Previously, we introduced the tactile sensors of humanoid robots. In fact, in addition to tactile sensors, robots require a total of four types of sensors: force sensors, tactile sensors, inertial measurement units, and visual sensors. Today, we will elaborate on these.
Force Sensors
Force sensors convert mechanical signals into electrical signals and are core to the robot’s perception and interaction. Six-dimensional force sensors can measure multi-dimensional forces/torques simultaneously and are primarily responsible for end-effector force control.
The current mainstream technologies are metal foil sensors (such as Honeywell TJE) and silicon strain sensors (such as ATI Gamma). Silicon strain sensors have advantages in sensitivity. Domestic company Kunwei Technology has achieved a crosstalk of <0.5% FS for six-dimensional force sensors, while Lingyun Co., Ltd. has collaborated with the Chinese Academy of Sciences to develop a low-cost solution.
Currently, ATI holds a market share of 22.4%. In the domestic market, Kunwei Technology has over 50% market share in collaborative robots, focusing on high-precision tactile sensors, with products covering a range of 30N-20KN. Clients include JAKA Robotics and Xiaomi. Lingyun Co., Ltd. has an annual production line of 3,000 units and is collaborating with the Hefei Institute of Physical Science to develop six-dimensional force sensors, aiming for sensor business revenue to account for 15% by 2024. Keli Sensor is investing in MEMS silicon-based strain gauges and strategically investing in Kepler Robotics, with other physical quantity sensor revenue expected to increase by 331% year-on-year in 2024.
Tactile Sensors
These are electronic skins that achieve pressure, temperature, and humidity perception through flexible tactile sensors, primarily focusing on bionic touch.
The current mainstream solution is resistive type, with Fulei New Materials using coating processes to keep costs low; Hanwei Technology has developed high-sensitivity capacitive sensors with faster response times, with delays as low as 5ms. International leaders Novasentis and Tekscan hold 57.1% of the global market share, while Hanwei Technology focuses on medical monitoring scenarios. Its subsidiary, Suzhou Nengsida, is mass-producing flexible resistive sensors with full orders, expecting sensor revenue of 384 million yuan in 2024, a 68% year-on-year increase. Fulei New Materials’ flexible sensors are applied to robot fingertips, with pilot production lines launched, expecting revenue of 2.98 billion yuan in 2024, a 19% year-on-year increase. Riying Electronics is migrating automotive electronic technology, developing automotive-grade resistive sensors, and plans to expand into the overseas robot market through its U.S. subsidiary.
Inertial Measurement Units
The Inertial Measurement Unit (IMU) perceives motion posture through accelerometers and gyroscopes, primarily used for balance control.
The current solution is MEMS technology, which offers miniaturization and low-cost advantages. There are also laser and fiber-optic inertial navigation solutions, which are highly accurate but very expensive, suitable for aerospace and other fields.
Currently, the high-end market is dominated by Honeywell, Northrop Grumman, and Safran, which together occupy nearly 70% of the market. Domestic company Huayi Technology plans to launch a humanoid robot-specific IMU (ARU8010) in 2025, supporting a dynamic range of 4000°/s. Its subsidiary, Huayi Intelligent Manufacturing, expects revenue of 22.76 million yuan in 2024, while Xindong Lianhe’s MEMS IMU is applied in consumer electronics, with R&D investment expected to exceed 20% in 2024.
Visual Sensors
Simply put, these are cameras and LiDAR used for environmental perception. Currently, 2D and 3D solutions are developing in parallel. For example, Tesla’s Optimus uses a 2D solution, utilizing multiple cameras and Transformer algorithms, while 3D solutions include Yushutech’s RealSense D435i and UBTECH’s LiDAR + RGBD fusion multimodal data.
In terms of market, international manufacturers Sony and ON Semiconductor dominate the image sensor market, while domestic manufacturers like Weir Group are gradually breaking through. Currently, Weir Group is the second largest global supplier of automotive CIS, planning to launch automotive-grade 3D ToF sensors in 2024. Additionally, Orbbec is a supplier for UBTECH’s Walker X, with 3D vision technology covering the service robot field. Hongsoft Technology’s AI vision algorithms empower robot target recognition, with clients including Tesla and Xiaomi.
Unit Cost
The cost of sensors for a single robot depends on functional requirements and configuration complexity. Below is a cost breakdown for a typical configuration (using humanoid robots as an example):
|
Sensor Type |
Quantity per Unit |
Cost per Unit (Yuan) |
Value per Unit (Yuan) |
Proportion of Total Sensor Cost |
|
Force Sensors |
28 (1D) 4 (6D) |
1D:500 6D:20,000 |
7,000 (1D) 80,000 (6D) |
87% |
|
Electronic Skin |
10 (fingertips/pads) |
400 |
4,000 |
4% |
|
IMU |
2-3 |
1,000 |
2,000-3,000 |
2-3% |
|
Visual Sensors |
3-4 (including LiDAR) |
Cameras:200 LiDAR:4,000 |
600-16,000 |
6-16% |
|
Total |
100,000-200,000 |
100% |
Force sensors account for over 80% of the single-unit sensor cost, making them a core expenditure item. The unit price of six-dimensional force sensors can reach 20,000 yuan each. The cost of visual sensors fluctuates significantly; if the LiDAR configuration is high, it can exceed tens of thousands, but if only cameras are used, the cost can be reduced to 600 yuan. The costs of IMU and electronic skin are relatively low, but there is potential for incremental growth with functional upgrades (such as multimodal perception).
Incremental Driving Factors
The mass production of humanoid robots is accelerating, with global shipments expected to reach 350,000 units by 2030 (according to GGII forecasts), driving large-scale procurement of sensors. As scale effects push down sensor prices (for example, the price of six-dimensional force sensors is expected to drop from 20,000 yuan each to 5,000 yuan each).
Technological iterations are driving the demand for multimodal perception. For instance, future interactions in complex scenarios, dexterous hand grasping, and dynamic balancing will require high-precision force feedback, which will increase the penetration rate of six-dimensional force sensors.
Additionally, as environmental adaptability requirements increase, electronic skin will need to enhance tactile sensitivity, IMUs will need to optimize motion control, and visual sensors will need to achieve navigation and obstacle avoidance.
According to institutional forecasts, the humanoid robot sensor market is expected to exceed 11.9 billion yuan by 2030, especially with a compound annual growth rate of 60% for force sensors from 2024 to 2030.
|
Sensor Type |
2024 Market Size (Billion Yuan) |
2030 Estimate (Billion Yuan) |
CAGR (2024-2030) |
Core Growth Logic |
|
Force Sensors |
7.9 |
79 |
60% |
Increased penetration rate of six-dimensional force sensors (55% → decrease in unit value) |
|
Electronic Skin |
0.5 |
5 |
17% |
Breakthroughs in flexible materials + expansion from dexterous hands to full-body coverage |
|
Inertial Measurement Units (IMU) |
1.1 |
11 |
15% |
MEMS technology replacing laser inertial navigation + acceleration of humanoid robot mass production |
|
Visual Sensors |
2.4 |
24 |
18% |
Increased penetration rate of 3D visual solutions (LiDAR + camera fusion) |
This article is compiled based on publicly available information and does not constitute investment advice.
The Investment Research Uncle focuses on uncovering reliable first-hand information in the industry.
Following is effortless; missing out on news you want to see would be a pity!
Regarding consumption, the latest research summary
The “shovel” in the AI computing power realm—server power supplies
Explosions imminent! Driven by policy, technology, and demand