At the artificial intelligence conference CCAI2016 hosted by CSDN, Wang Wei’s former tutor, now Chairman of DJI Innovation Technology, and Professor Li Zexiang of Hong Kong University of Science and Technology, shared with participants that they had obtained the promotion of robots in helping people in practical applications. experience of.
Li Zexiang in his speech
Li Zexiang first introduced some challenges and researches on robot mathematical theory. He said that the most important thing about the robot itself is the mathematical model of the robot's movement state. The state of motion of robots is ever-changing, the state space of these robots' motion is clear, and the establishment of relevant mathematical models is the key to establishing a good robot control system. The main types of challenges facing robots today are:
The state space of rotational motion, and dynamics and stability based on this.
State space of industrial robots
Robots, because of the high degree of flexibility required by the hands and the many tasks to be accomplished, their control is also very difficult, and it is necessary to establish a suitable mathematical model.
Robots that need to perform three-dimensional movements have a very complex state space
Some of the joint structures of the human body (or other animals) are very wonderful. For example, the shoulder joint has a rotating space of plus or minus 90 degrees. It is not easy to establish these joint models and analyze their control mechanisms.
The state space of robots in high-end manufacturing is a new challenge compared with the previous ones.
Non-completely constrained motion planning, which usually has certain constraints, such as how to operate a pen and an object flexibly? How can a cat automatically adjust his posture to fall to the ground when the angular momentum is conserved when landing? This is a mathematical problem. It is necessary to figure out the mathematical model of these problems and then make plans and controls in order to apply them to robots.
The error and quality issues that arise in manufacturing are also issues that affect the effectiveness of the control. At present, we need a product quality standard based on mathematical models to define the conformity of robotic products to prepare for larger-scale manufacturing in the future. Most machines actually have a variety of errors (including manufacturing equipment). How do you locate the relationship between it and its machining parts? This is also a problem. These quality standards and tolerance standards need to be accurately defined and tested using mathematical models.
Finally, some current robots are no longer acting alone. It can be foreseen that more robotic collaborations will occur in the future. What is the state space of the robot with two arms when the robot and other equipment are working in the same environment? It is more complicated. There are multiple robots, sensors, and various other things in one working unit. When the equipment, as a cooperative body, how to describe their state space?
In response to these problems, Professor Li introduced the differential geometry theory of robots and considered the movement of a single body as a differential structure composed of a differential manifold and a unit ball, such as the Earth, which will have all the points on Earth. The coordinate method shown, and the relationship between them when there are multiple coordinate points, is the differential structure of the Earth. And these methods of representing coordinates are differential manifolds. For robots, using the Euclidean space's geometric features, the optimization of the robot's motion to make some maximum and minimum values ​​is the key to this theory.
Li Zexiang summed up his valuable experience gained from many years of industrialization and summed it up mainly in three points:
One: Understand what kind of students you need to train. What kind of students are best for starting a business? Their previous standard was: Good students who scored well in the school, but later found that this road was impossible. Later, Li Zexiang opened a course in robot design competition at HKUST, valuing practice and asking different students to come together to carry out the whole process. In these processes, students learned the design of the system, the management of time projects, and teamwork. In order to speed up the cycle, Li Zexiang also allowed students to go to Huaqiangbei to do PCB processing, which greatly accelerated the iteration time. Students who are trained in this way will have fewer problems if they want to start a business later.
Two: need to find the direction. Robots and AIs complement each other. Li Zexiang made a good score. There is a saying: “Whether a black cat or a white cat can catch mice is a good cat.†Robots and AIs are like black cats and white cats. There is no need to distinguish between black and white, but they focus on how they work together to catch the “ratsâ€. These mice are currently the most valuable applications for AI and robotics. Li Zexiang predicts that several “big mice†in the next few years will have automated driving and manufacturing automation in the 3C industry, which is also a very difficult problem in the field of robotics or manufacturing. The current industrial robots can only complete some of the simplest grinding, polishing and other operations, but this area is very promising in the future.
Three: It is necessary to have such an atmosphere. Such a brand and culture, Li Zexiang said that he has been leading students to complete internships and research at HKUST, and then completed industrialization in Shenzhen. Many of his friends in the United States came to see this scenario and said: This place feels like the Hollywood of Makers and can turn any dream into a reality. Based on these years of company and research experience, Li Zexiang feels that more Chinese entrepreneurs and even entrepreneurs all over the world need such a brand. Therefore, Li Zexiang and his friends created the Songshan Lake Industrial Base for entrepreneurs. They provide a wide range of parts and components, industry chain support, funds, and guidance to help them turn their dreams into reality. This is the positioning of Songshan Lake: a brand that belongs to robot entrepreneurs and a platform to help them realize their dreams.