Professor Wang Tianmiao, a well-known expert in the robot industry: four major trends in the development of service robots

On June 30, the Ecovacs Robot Conference and Suzhou Bay Service Robot Industry Summit officially opened in Suzhou. Wang Tianmiao, a professor at Beihang University, was invited to participate in the robotics industry sub-forum and gave a wonderful report on the core technology and development trends of service robots.

As ultra-long cycle tracks, such as mobile Internet and smartphones (2005-2020), new energy vehicles and smart cars (2015-2030), digital economy and intelligent robots (2020-2050), etc., they have always been the focus of governments. The world, industry, academia, investment community, etc. have paid close attention to it, especially for China. As the market dividend and demographic dividend gradually weaken, the technological dividend has become the core element for the re-emergence of China's economy and the sustainable and rapid development of comprehensive national strength. Among them, the Artificial intelligence, intelligent robots, high-end manufacturing of new materials, carbon neutral new energy, biotechnology, etc. have become important engines for new industrial transformation and new economic development in the future.

Social development and cutting-edge multidisciplinary innovation are constantly stimulating the evolution and development of intelligent robots from technology to form.

Demand for large-scale industrial development and urban agglomeration: On the one hand, efficiency and quality drive, labor force decline and cost increase drive the development from the secondary industry to the tertiary industry, primary industry application, etc. At the same time, the Belt and Road Initiative has become a Important profit channels for my country's robotics and automated production line companies. On the other hand, the agglomeration of population and logistics in big cities, food, agricultural and sideline products, prepared vegetables and fresh food, garbage and sewage and environmental protection treatment, autonomous driving and intelligent transportation, intelligent energy management and energy storage and battery exchange, AIot and safety monitoring, dangerous disaster relief robots , as well as consulting, logistics, cleaning, hotel, exhibition, coffee and other robots have become urgently needed service and product robots.

The acceleration of the aging society and the needs of the new generation of entertainment, cultural, creative and sports: On the one hand, the demand for chat, escort, assistant, elderly care, rehabilitation, traditional Chinese medicine and other robots is also becoming more and more urgent, including digital chronic disease medical treatment and AI virtual robots, fitness rehabilitation and traditional Chinese medicine massage robots , barrier-free mobile robots, turning massage and toilet disposal robots, etc. According to incomplete statistics, 15% of people over 65 years old, 25% over 75 years old, and 45% over 85 years old need this service. On the other hand, robots for young people in technology, cultural and creative, entertainment, sports and other fields include virtual human agency and communication, human-machine hybrid intelligent robots, emotional companion robots, cooking robots, cleaning robots, VR personalized fitness robots, stem cells and beauty. Injection robots, entertainment dancing robots, etc.

Irreplaceable robots in special scenarios: On the one hand, interstellar exploration, precise treatment operations, and biological tissues are driven by advanced needs, including space exploration and immigration, brain interfaces and consciousness, surgical robots and vascular nanorobots, myogenic life tissues and organs, and health Happy biochemical technology, immortality of life and soul, etc. On the other hand, hazardous operations and local wars stimulate demand, including hazardous operations, emergency and disaster relief, research and development of drones, unmanned tanks, unmanned ships, intelligent weapon systems, robot warriors, etc.

Dynamic 1: Basic research on cutting-edge hot topics, especially new materials and rigid-flexible coupling soft robots, NLP and multi-modality, brain-computer interface and cognition, basic software and platforms, etc. are particularly critical, because basic original breakthroughs are expected to change Robot form, product functions and service model

1. Humanoid robot technology, life-like bodies, artificial muscles, artificial skin, myoelectric control, tissue organs, soft robots, etc.;

2. DNA nanorobots and new material micro-nano components, nanomaterials, micro-electromechanical, 3D printing, intelligent prostheses, micro-nano manufacturing and assembly, drive energy conversion, force feedback interaction, etc.;

3. Biological sensing technology, audio and visual touch sensors, edge AI computing, rigid-flex coupling, perception-driven integration, etc.;

