Why China’s robotic mowers could decide the LiDAR race

Ni Tao is IE’s columnist, giving exclusive insight into China’s technology and engineering ecosystem. His monthly Inside China column explores the issues that shape discussions and understanding about Chinese innovation, providing fresh perspectives not found elsewhere.

Forget cars. The most important testing ground for LiDAR right now isn’t on highways but on lawns.

Robotic mowers, a niche consumer product, are forcing LiDAR to evolve faster than automotive markets ever did. To work in a compact, low-cost mower, sensors must be smaller, cheaper, and more reliable, with mass-market readiness baked in. That pressure is pushing Chinese suppliers like Hesai and RoboSense into breakthroughs that are reshaping LiDAR itself.

What most engineers overlook is that the race to democratize LiDAR won’t be won on luxury vehicles. It’s consumer robotics, the humble lawn mower included, that is driving the leap from bulky, analog systems to sleek, digital solid-state LiDAR.

Exactly how that happens is the result of technological breakthroughs and engineering refinements. Conversely, it also offers a valuable lesson for LiDAR engineers looking to scale beyond automotive: how market demand for a new category can fast-track efficient R&D and bring mature, cost-effective products to market in remarkably short order.

Pushing the boundary

Robotic mowers aren’t a gadget fad — they’re already a fast-growing global market. According to Mordor Intelligence, the global market grew from $1.5 billion in 2021 to $2.6 billion in 2024, with a CAGR of 20 percent. Fortune Business Insights projects the market will reach $4.04 billion by 2028.

Until 2023, two main approaches dominated: boundary-wire and boundary-free systems.

Boundary-wire systems defined the industry’s early phase. A thin electromagnetic wire is buried around the lawn perimeter, carrying a weak current to create a magnetic field. When the mower detects the wire, it changes direction, staying within the designated “geo-fence.” But installing the wires is cumbersome and prone to wear and tear.

Soon, boundary-free mowing took hold, powered by new technologies like pure vision, UWB, and RTK—each with merits and drawbacks. Pure vision uses high-definition cameras to perceive and navigate, but performance falters in poor lighting conditions, and positioning is often lost.

UWB (ultra-wideband) is a short-range wireless technology that uses Time Difference of Arrival (TDoA) or Time of Flight (ToF) methods to transmit low-power, high-bandwidth data for precise tracking and instant device-to-device communication. With at least three base stations, a mower can determine its coordinates to centimeter-level accuracy, reducing overlap and improving mowing efficiency.

Paired with sensors like cameras, UWB helps the robot to avoid obstacles such as trees, steps, and fences. Yet it requires at least three signal poles, adding cost, and frequent battery replacement causes inconvenience for users. Signal blockage also significantly undermines positioning accuracy.

Meanwhile, RTK (real-time kinematics) employs an antenna to capture signals from GPS/GNSS/BeiDou satellites, supplemented by base-station differential data. This enables the robot to figure out its precise navigation and obstacle avoidance coordinates, with an accuracy down to 2 centimeters.

However, poor weather can interfere with signals, prompting many mower manufacturers to add sensors such as ultrasonic radar and an inertial measurement unit. Even so, RTK could ultimately be overtaken by LiDAR-led solutions.

From mechanical to digital

For autonomous driving enthusiasts, LiDAR is nothing new. From clunky mechanical systems, LiDAR has evolved into semi-solid state and now into its most advanced form: digital solid-state LiDAR.

In essence, digital LiDAR addresses the biggest shortcoming of traditional designs: low digital conversion efficiency, poor signal-to-noise ratio, and unwieldy complexity.

The result: simplified signal-processing chains, stronger anti-interference, improved signal integrity, and smaller, lighter devices as a result of eliminating redundant mechanical parts.

Traditional analog LiDAR depends on sensors like avalanche (APDs), PIN photodiodes, or SiPMs (silicon photomultipliers) to convert received light into analog signals. These faint signals require amplification and filtering before being digitized by an analog-to-digital converter (ADC) for processing.

