Wraparound:
When it comes to robots, one word that has been frequently mentioned lately is artificial intelligence. Artificial intelligence is a discipline that uses computers to achieve human-like intelligent behavior. Robots themselves are one of the ultimate targets of artificial intelligence. So when it comes to robots, it's easy to think of artificial intelligence. Artificial intelligence is really important for robots.
Today's robots are capable of performing precise, repetitive tasks, but they often don't have the flexibility to adapt themselves to new tasks or unfamiliar or uncertain situations. However, these things are changing, and robots are getting smarter. We can't help but ask, what are the key technologies that make robots smart? How far will these technologies evolve in the next decade? What are the application prospects for the entire industry?
With the increasing demand for household robots, the continuous progress of artificial intelligence-related technologies, and the growth of hardware performance, service robots have begun to move from the laboratory to the home in recent years, and have developed from single-function robots such as sweeping robots to multi-functional personal robots.
Perceptual, cognitive, and behavioral control
According to the common understanding, robotics is divided into three parts: perception, cognition, and behavioral control. The perception part is mainly based on visual, auditory and information processing of various sensors; The cognitive part is responsible for more advanced semantic processing, such as reasoning, planning, memory, learning, etc.; The behavior control part is specifically responsible for controlling the behavior of the robot.
When it comes to robots, one word that has been frequently mentioned lately is artificial intelligence. Artificial intelligence is a discipline that uses computers to achieve human-like intelligent behavior. The robot itself is one of the ultimate application targets of artificial intelligence. So when it comes to robots, it's easy to think of artificial intelligence. Artificial intelligence is really important for robots, and all three of the technologies mentioned above are related to artificial intelligence.
From the application level, robots have a certain degree of autonomy and can interact with people and the environment, and have higher requirements for intelligence than previous computing devices (including computers, mobile phones, etc.), which is one of the reasons why artificial intelligence has gradually received attention.
Traditional artificial intelligence, as a discipline, originated from the Dartmouth Conference in the 50s of the 20th century, and has accumulated rich achievements in basic theories and methods after several ups and downs. From the early symbolic computing system to the expert system, to the machine learning and big data analysis developed in the 90s of the 20th century, it can be regarded as the category of artificial intelligence. Some related applied research has been derived in the fields of image, speech, search, data mining, and social computing. Among them, computer vision, speech and natural language processing, agents, etc., which are more closely related to robots.
From a technical point of view, AI still has a long way to go before it reaches the human level of intelligence, because the mechanism of human intelligence has not been studied clearly. However, from a practical point of view, according to the current technological progress, if it can partially simulate human intelligent behavior (such as identifying the owner and interacting accordingly), and achieve a good user experience, a breakthrough will be achieved in the short term. Of course, this also needs to further solve the practicability and robustness of the technology in technology research and development.
After all, in the past, many robots were only developed and tested in laboratories or restricted environments (such as nursing homes), while emerging home service robots will complete certain services independently or semi-independently (through collaboration with people) in the home environment, which puts forward higher requirements for the robustness of the technology. Some core technologies, such as computer vision and speech recognition, are still being perfected and have not yet developed to full maturity. All of this determines that in order to realize a truly practical and intelligent home service robot system, quite in-depth research and development work is required.
A robot is more than just a bunch of machines and chips
In past research on service robots, a typical target application is that robots can do housework, which requires robots to be able to manipulate objects with their arms (grasp and release). Although there is a lot of research in these aspects, judging from the current technological progress, there are still many challenges in realizing robots to do housework in a few years.
1. At present, the cost of robots, especially humanoid robots, is too high, for example, a robotic arm costs tens of thousands of dollars, and a manipulator costs tens of thousands of dollars, and the cost of the entire robot has exceeded the affordability of ordinary families.
2. The flexibility of the robotic arm is still difficult to match with the human hand.
3. There are also security problems, once a robot with a steel torso enters the home, it may accidentally wave or stretch its arms due to a procedural error, causing great harm to flesh and blood people.
If a robot can't do housework for a while, then what's the use of it? This is a question that many people are concerned about. If it can't be applied, a robot is just a bunch of machines and chips. At present, in addition to robots that can achieve relatively single functions such as sweeping robots, many other uses of robots are still under development.
As a new type of smart device, what ordinary users are most concerned about is what robots can do for them? As mentioned earlier, it is still difficult for robots to realize applications such as housework, and it will take a long time to reach the level of human intelligence. Therefore, in terms of application, it is also necessary to design for the current situation of artificial intelligence technology, especially perception and cognitive technology, and make full use of the latest technology to achieve functions that seemed impossible in the past. It is expected that in the next ten years, the service robots that will appear on the market will have the following characteristics.
1. Robots will achieve low cost. As mentioned in the previous article, the cost of service robots must be low enough if they are to enter the field of professional or home services. The cost of a professional service robot can be relatively high, while the cost of a home service robot must be tightly controlled.
2. The robot will have multi-functional characteristics. At present, the service robots used in the home are mainly robots that can complete a single function, such as sweeping robots. In the future, there will be greater development of multi-functional robots. Some of the processors in the robot will be used for a variety of applications due to cost considerations. At the same time, with the development of technology, the application of robots will be more and more, and robots will be able to complete different tasks like humans.
3. Unique app. This is the main motivation for users to purchase and use service bots. These applications are not simply ported from existing computers or mobile phones, but have characteristics unique to robotic applications, such as autonomy, human-like control, and the ability to interact with humans.