4. Natural language understanding, emotion recognition and human-computer interaction technology, conversational intelligent interaction technology, emotional interaction, solitary chatting, children and elderly care, etc.;

5. Brain-computer interface and biomechanical integration technology, brain science, neural consciousness, myoelectric signal, knowledge map, cognitive recognition, machine reasoning, etc.;

6. Yuanverse virtual human and robot fusion technology, next-generation Internet, entertainment interaction, agents, contextual sensing, remote operation, etc.;

7. Composite robot technology "hand, foot, eye and brain" integration, consisting of a mobile platform, robotic arm, vision module, end effector, etc., integrating environmental perception, positioning and navigation, intelligent control, etc., unstructured environment recognition, multi-machine Collaboration, intelligent transportation, etc.;

8. Super software automation, robot operating system, soft robot, RPA, property, finance, accounting, government affairs and other automation, etc.;

9. Cloud service robotics, distributed cloud services, cloud processing centers, artificial intelligence and machine learning, interpretable artificial intelligence, remote rental services, remote teaching services, robots as a service RaaS, etc.;

10. Ethics, robots for good, employment, privacy, morality and law, etc.

Dynamic 2: Robot +, among which sensors and core components, high-frequency standardized commercial applications (indoor and outdoor logistics, cleaning, emotional care assistants, etc.), Raas and App software are particularly critical, because these are expected to exceed more than 10 million units per product or more Form a subscription business model

High value-added core components include AI vision, force and touch, RV, motor, AMR, design and application software, etc.; super software automation tools AIops, RPA, Raas and other vertical large models, including leasing, training, processing, and application development and other cloud service platforms Raas; medical robots; mobile composite robots for loading and unloading, operating logistics or cleaning; service robots for entertainment, catering, massage, moxibustion, escorting, etc.; for agriculture, construction, recycling, disassembly and assembly, energy, Unmanned systems such as the nuclear industry.

In terms of robot + commercial applications, there are currently some companies in my country's robot complete machines and core components that are emerging. It is expected to have broad application prospects in new energy, automated logistics, agricultural and sideline products and consumer goods, biotechnology, public services, housekeeping services, etc., showing a promising future. Segments are developing explosively.

The "14th Five-Year Plan" Robot Industry Development Plan mentions that the average annual growth rate of the robot industry's operating income during the 14th Five-Year Plan will exceed 20%, and the density of manufacturing robots will double. The application scenarios cover to G-end, to B-end, to C-end, etc. Multiple dimensions. Environmental standards, high-frequency space, and labor costs also make “machine substitution” a pain point in some scenarios.

Dynamic 3: Large model + robot, it is expected to integrate general large models with vertical large models of specific robot applications in embodied intelligent interactive, knowledge-based, and standard application scenarios, greatly improving the level of robot intelligence and making it more in-depth and broad. application

As we all know, general multi-modal, NLP, CV, interactive and other large AI models are innovating robot perception methods, environmental cognitive complexity, knowledge-based fusion decision-making and control, etc., and are expected to bring great improvements to the level of robot intelligence and a wide range of application fields. will be greatly improved, especially the integration of the general model of application scenarios of embodied intelligence, interactivity, knowledge, and standardization with the vertical model of professional robot applications, including science and education, assistants, nursing, elderly care, etc., as well as guided operations, Cleaning, logistics, etc. are expected to be the first to achieve breakthrough progress.

The Google DeepMind team demonstrated the robot AI intelligent agent "RoboCat" on June 20, 2023. It is "universal". RoboCat can learn to control the robotic arm to complete tasks such as "ringing", "building blocks" and "grabbing fruits". Verification The success rate of completing a new task has increased from 36% in the initial stage to 74%, enabling rapid learning capabilities and reducing the need for human supervised training. Google also successfully applied the PaLM language model to robots, increasing the success rate of robots executing commands from 61% to 74%. This model determines the probability of successful learning of a specific skill by multiplying two probabilities: one is task-grounding, which is the language description of current professional skills; the other is world-grounding, which is the accessible path of various skills and knowledge currently accumulated. feasibility.

