There is some sort of mystery available in the history of industrial automation. It has to do with the contradiction between technological advancement and marketing this technology to the public. First thing to know is, that robotics can be realized with today's technology. It is possible to build kitchen robots, delivery robots and even humanoid robots. The engineers have mastered the problem of biped walking, object grasping and pathplanning. At the same time they haven't build anything and this contradiction should be explained in detail.
To explore the difference between technological capabilities and the ability to sell robots on the market further we have to focus only on commercial robots. That are robots which are sold by customers to customers. There are a lot of example available from the last 50 years in which commercial robots were presented. The interesting situation is, that not a single robot was a success. The prediction is, that because of hard to understand facts it won't be able in the future to sell more advanced robots.
Perhaps it makes sense to imagine how a newly started robot whould looks from a technical and from a marketing perspective. First thing to do is to build and program a robot. A typical entry level example would be a delivery robot. The robot is able to pick up objects and deliever them to a location somewhere in the city. Writing the software which controls the robot autonomously is not very complicated. It is some sort of advanced path planning algorithm which is used to control ingame AI in a racing simulator. It is possible to show that the robot's program is working correctly. This is equal to recognize, that the technological development process was succeeded.
Now let us take a look into the marketing perspective. The question is how to advertise such a robot so that customers will pay real money for the device. And here is the bottleneck located. Because humans have no need for a delivery robot. And if the companies who has developed the robot will explain in detail how it works and what to do with the machine, the customers gets irritated even more.
From an outside perspective, a common argument to explain the mismatch is the assumption that the mainstream customer doesn't understands robotics or he is not interesting in technology at all.But this is not true, the customers are understanding the details very well and they are want to make advanced automation a success. Nevertheless they won't buy a robot, especially if the machine is able to do real life tasks.
Let us go into the details. Robots are mostly advertised as a productivity machine. The idea is, that a robot can do the task of a human worker but much faster and cheaper. The sad reality is that they are struggling exactly in this context. Not because of the robot itself, but because of the description of the robot.
Now let us imagine the same robot is marketed with a different objective. This time, the robot is sold to university and the promis is, that the robot will produce large amount of costs, and it will require 10 computer science and 100 phd holders to understand the system. This time the robot can be sold easily to the customer. Again, it is the same delivery robot which is working with the same hardware and software. The only difference is, that this time the marketing claim is different what the purpose of this robot is.
There is a demand for overpriced useless robots which will reduce the productivity. These robots can be sold to space agencies like the NASA or to a research center at the university. it will need millions of US$ to buy such a robot and it will require hundred of manyears to program these machines and write papers about it. The interesting situation is, that such an interaction works very well because the advertised product and the delivered technology is the same.
The only problem is available is robots are sold as cost saving productivity devices which is able to replace human workers. Such a promise can't be fulfilled by today's robots. No matter if it is a biped robot, a self driving car or a household robot. None of these device can operate in this mode. Robots are always research projects which are producing cost and have a demand for highly qualified researchers. A technical university is the best source which can deliver money and highly experienced computer science students.
It is possible to verify this theory by a short look into the reality. The amount of robots which are utilized to reduce labor costs or automate the workflow in a factory is exactly zero. Not a single company nor a household owns such robots. At the same time, robots have become a widely accepted technoogy within research universities. Nearly 100% of the universities have at least one biped in the laboratory and for educational purposes. The prediction is, that this trend is stable. That means in 20 years from now, the amount of robots on the road and at home remains at the same level which is 0.0 while at the same time, the amount of robots especially highly expensive one in universities will grow further.
To understand this paradox we have to ask some basic question. What is the reason why a computer science laboratory at the university has a need for starting a robot project? Because it helps to explore new algorithm, teach the students how to program robots and perhaps it helps the university to promote itself as a technology friendly institution. That means, a modern biped robot in a university classroom fits very well together. On the other hand, the same robot model won't fit to the needs of households or companies. Both stakeholder have a demand for productivity increasing machines and they want to save the costs. Also they have no interests in programming the device but they want to buy something which works out of the box. This is not the case for a robot especially not for a recent robot which is working with neural networks.
