Current robots have been able to perform precise, repetitive tasks, but many times they are not flexible enough to self-adjust for new tasks, nor can they cope with an unfamiliar or uncertain situation. However, these conditions are changing and robots are becoming more intelligent. We can't help but ask, what exactly is the key technology for robots to realize intelligence? How far have these technologies developed in the last decade? What will be the application prospects of the entire industry? With the increasing demand for application of home robots, the continuous advancement of artificial intelligence related technologies, and the growth of hardware performance, service robots have begun to move from the laboratory to home in recent years, and have evolved from single functions such as sweeping robots to multi-functional personal robots. 01Perception, Cognition, and Behavior Control The robot's technology is divided into three parts according to the usual understanding: perception, cognition and behavior control. Perception is mainly based on visual, auditory and various sensor information processing; the cognitive part is responsible for the higher level of semantic processing, such as reasoning, planning, memory, learning, etc.; behavior control part specifically controls the behavior of the robot. When it comes to robots, one of the most frequently mentioned words is artificial intelligence. Artificial intelligence is a discipline that uses computers to achieve human-like intelligence. The robot itself is one of the ultimate application goals of artificial intelligence. So when it comes to robots, people think of artificial intelligence. Artificial intelligence is indeed very important for robots. The three parts mentioned above are all related to artificial intelligence. From an application point of view, robots have a certain degree of autonomy and can interact with people and the environment. Compared with previous computing devices (including computers, mobile phones, etc.), robots have higher requirements for intelligence, which is gradually attracting attention from artificial intelligence. One reason. Traditional artificial intelligence is a discipline that originated in the Dartmouth conference in the 1950s. After several ups and downs, it has accumulated rich results in basic theories and methods. From early symbolic computing systems to expert systems, to machine learning developed in the 1990s, and big data analysis, they can all be regarded as artificial intelligence. In the fields of image, speech, search, data mining, and social computing, some related applied researches have been derived. Closer ties to robots include computer vision, speech and natural language processing, and agents. From a technical point of view, artificial intelligence must reach a human-level intelligence, and the road to take is still far away, because the intelligence mechanism of human beings has not yet been studied clearly. However, from a practical point of view, according to current technological advances, if we can partially simulate people's intelligent behavior (such as identifying the owner and interacting accordingly) and achieve a better user experience, we will make a breakthrough in the short term. Of course, this needs to further solve the practicality and robustness of technology in technology research and development. After all, many robots in the past are still under development or testing in a laboratory or a limited environment (such as nursing homes). The emerging home service robots will be independent or semi-independent (through collaboration with people) in the home environment. Completing certain services puts higher requirements on the robustness of the technology. Some of them, such as computer vision, speech recognition and other core technologies are still constantly improving, and have not yet reached full maturity. All of these factors have determined that in-depth research and development work is required to achieve a truly practical and intelligent home service robot system. 02 robots are not a bunch of machinery and chips In the past research on service robots, a typical target application is that robots can perform housework, which requires the robot to use the arms to operate objects (pick and place). Although there are many studies in these areas, from the perspective of current technological progress, there are still many challenges in achieving robotic housework within a few years. 1. Current robots, especially humanoid robots, are too costly. For example, a robot's arm may cost tens of thousands of dollars, while the robot's hand needs more than 10,000 US dollars. The cost of the entire robot is more than ordinary families can afford. 2. The robot's dexterity is still hard to compare with the human hand. 3. There are also problems that need to be solved in terms of security. Once a robot with a steel body enters a home, it may cause great harm to people with flesh and blood due to erroneous procedures, inadvertent wave and extension of arms. If robots are temporarily unable to do housework, then what is the use? This is a concern for many people. If it cannot be applied, the robot is just a bunch of machines and chips. In addition to robots that can perform single functions, such as sweeping robots, there are still many robots that are being developed. As a new type of smart device, robots are most concerned about what robots can do for them. As mentioned earlier, robots are still difficult to implement applications such as housework. To reach the level of intelligence, it also requires a longer period of development. Therefore, it is also necessary to design the artificial intelligence technology, especially the cognitive cognitive technology, in the application. It is necessary to make full use of the latest technologies to realize functions that were previously impossible to achieve. It is expected that service robots appearing on the market in the next decade will have the following features. 