can agree on the definition of a robot. I cant even agree with my own definition! But when I talk about robots, I refer to what the general (possibly uninformed) public percieves as a robot. Basically, if it cant battle Arnold Schwarzenegger, then its not a true robot 

The components in a washing machine are no different from whats in any robot, and it falls in the mechatronics category like other robots. But I think most people wont call it a robot unless they can see 'life' or a 'personality' coming from it, and that its capable of moving from a stationary position on its own. I guess I want to lean to that definition myself because frankly, washing machines un-glorify my hobby 

[Gopher]

Nah, that's not true. You make the coolest kind of robots: mobile robots! I didn't make the point to contradict you, anyway, just to avoid confusion from the mixing of my use and your use of "robot" in the thread.

[shurman-a]

Gopher said:
 In fact the server needs only to give the most basic information to the bots. It doesn't have to "control" the vacuum, just say "You can go vacuum now" and maybe "Stop and go recharge." Neither of these decisions requires much intelligence or reasoning on the supervisor.

In order that the server/supervisor gives the basic information to the bots, it must be full-autonomous and it must be so intelligent that the bots do right jobs at right place and right time, and it must have basic knowledge of functions of the bots. My method for steering and controlling the motivated autonomous robot is just such an autonomous supervisory system. If your system should work properly, the supervisor/server cannot work in so simple way as you think, without having knowledge and intelligence concerning its surrounding.
Example, the washing machine: Of course, the supervisor implemented in a computer can give wireless instructions to the washing machine and bots ( actually computer with such wireless communication, supervisor and controlling operations is a robot). But the problem is when the supervisor should give these instructions. Your mentioned "hamper-bot"  cannot be simple as you say, it must at least have the intelligence of (i) navigating in the flat, (ii) taking the right clothes in the washer, and (iii) take the clothes out from the washer and put it in the drawer. Actually, it is not a bot but an intelligent robot. And what with operating other devices? Another such bots/robots would be needed?- or the washer bot/robot would manipulate it too? In the last case, such bot/robot must be an all-round intelligent robot. In any case, such system is very expensive, uncomfortable and its complexity is greater than the complexity of the motivated autonomous robot: in your system, not only the supervisor must be intelligent and autonomous but also several other robots (bots according to you). As I see, it is cheaper and simpler to implement the needed intelligence and autonomy into one robot.
The cost of a motivated autonomous household robot: I suppose that, if it would be manufactured in mass, then the price would be about 6000US$.

Gopher:
But note that, if the supervisor's job is just to oversee a group of specialized robots, does it really need to be a humanoid either? Surely even in this case, a central computer controller that is just capable of remotely viewing the factory and communicating with the bots wirelessly could do the job as well?...........
..... A body, humanoid or otherwise, would increase the cost and complexity, and decrease the reliability, of the supervisor, without seeming to provide much added benefit, Seems better to just wire the entire facility with a variety of sensors and cameras, so that in effect the entire factory structure is the supervisor bots' "body." In effect, though, this reduces the supervisor to just another specialized robot; it doesn't do much of anything, it simply collects information and turns it into instructions, which the other specialized 'bots know how to execute. I may be wrong, but this seems to be within the range of an expert system, and not one really requiring what I think of as an "autonomous" robot at all?


Indeed, for a factory setting, a motivated autonomous system must not be a humanoid robot. However, your described central computer controller with bots, drones and cameras is a robot operating and controlling it; and it must be autonomous in order to operate and control the production properly. My steering and controlling method for a robot just specifies how such system works intelligently and autonomously.
 
A.Schurmann/shurman-a

[Gopher]

I still don't understand why you feel the robot supervisor can't be a simple expert system. Expert systems can be "autonomous." Autonomy is not, in my mind, the holy grail you seem to think it is. The acheivement is not autonomy, it's a combination of autonomy with extreme versatility that is hard to acheive. An expert system supervising and directing the operation of a group of specialized single-purpose bots doesn't seem to require much versatility to me, though I admittedly have not studied the problem exhaustively.  What I know is that the number of situations that occur in a factory (that apprently wouldn't even contain people, the most unpredictable element in an ordinary factory) is definately finite, and could be programmed into an expert system.

[shurman-a]

Gopher: Expert systems can be "autonomous." Autonomy is not, in my mind, the holy grail you seem to think it is. The acheivement is not autonomy, it's a combination of autonomy with extreme versatility that is hard to acheive. An expert system supervising and directing the operation of a group of specialized single-purpose bots doesn't seem to require much versatility to me, though I admittedly have not studied the problem exhaustively.  What I know is that the number of situations that occur in a factory (that apprently wouldn't even contain people, the most unpredictable element in an ordinary factory) is definately finite, and could be programmed into an expert system.

Well, I suppose your system works autonomously in standard situations. However when a non standard situation occurs, this system wouldn`t work properly ( may be so wrong that a great damage appears), because the system is not intelligent enough. The central computer controller have not the needed knowledge of the surrounding and it cannot do complex activities (about the  intelligence of the motivated autonomous robot see  http://mywebpage.netscape.com/alschurm1/page6.html ). Here are examples when the system wouldn`t work properly:

E1: A machine M1 does not work properly. The controller/supervisor orders a bot-robot, R1, to repair it, which it does. After 2 min a second machine, M2, of kind M1 does not work properly too, but it couldn`t be repaired because the robot R1 is repairing M1. When machine M1 was making a not so important work and machine M2 very important one (because e.g. of delivery time) then the system works not intelligent.