It seems that from an abstract point, a robot is a machine which is consuming a lot of money and lot of intellectual capital and at the same time it doesn't increase the productivity and it doesn't help to reduce the costs.
There is no explicit rule available that a self driving car can only be controlled by computer science students. It is possible that non technicans buy such a car to have fun with it and reduce the amount of accidents. But in the reality, such a use case is not there. That means, self driving cars are only available for computer science students but never for non technicans. The reason is, that there is no sense to use a self driving car for something important. Such a machine works only in the university context.
Perhaps it makes sense to describe the term working in detail. A working robot isn't defined by it's hardware nor its software but it has to do with social roles. A robot makes a lot of sense in an academic paper, A roughly estimation comes to the conclusion that there are at least 100k papers written in the past 50 years about all sort of robots. And these papers were created always in an academic context. So it seems, that the combination of a robot plus a university context makes a lot of sense. At the same there is not a single case available in which robots were used in a non academic context for example in a house hold or by a company to increase the productivity. So it seems, that such interaction doesn't make sense.
In simpler words it is possible to say, that robots are everywhere except outside the university campus. On the university robots are used in a simulator, for testing out real hardware, for building self driving cars, implement qlearning algorithm and even for programming vision systems for drones. It is possible to select randomly a technical university somewhere in the world, and the chance is high that this university has not a single and not 10 different robots but more than a dozens of them which are used frequently in projects and documented in papers.
On the other hand, the amount of robots outside of university projects is little and even zero. It seems, that the technological progress in the last 20 years doesn't help to introduce robots in the real world but has increased the productivity paradox. There are no companies available which are selling robots and there are no households available which have a demand for these machines. 100% of the house hold kitchens and all the public roads are working fine without any robots. And the chance is high that this situation will remain the same for the next 20 years. With the robot itself every thing is working fine. The only problem is the marketing effort to sell robots for practical applications. Robots are useless as self driving car, as delivery drones, they have no advantage for self driving ships or as security devices in a shopping mall. Robots are useless in a restaurant and they no advantage in a kitchen. Robots have failed at the assembly line fore example in the automotive industry and robots are a poor choice in a hospital. Basically spoken, robots are useless in every context except for educational purposes. The only sense making application for a robot is if the device is sold to a university.
The funny thing is, that most computer science students are empathizing that they have developed practical robots which have greatly improved over the years. What these science students are never telling to the world is, that the robots won't leave the university lab. All the biped, grasping and flying robots can't be utilized for practical applications. And if someone is trying to do so the project will fail.
Let us try to understand the paradox from a positive perspective. Instead of asking what robots can't do let us focus on successful robot project. A simple game might help to understand the reality. The works in a way, that we are imagine a city or country randomly and then ask google if the university in this location has at least one robot project. First trial is the city “london”. After entering the search term “university london robot”, not one and not ten hits are shown but hundreds of them. all the colleges, university and technical courses in London have robot projects. They have biped robots, UAV drones and pick&place robots. The students can learn how to program these robots in java, in C and even Prolog. Next trial is to imagine the city of Los angeles. The situation is nearly the same. Lots of technical university are explaining what robots are and how to program them. The user can decide to read the website, the paper or even watch videos from one of the universities. That means, in los angeles, there are robots everywhere.
On the other hand, it is very hard to find robots not connected to a university. For example in a hospital or in a logistics company. It seems, that in the same city the robotics technology is well researched at the university but the same technology is never deployed to the business world. This is not a coincidence or the result of single decision making process but it is some sort of unwritten rule which is valid for the last 50 years. The rule basically says that robots are great at the university but are doesn't produce sense for the world outside the ivory tower.
Sure, many attempts were made to deploy robots in real world companies or private households. but mostly the project was terminated after a while and it was always a failure. The reason is, that universities and normal companies have a different use case. A normal companies is interested in increasing their productivity, lower their internal complexity and increasing their profit. But a robot can't fulfill these requirement. What robots are doing is to produce costs, increase the complexity and they have a demand for computer programmers. So it seems, that robots aren't producing something but they have a demand for money and knowledge.
