Academics play an important role in the field of artificial intelligence, and the laboratory is a miracle. We see a large part of commercial robots, drones, and driverless cars. How is an artificial intelligence project commercialized?
At the CCF-GAIR conference on smart driving, Professor Alonzo Kelly, director of research at the National Engineering Research Center (NREC) at Carnegie Mellon University, shared his views on the commercialization of robots.
In fact, universities do not directly produce robots, but robots also occupy a very important position in the market. Therefore, we must carry out commercialized transformations to help companies better produce robots for better use.
At present, NREC is mainly engaged in the research of mobile robots. Driverless cars are a part of it. Most people's understanding of driverless cars is Google's self-driving cars. Actually, driverless cars actually have various displays 15 years ago.
The key factors in the commercialization of robotsIn the project, we found that there are many patent applications that need to be taken into consideration when creating a packaged solution. Therefore, we consulted some companies to provide them with professional consulting services to make them realize that the product is on the market or is Before the commercial promotion, it should apply for the protection of its patents and intellectual property at the research stage. This is a typical workflow.
At the beginning of our time we saw development through the School of Machines and Research Institutes. There was a mastery of knowledge, mastery of innovation, education, and publication of related papers. There was a series of concepts in between. At the EC stage, we We will talk about the development, integration and performance of this product, reliability, etc., as well as product testing. In the end Carnegie Robotics Institute we will focus on its production, after-sales service, etc., and we consider funding sources, like Google, they all have no It may be a hot thing in the future that people drive the industry to make acquisitions or investments.
TalentIn addition, robots, including unmanned talent, are also scarce, especially engineers. It's not easy to retain these talents. In many cases we find students who graduate from our own college. We have to keep them in school is not an easy task. We found that many companies have encountered huge challenges from employees and a very high Talent pool of talented people reserves.
Some companies around Pittsburgh are successful. There are also some companies that are in the process of incubating, and some are based on entrepreneurship. In the past ten years through our Carnegie Mellon cooperation with some companies, some of them are companies that may be acquired or merged by big companies in the same industry.
solutionLooking at the engineering solutions for building a robot system, we have a very solid framework. This makes a lot of projects easier. There are many problems in the engineering field that are very easy to solve. At the beginning, everyone was more conscious about the project. The perspective and vision provide solutions to the problem.
The robots we use today are often synchronized with the software we use. At this time, we have made some tasks easy to accomplish. Many times we say that in order to avoid the increase in costs, the robot's functions may become very simple and the intelligence is very good. Low robot. Everyone thinks that it is better implemented in the project, but their function is limited.
I was often asked that our robot does not need some technology. My answer is otherwise. If you want to completely escape the GPS in such a robot is impossible, we say this not to increase the cost, or the cost The control is not sensitive enough.
Looking at the example, first of all, we should say that the form should be meaningful in the form of a function. For example, like some robots with tasks, they may look very large. If not, they may be more convenient, and some robots. The above cost can be well controlled.
It may be cheaper for a robot to complete a task than to operate such a construction machine because you see that if it is such an engineering machine, the person operating on it should be comfortable, and the cost is very large, so this kind of misunderstanding and The idea that we are going to change is not like the robots that replace the original machinery and machine products are more expensive than the original products. This is wrong.
We have successful examples to prove such an argument. From a biological point of view, we would say that in the field of biology there is a certain degree of certain areas that can be used by robots, but this is not the case. Many times it is not the evolution of living beings that are the only way. We sometimes Can rely on other products like laser expansion machine, we can see most of the robots in the game when the barriers to the barriers fell, our robot can grasp and achieve such a very precise With the action, we have got everyone's attention on the robot's field and we are very stable.
We can use manual tools to turn on the switch and complete simple tasks. If our robot can grab a drill, it can complete all the simple tasks and tasks. Our solution is a transitional and transformative program that we hope will enable sensors and robots to capture the biometrics and functions of human eyes.
There is a chasm now that the robots have analyzed and studied themselves and adopted a scheme that can drive the car out of such a ditch state. This is entirely outdoors, so there are many satellite positionings and we have our maps under the premise of map planning. The robot can complete instructions such as turning left and right, turning over the mountains. This is completely based on GPS maps. You can drive outdoors and you can let the robot drive your car.
Cooperation with enterprises, win-win!When we talk about the project, we actually go through the process of a project:
The first is whether this project is feasible. The second project is not lasting, because this project is not only scientific research, but also more of a commercial activity. We do not want to make this project a halfway process, or because of our customers. We do not like the products we produce that cause our business to die.
Sometimes we have to discuss a lot with our customers because they think that the robot we produce may not be used in the market, so we need to discuss two issues. The first is feasibility. Second continuity.
The big companies are technically very limited. Although some of the robots they produce can move freely up and down and grab them, they still need manual manipulation.
NREC will help them apply for invention patents. At present, our invention patents include planning, navigation, and drawing. Because most of the software in today's software is actually done by the machine. Imagine how many encodings are needed if we want to build a complete automation system?
About 1.35 million!
This includes simulation, control, planning, cruising, and other parts. The cost of each coding is 650 US dollars, so in the end our total cost is 1.35 million US dollars. This is actually a big fund to build an automated system. It requires high costs, so we will cooperate with several companies. In addition, the company is willing to cooperate with universities because it can reduce costs, and it can also reduce the government's waste of funds.
More than 30% of NREC's projects are supported by the government, so we will authorize companies to help them apply for government funding. In addition, if universities and companies cooperate, sometimes it is difficult to finalize contracts. The university has its own The management system, and the companies they hope to earn high profits, the university hopes to keep its own research and development secrets, the company needs to master these R & D mechanisms, so we will sign this kind of unreliable agreement with the company, for certain projects The university will not transfer this to them. We have actually done a similar thing with one of the companies. We mainly give them the perception system of the robot. The other systems are all of them doing their own, and we need to The system is subdivided, in the end who needs what part of the intellectual property. What's more special about this robot is that it not only locates the environment around it, it can also find out where the trees are in the environment or where the cars are.
Let's talk about authorization. Authorization is more complicated, because under normal circumstances our company is directly engaged in research and development in the United States. Therefore, if universities cooperate with companies, they must authorize them. For them, they do not need to directly To buy a patent, the cost is even lower, so that if the cost is low, the company's profits may be higher. University profits are mainly licensing fees and copyright fees.
Of course, everyone now has a variety of guesses and predictions about robots. Some people think that robots should be cheaper. Some people think that robots should run faster. To give an example, the left side is mainly supported by the government. We spent 40 million U.S. dollars, and built such a robot out of thin air for 7 years. The middle project was done from 2008 to 2015 and it took five years. The project on the right only spent 2 million US dollars, and it will be ready in 1 year.
Therefore, the cost of cooperating with enterprises to produce industrial robots is indeed lower, and companies have also solved some technical problems. This is a win-win process.
copper flex Membrane Switch, FPC Membrane Switch With Metal Dome,Membrane Switch With Embossed Button,Membrane Switch With Metal Dome
CIXI MEMBRANE SWITCH FACTORY , https://www.cnjunma.com