1. The robot will realize low cost. In the foregoing, it was mentioned that if the service robot is allowed to enter the professional or home service field, the cost must be low enough. The cost of professional service robots can be relatively high, while domestic service robots need to strictly control costs. 2. The robot will have multifunctional features. The service robot currently used in homes is mainly a robot capable of realizing a single function such as a sweeping robot. The future of multifunctional robots will have great development. For cost reasons, some processors in robots will be able to be used in a variety of applications. At the same time, with the development of technology, there will be more and more applications of robots, and robots can accomplish different tasks like humans. 3. Has a unique application. This is the main motivation for users to purchase and use service robots. These applications do not simply transplant existing computers or mobile phones, but have the unique features of robotic applications, such as autonomy, manipulation similar to humans, and being able to interact with people. . 4. Collaborate with others to complete the task. Because robots serve people, they need to interact with people in many ways in order to understand their needs for services. On the other hand, there is still a big gap between the intelligence controlled by robots and people. Therefore, some tasks cannot be completed independently, and they can only be completed with the help of human beings. 5. The robot will have high security features. This includes both information security and physical security. That is, it does not bring security risks to users, and it can even actively detect and stop some potential security threats. The famous three laws of Asimov stipulate basic rules for robot safety. Of course, how to achieve these specific security, especially physical security, also needs to work hard on the robot hardware and software. How the future of home service robots and professional service robots will develop has become a topic of concern in the industry. Home service robots will implement the following types of more typical applications. 1. The home service robot will implement various helper applications. Popular personal assistant software (Siri, Cortana, etc.) on smartphones will evolve from virtual intangibles to functions that physical shape robots can achieve. It can be imagined that the robot assistant can help the user to search for information such as: weather, limit line number, etc., or remind the user. It sounds like there is no difference from the experience on the phone, but in fact there have been many changes in the experience. Users interact with robots in more natural ways, like communicating with people, and even see the robot's expressions, making the service more personalized. In addition to these applications, service robots can also be used as human assistants in more scenarios. For example, a service robot can be used as a person's fitness assistant (see Figure 1). When a person is in the gym, the robot can provide some suggestions like a fitness instructor. I believe that such applications will become more and more abundant with the development of technology. 2. Home service robots will assist children's education. Because of the autonomy and mobility of service robots, it will bring infinite possibilities and wider application space for children's education. 3. Home service robots will enable elderly care. Many countries, including China, are entering or are about to enter an aging society. Old people in China generally hope that their families will be able to provide for the elderly while their children are busy with work. Even if they live together, they cannot always look after the elderly. Such as reminding the elderly to take medicine or timely assistance in an emergency situation, if completed by the family service robot, will greatly improve the quality of life for the elderly in the home to support the elderly, but also make the children more at ease. The care of the elderly will be an application hotspot of robots in the future and have great social value. There are many things that robots can accomplish. Apart from the functions directly related to health aids, they can help the elderly read and read newspapers, and simply chat with the elderly to perform emotional care. For professional service robots, their applications vary greatly depending on the field of professional application. 1. Cargo distribution at the logistics center. It is reported that this application has begun to be actually used by companies such as Amazon. The future will continue to be upgraded, such as the ability to automatically pick up the goods. 2. Serve customers in retail stores, counters, front desks, etc. These service robots can greet guests and answer some basic questions. They can also help customers with their bags, send water to tea and guide customers. 3. Robot photographer. Robot photographers can shuttle between the guests tirelessly to help people take high-quality photos. Of course, this function is also a fresh feature in home service robots. 4. Intelligent security robots. In places where there is insufficient police power, if there are intelligent security robots patrolling for 24 hours, they will have a good deterrent effect and reduce the occurrence of crimes. 