Educational robots
In the first path of this essay it was explained why robots are a poor choice for automating the retail industry or to drive on public roads. What was missing is a positive description in which domain a robot is useful. Some examples for educational robots are:
- the lynxmotion robot arm
- arduino mobile robots
- Legomindstorms NXT line following robot
- nao humanoid robot
- Willow garage PR2 kitchen robot
- robocup robots
- self driving car robots on a floor
- pick&place robot demonstrator to teach the PDDL language
- Duckiebots kit (self driving car on a floor)
- and so on
These robots are sold for a different price range which is starting at 200 US$ and has its upper limit at 10k US$. For example the NAO robot or the PR2 are very expensive one which is used in a university context. The interesting situation is, that all these robots can be recommended. They are used frequently in the past and lots of papers were written about ongoing projects. So it seems, that the mentioned robots are fulfilling the needs of students and teachers as well.
In addition it is possible to recommend some robot simulators which are not working with real hardware but are computer programs on a desktop PC. A typical example would be a pong game which is enhanced by an artificial intelligence.
The surprising fact is, that even the robots are highly successful they can't leave the university context. They are research only robots. Sure it is possible to demonstrate practical applications for example a self driving car or a pick&place robot, but these demonstrations are not useful for a company or in a houosehold context.
Let me give an example. Suppose a user is buying the PR2 robot with the idea to use the device in the kitchen at home. Such a usecase won't work. What the user can do is to create an academic robot project at home and write a paper about it. And this interaction is the reality for all the mentioned robots. It is simply not possible to modify a research robot so that it will fulfill a real life task.
Or let me explain it more precise. Al the mentioned robots like PR2, lynxmotion arm or the Nao model are a poor choice if the goal is to automate something or reduce the costs in a factory. The robots were not build for this purpose. The working thesis is, that no robot in the world can be used for this purpose. Even if the robot is constructed very robust and will need high amount of voltage it is useless in a factory or at home.
The insight is not completely new. In the 1908s some robot projects were started in the industrial context. The idea was to automate the production of a car factory or to increase the productivity in a hospital by introducing robots. All these projects have failed. And the chance is high that with modern technology and with more effort the project will fail again.
On the other hand it is surprisingly easy to initiate an educational robot project. For example to build a simple line following robot. Even if the robot has smaller bugs, it is possible to write a paper about it and it is for sure a real educational project. In a short sentence it makes sense to reduce the amount of robot projects in factories and increase the amount of robots in universities. Such a priority will make sense for all the participants.
A good example for a successful educational robot is perhaps the micromouse challenge. Micromouse is one of the oldest projects. It is repeated every year not by a single universities but by many of them worldwide. The amount of publications about the subject is very high so it seems, that the competition is a success.
Perhaps it makes sense to give a precise outlook into the future. Some new robot projects were started by companies all over the world. There were projects started about building harvesting robots in agriculture, self driving cars on public roads, pick&place for logistics applications and of course drone robots for delivering packages faster in the city. All these projects have failed already, or they will fail in the future. The reason is, that in all the cases the goal is to use a robot to automate something, reduce the costs and increase the productivity. It was explained already, that these goals can be fulfilled. it is correct that some engineers are promising that robots can be used for this applications but this is not correct.
The term educational robot was used many times. But what is it exactly? An educational robot is a very expensive device which costs 500 US$ and more. It is doing nothing but the machine will need electricity and somebody who can program it. And if the user is doing so, he can learn what Artificial Intelligence is about. Similar to a toy, an educational robot has no value and it can't improve something but it is the opposite. It is used always in an educational context for example to teach programming and control theory. A typical project which can be realized with an educational robot is, that the students are forming a group and have to program a software for the robot. Such a project will need 6 months and then it gets presented in a paper and a talk.
The interesting situation is most or even all these projects and robots are successful. That means, if the students have fulfilled a project in the last semester they want to repeat it with a slightly different challenge.
No comments:
Post a Comment