It takes at least five years for the humanoid robot to enter the family.

With the rise of Chinese power in the field of artificial intelligence research, more and more Chinese scientists are engaged in the process of commercialization of artificial intelligence technology.

Recently, Ubunt Technology announced that Tao Dacheng, a professor at the University of Sydney, is about to join the company and is the chief scientist of artificial intelligence. At the same time, in terms of robot motion control, Tsinghua University professor Zhao Mingguo will also serve as the "Chief Scientist of Humanoid Robots". Tao Dacheng, Zhao Mingguo and Yu Jian, founder of Zhou Bi, accepted media interviews including the “First Financial Daily” in Beijing to discuss a series of interesting issues such as the commercialization of humanoid robots.

Four thresholds for entering the family

The strategy for the humanoid robot is to complete the commercialization of high-torque, high-explosive servo servos in 2018 and launch a series of bipedal family service robots. At the same time, it is considered to launch the humanoid home service robot operating system within three years, and launch a humanoid home service robot based on commercialization purposes.

Zhou Jian believes that human needs are not specific. Some people may need real-life, emotionally dependent robots to accompany them. If they are single, the imagination space will be great. Commercialization of such robots within 5 to 8 years is possible. Future humanoid robots will be infinitely close to people in motion control and artificial intelligence.

"We have been discussing with Amazon for a long time. Amazon believes that the development of human-computer interaction in the future, the consumer center must be a humanoid robot." Zhou Jian mentioned that the main source of income for 2017 must be four, the sales revenue will reach 10 Billion to 1.5 billion yuan. One of them is the robot developed in cooperation with Amazon.

Tao Dacheng analyzed the research content of four aspects of humanoid robot from the perspective of software. He pointed out that the most critical thing about humanoid robots is visual output. More than 70% and 80% of the information we get every day comes from the eyes, so the first thing is to do image processing and how to make the acquired images clearer. After seeing it clearly, it is machine identification. As a home service robot, we must first know every member of the family so that we can create user portraits for them and then sum up the information of family members. To achieve stable image processing and machine vision, two very critical fundamental aspects of research, deep learning and multi-angle learning, are indispensable.

He said that robots are everywhere in the house, to analyze real-time scenes, and how to accurately achieve indoor navigation, there must be a lot of basic technology support. Understand the environment, the following important thing is to recognize people and understand people's behavior. This is the face recognition system of sports. People are emotional, you can be very friendly, you can be very angry. This requires an understanding of the human expression for the robot. Understand people's emotions, human behavior, we must know how to detect and identify objects.

Tao Dacheng believes that the commercialization of humanoid robots for the family is still to reduce costs. The biggest problem is to estimate the attitude of the person. If the camera and sensor are loaded with a certain amount of money, the cost will be increased. How to use the single-screen camera to obtain the desired Data is the question that needs to be studied. One of his team's recent research results is the calculation of videos downloaded from Youtube.

In addition, another key issue is model compression. During the movement of humanoid robots, there are a lot of computational overhead, electrical energy overhead, and the overhead of artificial intelligence, which leads to a large battery, which is difficult to overcome in many scenarios, so it is necessary to make a big The model becomes a small model.

Breakthrough within 5 years

Professor Zhao Mingguo, who is focusing on the research of biped robot motion control technology, believes that after the basic functions are achieved, it can be iteratively and truly produce a product that meets the needs. Within two to three years, the robot's "grab" function will be in place.

Take the function of “grab” as an example. There are many kinds of bottles in the house. Wine glasses, mineral water bottles, thermos bottles, etc. may be the same in the robot, but how to accurately identify these objects is needed. A scenario in which artificial intelligence technology and motion control technology work together.

Zhao Mingguo also said that the real humanoid robot is a robot with limbs, brain and torso, but its shape is not necessarily the first step. The part that completes the whole torso first can be exercised. With the concept of AI, and then more realistic skin, materials, and various sensors, the touch and temperature will improve.

"In the next five years, combined with the development of drive technology, control technology, and artificial intelligence, we will surely usher in a new breakthrough in humanoid robots in home applications, and will also usher in breakthroughs in industrial applications." Zhao Mingguo said.

Developing a robot, hardware and software technology for better integration is a key issue. Then, from the point of view of the development program, is there any hardware to develop the software, or can the two be carried out at the same time?

Zhao Mingguo believes that robot movement must have a good platform first, so the order of landing should go hand in hand. It is necessary to realize the commercialization of large-scale servo steering gear, and the corresponding software should be developed simultaneously. But the servo steering gear must be implemented at the beginning.

In addition, Tao Dacheng also told reporters that he has been engaged in in-depth learning for a long time. The weakness of deep learning is that when there is not a lot of data for training, the effect may not be too good. For example, the men in the room are short hair, the ladies are long hair, but I can't say that long hair is a woman, short hair is a man.

He believes that human understanding of the brain is still very shallow, and deep learning actually follows the hierarchical structure. We have seen breakthrough results, but there is still a process, just like we can't teach children's calculus for 5 years old, we need to advance gradually.

In addition, Tao Dacheng also pointed out that deep learning is a trend, but it does not mean that the original study should be abandoned. The future of statistical learning and deep learning will be organically combined to solve our needs.

Accessories For Power Line Stringing

Accessories For Power Line Stringing Series of various types including Continuous Duct Rods,Fiberglass Duct Rods,Earth Drill,Earth Anchor,Pulling Socks,Joint Protector,Pole Climber,Wrenches,Conductor Spacer,Spacer Bike,Aluminum Ladder,etc,which is widely used in power line project.All of them are made of high quality materials with reasonable volume,light weight,easy to operate.By high quality material and good design,these kind of tools can be durable and long service life.we are a professional Chinese exporter of Electric Power Line Stringing Tools,and we are looking forward to your cooperation.

Yangzhou Qianyuan Electric Equipment Manufacturing & Trade Co. Ltd is specialized in manufacturing and trade of electric power line transmission tools. Our main products are Anti-twisting Steel Wire Rope,Stringing Pulley,Hydraulic Crimping Compressors,Engine Powered Winch,Motorised Winch,Wire Grip, Gin Pole ,Cable Stand,Mesh Sock Grips,Cable Conveyor,Lever Chain Hoists and so on,which are mainly supplied to power companies,railroad companies and other industry fields.

fiberglass duct rodder, safety helmet, Joint Protector, pole climbers, Horizontal Ladder

Yangzhou Qianyuan Electric Equipment Manufacturing & Trade Co.Ltd , https://www.qypowerline.com