E2:  Machine M1 works not properly because the object, G1, it must to work on hasn't been made properly by other machines, although they are working properly. Neither the supervisor nor the bot-robot for machine M1 could find out why the object G1 has been made wrong. The supervisor needs more intelligence and knowledge about the machines and the object G1.

E3: Several wrong events F1, F2, ....,F10, appeared. No one of these events, separately, is very important and the supervisor orders appropriate bots-robots to handle each event Fi separately, which takes more than 30 min. However, as a whole these events are a dangerous situation, which neither the supervisor nor the bot-robots are able to recognize, and a damage similar to an explosion will occur. Without a working knowledge about its surrounding and the ability to perform complex activities, the supervisor/controller cannot function properly without human supervisors having the needed working knowledge.
 
A.Schurmann/shurman-a

[Gopher]

....


Responses to examples:

E1: Prioritization is a simple matter and any defined tasks (such as the ones M1 and M2 are responsible for) would be ranked by importance. An expert system could handle this no-problem.

E2: Rather than increase the complexity (and cost) of the system 50-fold, wouldn't it be cheaper for the factory to send out a red flag and call in a human repair man in these extremely rare circumstances? Besides which, if bots all think they're working correctly, but they are wrong, then this is a programmer error. Once a human identified the problem, they would not only repair it but fix the programming problem that made the mistake go unnoticed the first time. Much cheaper, up-front *and* in the long-term, than trying to make a 'bot capable of undirected troubleshooting of an entire factory full of specialized bots!

E3: What of "expert system" don't you understand? You would never leave an expert system in charge without programming in every possible dangerous scenario in. Of course, this is a complex and difficult task, and there's always a chance something could be missed, but that's really more an argument for not removing all humans from the situation, rather than an argument for trying a different approach to the bots? After all, the more complex your bot is, the more likely it is to fail! Expert systems can be predicted and controlled, a bot that uses it's own logic or reasoning isn't, and is far more likely than an ES to have surprising and negative responses to an unexpected problem.

E3:

[shurman-a]

Your response:

E1: Prioritization is a simple matter and any defined tasks (such as the ones M1 and M2 are responsible for) would be ranked by importance. An expert system could handle this no-problem.

It is known that prioritization is not a simple matter when directing and controlling complex systems, because (a) assigning constant priorities leads to unintelligent task performing and (b) dynamical priorities must be assigned and here arises the question how? Thus, your expert system must have procedure/method for assigning intelligently dynamical priorities to tasks/activities.

E2: Rather than increase the complexity (and cost) of the system 50-fold,...

It is a joke. Application of motivated autonomous robots would decrease the amount of hardware (reduce bot-robots, cameras e.t.c.) by 70%. Thus, not only the complexity would be reduced but also the cost. The cost of a staff of at least one autonomous robot and specialized robots consists of (i) the cost of the hardware and (ii) the cost of the software. The cost of the software would be low, because once such a software is build it can be, with low costs, adapted to similar autonomous robot systems.

wouldn't it be cheaper for the factory to send out a red flag and call in a human repair man in these extremely rare circumstances?  

However, in order to send out a red flag, your expert system must recognize the bad/wrong situation - thus it must be able to recognize many unwished situations which may appear and do or order appropriate activities (note that sending often a red flag, when it is not necessary, is a wrong activity). The answer to the question what robots should do and what human workers, depends on the kind of considered manufacturing.

if bots all think they're working correctly, but they are wrong, then this is a programmer error.

Not at all. Does your specialized bots think? According to you, they are specialized to do some operations on a given machine; bot specialized in operating machine M1 is not able to operate machine Mx; a bot is not able to recognize situations in which other machines are involved (or you changed your mind and think now that such bot is not so specialized and is able to recognize situations concerning the production line?). To blame the programmer, because the whole system does not meet the requirements for supervising and controlling a manufacturing is not fair; I wouldn`t like to be a programmer supervised by you.

E3... You would never leave an expert system in charge without programming in every possible dangerous scenario in.

Well, I must conclude that your expert system-robot does: (i) it assigns intelligently dynamical priorities to tasks and situations, (ii) it stores and recognizes all important situations which may appear and (iii) it does or orders activities which handle all important especially unwished situations which may appear in its surrounding.
But just such robot and expert system is the motivated autonomous robot, and it is far not so difficult as you think to produce motivated autonomous robots.

shurman-a

[Admin]

Application of motivated autonomous robots would decrease the amount of hardware (reduce bot-robots, cameras e.t.c.) by 70%.

How did you quantitatively arrive at this value?

You should run some simulations on this, me and gopher just arent convinced 

[shurman-a]

In the staff of motivated autonomous robots, one robot does the work of approximately 4 - 6 bots. In this way, one needs 4 - 6-fold less robots. Also the number of cameras is reduced by about 70%, because they are not needed - the robots observe the manufacturing using only 30% of cameras.

shurman-a