Development trend of robot chips - accelerating innovation in the robot industry with powerful underlying chip capabilities

Enthusiast Network reported (Text/Li Ningyuan) Driven by the demand for digital transformation, the global industry continues to flourish. According to IFR data, the global robot market size will exceed US$50 billion in 2022, of which the market will reach US$19.5 billion, the service robot market will reach US$21.7 billion, and the special robot market will reach US$10 billion. It is expected that by 2024, the global robot market size will exceed US$65 billion.
Against the backdrop of favorable market conditions, the basic and cutting-edge technologies of global robots are also constantly improving. The development of robot technology now emphasizes the independent research and development and optimization of core components, the development of robot core soft platforms, and the integration of cutting-edge technologies.

This article takes a closer look at the development trend of the robot industry from a perspective closer to the upstream industry.

From 2D to 3D, hot and more powerful visual processing platforms

The popularity of machine vision is obvious to all. According to data from the China Machine Vision Industry Alliance, the market size of China's machine vision industry is also growing rapidly, reaching 14.4 billion yuan in 2020. With the continued growth of the market size of China's machine vision industry, the market size of the machine vision industry is expected to exceed 23 billion yuan in 2022.

The biggest trend comes from the breakthrough of 3D vision technology, which further promotes the application of vision technology in high-end robot scenarios. Traditional 2D machine vision is rapidly upgrading to 3D machine vision. At present, there are not many 3D processing solutions on the market. The more common method is to use a complete set of SoC or visual processing platform. Intel processor plus, equipped with OpenVINO toolkit and oneA, this visual platform supports array 3D cameras, processes three-dimensional position information, and carries oneAPI to optimize operators to increase speed and reduce computing delay. OpenVINO compensates for visual information through calculation. In this area, the competitiveness of domestic visual manufacturers has also gradually improved. Zhaoguan Electronics and visual chip products also occupy a considerable proportion in the domestic robot market.

Another trend is that after the integration of system and machine vision technologies, the entire signal processing process from receiving light to system output is independently completed. Relying on the improvement of visual processing SoC processing power, memory density and system integration, embedded vision will be more widely used in the field of robots.

+AI , promote robot innovation

5G and AI have been talked about in the robotics industry for many years. The RB6 robot platform released this year has allowed industry insiders to see the powerful strength of robots under the support of 5G and AI. The Qualcomm Robot RB6 platform combines the enhanced Qualcomm AI engine and 5G functions, supporting Sub-6GHz and millimeter wave bands in global mainstream, enterprise networks and private networks. The platform's flexible architecture can support evolving connectivity features through expansion cards, including support for 3GPP Release 15, 16, 17 and 18 features for the Qualcomm Robotics RB6 platform in the future through expansion cards. The platform brings top-level edge AI and processing capabilities through an enhanced Qualcomm AI engine, supporting 70 to 200 trillion operations per second.

Real-time interaction between the robot front end and the center With the support of 5G high bandwidth, many AI algorithms that originally needed to run locally can be ported to the backend cloud, greatly reducing the computing load of the front-end robot. While reducing the computing load, the robot can also add more to improve multi-dimensional perception capabilities; in addition, with the support of 5G, the huge data perceived by the robot can be transmitted back to the computing center in real time. With the huge computing power of the backend, it can provide faster, more accurate and more comprehensive AI analysis and improve flexibility.

Robots are developing towards M7 and heterogeneous architecture

MCU is the first category of robot core main control chips. Small robots value the cost and power consumption of MCU, general robots value the accuracy of MCU, and high-end robots need MCU to take into account both computing power and scalability. Last year, Electronics Fan Network had exchanges with many domestic MCU manufacturers on robot applications. From the information obtained, it can be seen that the layout of robot MCU by various manufacturers has a trend of developing towards M7 core or heterogeneous architecture.

For example, the MCU+MPU architecture chip is being planned in the layout of the robot industry, and some new IPs will also be developed to replace the role of FPGA in the servo field; Jiuhai Semiconductor adopts MCU+FPGA heterogeneous architecture, MCU is responsible for communication processing, system bus monitoring and temperature monitoring, human-machine interface driver and other functions, FPGA is responsible for logic processing and motion control and other functions.

M7 core MCU is also being laid out by many manufacturers to meet the diversified application needs of the high-end robot market. This trend is closely related to the current domestic robot market MCU competition pattern. At present, except for a small number of high-end robot markets, the competitiveness of domestic MCUs needs to be improved. In other robot markets, domestic chips have been in full competition with foreign chips. Domestic manufacturers are also laying out the M7 core to improve their competitiveness in the high-end robot market. With the advent of high-speed operation and control MCUs with M7 cores or heterogeneous architecture chips with M7 cores from various MCU manufacturers, the competition for MCUs in the high-end robot field will become more intense.

FPGA accelerates robot development with SoM hardware combination

FPGA has been used in robot controllers for many years. At present, FPGA performance is excessive for conventional robot servo drive and control performance. In order to fully explore the possibilities of FPGA in the robot industry, FPGA manufacturers are shifting from using FPGA to achieve local performance optimization (servo drive, machine vision) to achieve system architecture optimization.

In order to deeply apply FPGA in the robot industry, many FPGA manufacturers have introduced their products to the robot market in the form of SoM hardware combination to accelerate robot development and optimize the entire system architecture. For example, the long-famous Xilinx Kria SoM, the Kira SOM hardware combination provides low latency (fast calculation), determinism (predictable), real-time (punctual), security and high throughput of thorough hardware acceleration, bringing adaptive computing power to robots. The Kira SOM hardware combination is also specially adapted for adaptive computing for robot applications. This year, the ProMe SoM released by E-Fintech has also made better optimization and design in terms of power consumption, size, ecology and flexibility. It only needs to define the required interfaces on the motherboard according to its own needs. At present, FPGA manufacturers are very fond of entering the robot system in the form of System on Module, which not only realizes the local performance optimization of applications such as video processing, sensing, and machine vision, but also performs systematic optimization based on the overall architecture of the robot.

Written in the end

Semiconductor companies have accelerated the innovation and upgrading of robots and market development with their powerful underlying chip capabilities. Chips and underlying intelligent platforms have become the key links in enhancing the value of robots. Semiconductor companies around the world are bringing new changes to the robot industry by building higher-performance underlying software and hardware platforms for robots.