4. Able to cooperate with others to complete tasks. Since robots serve people, they need to interact with people in many ways to understand the owner's needs for services. On the other hand, due to the large gap between the intelligence of robots and humans, some tasks cannot be completed independently and can only be completed with the assistance of humans.
5. The robot will have a high level of security. This includes information security and physical security. That is, it does not pose a security risk to users, and can even proactively detect and block some behaviors that may pose security threats. The famous Asimov's Three Laws set out the basic rules for robot safety. Of course, how to implement these safety in concrete terms, especially physical security, requires efforts on robot hardware and software.
How home service robots and professional service robots will develop in the future has become a topic of concern in the industry.
Home service robots will achieve the following typical applications.
1. Home service robots will realize a variety of assistant applications. Popular personal assistant software on smartphones (Siri, Cortana, etc.) will evolve from virtual and invisible to what robots with a physical appearance can do. Imagine a robot assistant that can help users look up some information such as the weather, restrict license plate numbers, etc., or alert users. It doesn't sound like the same experience on a phone, but there's actually a lot of variation in the experience. When users interact with the bot, they can communicate with the bot in a more natural way, just like they are communicating with a human, and even see the bot's expressions, making the service more personalized. In addition to these applications, service robots can also act as human assistants in many more scenarios. For example, a service robot can act as a human fitness assistant (see Figure 1), and when people are exercising, the robot can provide some suggestions like a fitness trainer. It is believed that with the development of technology, such applications will become more and more abundant.
2. Home service robots help children's education. Due to the autonomy and mobility of service robots, they will bring infinite possibilities and broader application space for children's education.
3. Home service robots will be able to provide care for the elderly. Many countries, including China, have entered or are about to enter an aging society. Elderly people in China generally want to live at home, but their children are busy with work, and even if they live together, they cannot always be around to take care of the elderly. If the home service robot can be used to remind the elderly to take medicine, or provide timely help in an emergency, it will greatly improve the quality of life of the elderly at home and make their children feel more at ease. Elderly care will be a popular application of future robots and has great social value. There are many things that robots can do, in addition to functions directly related to health assistance, they can also help the elderly read books and newspapers, have simple chats with the elderly, and carry out emotional care.
For professional service robots, their applications vary greatly depending on the professional application field.
1. Distribute goods in the logistics center. It is reported that this app has begun to be used by companies such as Amazon, and will continue to be upgraded in the future, such as automatic pick-up.
2. Provide services to customers at retail stores, counters, front desks, etc. These service robots can greet customers and answer basic questions. They can also help customers with bags, bring tea and water, and give them directions.
3. Robot photographer. Robot photographers can tirelessly shuttle between guests and help people take high-quality photos. Of course, this feature is also a new feature on home service robots.
4. Intelligent security robots. In some places where the police force is insufficient, if there are intelligent security robots patrolling 24 hours a day, it will play a good deterrent role and reduce the occurrence of crime.
Key technologies and challenges
It is foreseeable that there will be many interesting applications of service robots in the future, but there are still many challenges from the perspective of technical implementation.
The three major technical fields of robotics mentioned above include perception, cognition and behavior control, among which behavior control technology has accumulated more technologies. As for the service robots that can be truly applied on a large scale in the near future, they are more inclined to develop low-cost robots. For example, wheeled robots with relatively simple mechanical structures (only wheeled mobile platforms and movable heads, generally without arms or arms are relatively simple), because of their relatively simple control parts, perception and cognitive technology is particularly important. Based on the preliminary analysis of the previous research progress and application in the field of robotics, it can be concluded that the following perception and cognitive technologies will be the key to realize the application.
1. Three-dimensional navigation and positioning technology. No matter what kind of robot is mobile, it needs to be navigated and positioned in the home or other environment. Among them, SLAM (Simultaneous Localization and Mapping) technology can be used for localization and mapping at the same time, and a large number of technical accumulations have been achieved in academic research. However, for the actual system, due to the real-time and low-cost requirements (such as the impossibility of using more expensive radar equipment) and the dynamic changes of the home environment (the placement of items), higher requirements are put forward for navigation and positioning technology, and further research and development are still needed.
2. Visual perception technology. Including face recognition, gesture recognition, object recognition, emotion recognition and other related technologies. Visual perception technology is a very important technology for the interaction between robots and humans.
3. Language interaction technology. Including speech recognition, speech generation, natural language understanding, intelligent dialogue systems, etc.
4. Text recognition technology. There is a lot of text information in life, such as books, newspapers, and item label information, which also requires robots to be able to recognize text through cameras. Compared to the traditional text recognition after scanning, 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 prediction, and become more intelligent.
Judging from the current research status, great progress has been made in the key technologies faced by service robots, but there are still many problems to be solved.
Promote the combination of technology research and development and practical application
When it comes to robots, one of the questions that everyone is concerned about is whether they need a dedicated artificial intelligence chip.
One direction that is currently being explored is hybrid computing, which uses general-purpose processors and other architectures for computing, but it is still in the early stages of exploration.
The success of an architecture for a service robot depends on whether its performance meets the needs of the application and whether the power consumption is appropriate, which is a choice that needs to balance multiple factors.
It can be said that the current general-purpose processors and some other existing computing units (such as vector processing units, GPUs) can already provide an ideal hardware architecture for service robots, especially multifunctional service robots with perception and cognitive functions. At present, the main bottleneck of robots is still in the application requirements and the development of perceptual cognitive technology. Of course, the architecture of the future will continue to innovate as the application develops. Hardware design is increasingly challenging and needs to evolve in the future. At the same time, you can also consider using the FPGA architecture to provide acceleration for some more specialized applications, so that when the application changes, there will be more flexibility.