Sensors & Perception for Robots: Market Size and the LiDAR Race

01The perception layer

Perception hardware splits into two very different jobs. "Seeing" — mapping the environment, detecting obstacles, localizing the robot in space — is dominated by LiDAR (light detection and ranging) and cameras, and it's the more mature, better-funded half of this segment, having spent the past decade being built out for the self-driving car industry before robotics ever needed it. "Feeling" — sensing grip force, contact, and slip through a robot's hands and feet — is the newer, smaller half, and it's being built specifically for humanoid robots rather than inherited from an adjacent industry, since no other market previously needed a five-fingered hand to sense the world through touch.

02LiDAR's pivot to robots

Modern spinning LiDAR traces back to 2005, when Velodyne's David Hall built a 64-laser sensor for the DARPA Grand Challenge autonomous-vehicle competition — the technology that made point-cloud 3D mapping practical for the first time. Hesai and RoboSense (originally Suteng Innovation Technology), both founded in 2014, built their businesses on the autonomous-vehicle LiDAR boom that followed, competing for years primarily on cost-per-unit for self-driving car and ADAS customers. Hesai listed on Nasdaq in February 2023; RoboSense listed on the Hong Kong Stock Exchange in January 2024. Velodyne itself, the pioneer of the category, merged with Ouster in early 2023 rather than surviving as a standalone automotive-LiDAR pure play — a sign of how brutal the price competition in that original market had become. As humanoid and quadruped robots emerged as a new, higher-margin customer base from around 2024 onward, both Hesai and RoboSense pivoted a meaningful share of their product roadmaps toward robotics-specific LiDAR, rather than treating it as a side business to automotive.

03LiDAR market size

MarketsandMarkets estimates the global LiDAR market (all applications — automotive, robotics, geospatial, and industrial combined) at $3.27 billion in 2025, growing to $12.79 billion by 2030, a 31.3% compound annual growth rate. That figure spans far more than robotics; no research firm was found publishing a market-size series isolating robotics-specific LiDAR revenue on its own, which is why the shipment-volume and market-share figures below — both robotics-specific — are the more directly relevant numbers for this segment.

Data note: Hesai and RoboSense both publicly claim the #1 spot in robotics LiDAR, but their claims use different category scopes and aren't directly comparable. Hesai's figure (239,273 units, +425.8% YoY) is specifically for humanoid and quadruped robots. RoboSense's figure (303,000 units) spans a broader category — humanoid robots, robotic mowers, and unmanned delivery vehicles — and RoboSense separately reports partnerships with roughly 50 humanoid/quadruped makers including Unitree, AgiBot, Galbot, ENGINEAI, and Dobot. Both figures come from the companies' own disclosures (Hesai's SEC filing, RoboSense's press release), not an independent third-party count, and shouldn't be summed. Market-size figures here cover LiDAR specifically, not the full sensor/perception stack — cameras, IMUs, and tactile/force-torque sensors (a separate, smaller sub-market covered in article prose) aren't included, since no single research firm was found publishing a consistent market-size series spanning all of them.

Market data as of 2026-07

04Key players

Hesai and RoboSense each publicly claim to be the top supplier of robotics LiDAR, and both claims are backed by real, if not directly comparable, numbers. Hesai reported 239,273 robotics LiDAR units shipped in 2025 for humanoid and quadruped robots specifically, up 425.8% from 45,503 units in 2024. RoboSense reported 303,000 robotics LiDAR units shipped in 2025 across a broader category spanning humanoid robots, robotic lawn mowers, and unmanned delivery vehicles, up 1,142% year-over-year, and separately disclosed partnerships with roughly 50 humanoid and quadruped makers including Unitree, AgiBot, Galbot, ENGINEAI, and Dobot. Because the two companies count different categories of robot, their figures shouldn't be summed or read as a settled ranking — RoboSense's total is larger, but Hesai's is the narrower, humanoid-and-quadruped-specific number. RoboSense's momentum continued into early 2026: the company reported 330,300 total LiDAR units shipped in the first quarter, with robotics exceeding 50% of its total shipments for the first time.

05Tactile sensing

Alongside LiDAR and cameras, humanoid robots need a sense their industrial-robot-arm predecessors mostly didn't: touch. A humanoid hand has to modulate grip force on an egg versus a wrench, and detect when an object is starting to slip, which requires tactile sensors across the fingertips and palm plus force-torque sensors at the wrist. Industry market-research estimates for these sub-markets are thinner and less independently verified than the LiDAR figures above, so they're worth reading as directional. One specialty market-research estimate sizes the humanoid-robot torque-sensor market at $515 million in 2024, growing to a projected $1.28 billion in 2025 and $7.87 billion by 2032 (a ~41% compound annual growth rate over that 2024–2032 span) — a much steeper curve than the roughly $168 million to $185 million to $361 million trajectory estimated for the humanoid-robot tactile-sensor market over the same years, reflecting how much earlier-stage tactile sensing still is relative to force-torque sensing.

The supplier base looks nothing like LiDAR's China-concentrated field. ATI Industrial Automation and FUTEK, both US-based, are established force-torque sensor makers extending into humanoid wrists from an industrial-automation base. Switzerland's Bota Systems has built specifically ROS2-calibrated force-torque sensors aimed at the cobot and humanoid wrist market. On the tactile side, Pressure Profile Systems (US) is an established resistive-tactile-sensing supplier, while Sensor Products Inc. (US) and Aodiwei (China) compete on lower-cost capacitive tactile sensors. In August 2025, China's XJCSENSOR unveiled a MEMS-based force-torque sensor at the World Robot Conference in Beijing — one signal that Chinese suppliers are starting to contest a category so far led by US and European specialists.

06Challenges ahead

On the LiDAR side, the open question is whether Hesai and RoboSense's rival claims converge into one consistent ranking as robotics becomes a larger share of both companies' revenue and reporting gets more standardized, or whether they keep drawing category boundaries that suit their own strongest number — a pattern this site has already seen in humanoid shipment counts and drone-market share. On the tactile side, the question is more basic: whether any single sensing technology (capacitive, resistive, or another approach) becomes a default the way LiDAR did for 3D mapping, or whether humanoid hands keep shipping with a patchwork of sensor types from different suppliers, the way the segment looks today.

07Timeline at a glance

08FAQ