03 Key Technologies and Challenges It can be foreseen that there are many interesting applications for service robots in the future, but there are still many challenges from the perspective of technology implementation. The three major technical areas of robotics were mentioned earlier, including perception, cognition, and behavioral control. Among them, behavior control technology has relatively many technical accumulations. For service robots that can be practically used on a large scale in the near future, they will be more inclined to develop low-cost robots. For example, a wheeled robot with a relatively simple mechanical structure (only a wheeled mobile platform plus a movable head and generally no arm or arm) is relatively simple, so the technology in perception and cognition appears to be relatively simple. Particularly important. Based on previous research progress in the field of robotics and preliminary analysis of applications, the following perception and cognitive technologies can be considered as the key to application. 1. Three-dimensional navigation and positioning technology. Regardless of the robot, as long as it can be moved, it needs to be navigated in the home or other environment. SLAM (Simultaneous Localization and Mapping) technology can simultaneously locate and build maps, and there have been many technical accumulations in academic research. However, for the actual system, due to the requirements of real-time low cost (such as the inability to use more expensive radar equipment) and the dynamic changes in the home environment (positioning of items), higher requirements are placed on the navigation and positioning technology, and further research and development is still required. . 2. Visual perception technology. These include related technologies such as face recognition, gesture recognition, object recognition, and emotion recognition. Visual perception technology is a very important technology for robot and human interaction. 3. Language interaction technology. It includes speech recognition, speech generation, natural language understanding and intelligent dialogue systems. 4. Text recognition technology. There is a lot of text information in life, such as the label information of books and objects, which also requires the robot to be able to recognize characters through the camera. Compared with the traditional post-scan text recognition, the text can be recognized by the camera. 5. Cognitive technology. Robots need to gradually realize cognitive functions such as planning, reasoning, memory, learning, and forecasting, thereby becoming more intelligent. From the current research status, the key technologies facing service robots have made great progress, but there are still quite a few problems to be solved. 04 Combination of R&D and Practical Application of Promoting Technology An important research direction of Intel China Research Institute is serving robots. Its research scope mainly includes robust three-dimensional navigation and positioning, visual perception of people and the environment, and human-robot interaction (including the combination of perception, cognition, and behavior control). Committed to providing Intel's partners with advanced service robotics. The research method starts with a specific target application field and promotes the research and development of technology and testing in practical applications. The above-mentioned helper applications, children's educational applications, and nursing care applications for the elderly are the research priorities. For application testing, Intel China Institute established a prototype of the actual robot hardware. Figure 3 shows the prototype of the tablet robot. On the left is its internal structure. A computer with an Intel Realsense camera controls the rotation of an omnidirectional mobile chassis and head. Its height is adjustable to accommodate the interactive requirements of adults and children. At present, the institute has achieved phased results in areas such as visual perception. As shown in Figure 4, it shows the use of Realsense camera and SLAM technology to generate a home environment map, which can be used to position navigation afterwards. The Institute has also developed a robust tracking technology that can accurately track a specific user for a long time in the presence of multiple people. In the future, more related technologies and applications will be developed. These technologies will also serve industry partners to help develop advanced service robot products. 05 Combination of R&D and practical application of push technology For robotics, one of the concerns is whether it requires the use of dedicated artificial intelligence chips. One direction currently explored is called hybrid computing, which refers to the use of general-purpose processors and other architectures to work together to perform calculations. However, it is still in the early stages of exploration. Whether an architecture can be successfully applied to service robots depends on whether the performance meets the requirements of the application, whether the power consumption is appropriate or not, which is a balanced choice considering multiple factors. It can be said that today's general-purpose processors and some other existing computing units (such as vector processing units, GPUs) can provide an ideal hardware architecture for service robots, especially multi-functional service robots with perception and cognitive functions. . At present, the main bottleneck of robots is still the development of application requirements and perception and cognitive technologies. Of course, the future architecture will continue to innovate as applications are further developed. Hardware design faces more and more challenges and needs to continuously adapt to the evolution of future development. At the same time, you can also consider using FPGA architecture to provide acceleration for some more specialized applications, so that when the application changes, there will be more flexibility.