Dynamic 4: Humanoid (bionic) robots are expected to form a unified form of single robot products, which is expected to stimulate the reconstruction and large-scale development of their robot AI chips, various sensors, and robot parts supply chains.

With the advent of the "Robot+" era, bionic robots are embracing the 100 billion track. With the aging of the population and the booming development of smart manufacturing, robots, artificial intelligence and cloud service big data are entering a disruptive development stage. Bionic robots are promoting another modular, intelligent and cloud service development path. Large-scale industrial development of intelligent robots. Among them, humanoid and quadruped robots will be the two most growing segments of bionic robots. According to optimistic estimates, if 3-5% of the global labor gap from 2030 to 2035 may be replaced by bionic humanoid robots, the demand for humanoid robots is expected to be about 1-3 million units, corresponding to a global market size that will exceed 260 billion yuan. China The market will exceed 65 billion yuan.

Bionic robots still focus on flexible motion stability and dexterous operation operability to overcome key technical difficulties. Different from traditional robots, in order to move flexibly and operate dexterously in unstructured environments, bionic and humanoid robots have more urgent needs for system stability and high-end core components. Among them, the key technical difficulties are high torque density drive units, intelligent In terms of technologies such as motion control, real-time perception of the environment, and human-computer interaction, the academic community is also actively exploring new smart materials, rigid-flexible coupled artificial muscles, artificial sensory skin, soft robots, etc.

"ChatGPT+Bionic Robot" allows robots to change from "similar in appearance" to "similar in spirit". Open AI invested in 1X Technologies humanoid robot company to officially enter the robot industry, explore the application of ChatGPT in the field of robots, explore multi-modal large language models, and promote the integration of humanoid robots in human-computer interaction text knowledge and application process knowledge in the operating environment. Iteratively learn cognitive models to solve the challenging problem of severe lag in the combination of basic end-end framework algorithms of software in the robotics industry and perceptual front-end AI edge computing.

Although humanoid robots have fatal weaknesses in terms of efficiency and energy, application and convenience, maintenance and price, we have to pay attention to the unexpected progress of Tesla's rapid iteration of humanoid robots. The reason is that Tesla has changed from its own Starting from the specific application scenarios of certain links in ultra-large-scale automobile manufacturing in Germany, China, Mexico, etc., humanoid robots have been completely redefined and designed, especially the mechanical structure, electronic drive, and 40 joint parts have been newly designed and even some are subversive. characteristics, including different output torques, output speeds, positioning accuracy, rotational stiffness, force sensing, self-locking, size, etc. These original innovative breakthroughs are expected to drive humanoid robots to universal "perception capabilities, interaction capabilities, and operation control capabilities" The development of professional vertical large-scale models of computing power models and applications has spawned the rapid development of its robot AI chips, various sensors, and robot parts supply chain reconstruction and scale. Only in this way can it be possible to grow from Tesla Robots, which now has more than 1 million US dollars, to Gradually iterate to reduce costs and approach the sales price of US$20,000. .

Finally, look at the development of history and social forms, analyze the future trend of multi-disciplinary disruptive technological innovations such as new materials, new energy, biology, and AI, focusing on the world's aging, urbanization, population changes, and networking, intelligence, Create new market demand on a large scale. In the next 10 years, the global service robot market will break through the trillions of market development space. There are still uncertainties. There are three major debates: First, the evolution path of form? Industrial, commercial, humanoid, large model, or divergent application; secondly, sustainable driving of commercial value? Operation, training, integration, complete machines, parts, platforms, etc., authorized IP, sales, leasing, services, subscriptions, etc., relevant collaborative policies of universities, private enterprises, state-owned enterprises, innovation, supply chain, capital, government, etc.; the three robot ethics ? How can robots do good? Also includes employment, privacy, morals, ethics, and corresponding legal issues.