Analog designs also rely upon mechanical rotation for scanning—one reason early models were bulky, costly, and less reliable.

By contrast, digital LiDAR is solid-state. Beam steering is handled by optical phased arrays (OPAs) integrated on silicon chips. Designed to rapidly and electronically steer laser beams without mechanical parts, this crucial optical device controls light direction via phase differences between adjacent waveguides. Fast and precise, OPAs are promising for both optical communication and LiDAR.

Another leap: digital LiDAR bypasses analog-to-digital conversion. Incoming light generates digital signals directly, which are transmitted and processed into point clouds.

The single-photon avalanche diode (SPAD), a solid-state photodetector, is at the heart of this innovation. Sensitive enough to trigger a digital pulse from even a single photon, SPAD shifts much of the processing into the digital domain.

The simplified architecture also enhances integration. Compatible with CMOS processes, SPAD arrays can be integrated alongside time-to-digital converters (TDCs) and digital logic into one or a few SoCs—slashing size, power consumption, and cost, while making mass production feasible.

LiDAR for robots

When San Francisco–based Ouster launched a solid-state LiDAR in 2020, the industry’s first, it sparked an industry-wide race to iterate and commercialize.

Chinese companies moved fast. Firms like Hesai, RoboSense, Huawei, and Innovusion have developed smaller, higher-channel, wider-field-of-view LiDAR that meets automotive standards. They then quickly branched into other segments, in particular robotics.

Beyond industrial and humanoid robots, consumer robots such as mowers are attracting attention thanks to their rapid expansion.

Two standout examples are Hesai’s JT series and RoboSense’s Airy. The Hesai JT weighs only 200 grams, with a minimum diameter of 55 millimeters—about 70 percent smaller than comparable products. It offers up to 256 channels, consumes a maximum of 4.3 watts, and has a 60-meter range. Each frame covers an area equivalent to 1.5 football fields.

RoboSense’s Airy, a close rival, delivers similar performance but is slightly outmatched by the JT. Weighing under 250 grams and measuring 60 millimeters across, the cylindrical Airy consumes less than 8 watts, while supporting up to 192 channels that reach as far as 60 meters.

Installed in robotic mowers, these two LiDARs deliver obvious benefits, such as compact size. Hesai’s JT, for example, protrudes just 30 millimeters from the mower body, barely affecting design.

By extension, space constraints are even tighter in consumer electronics such as smartphones and drones. Broader LiDAR deployment will drive down supply-chain costs while real-world usage feeds valuable data into R&D.

In May 2025, RoboSense struck a deal with yard robot startup Mammotion to supply 1.2 million automotive-grade solid-state LiDAR units for robotic mowers over three years. Partnerships like this help to dispel myths and skepticism surrounding LiDAR’s versatility.

Toward chip-based integration

In 2019, Elon Musk dismissed LiDAR as “a fool’s errand,” citing its high price—often more than $10,000 per unit. Today, Hesai’s automotive-grade LiDAR can cost under $200–250. Musk did not foresee the price collapse for LiDAR, or the technological leap: greater accuracy, stronger performance, and higher reliability.

Wider adoption across industries—from cars and robots to ports, drones, and agricultural machinery—has reinforced the case for LiDAR. But this is more a result than a cause. The real catalyst is technological progress. The shift from analog to digital marks a clear industry trend: chip-level integration.

Automotive applications have been pivotal. Early rooftop “watchtower” designs were unsightly and increased drag, spurring engineers to miniaturize LiDAR into sleeker, built-in systems that remain discreet while providing safety redundancy. This demands relentless refinement of transmitters, scanners, and receivers.

At RoboSense, engineers restructured LiDAR architecture, consolidating discrete components into single chips and reducing assembly costs. Its MX series leverages ASICs (application-specific integrated circuits) instead of FPGAs (field programmable gate arrays). ASICs are less flexible but cheaper to build at scale, bringing unit costs under $200 and enabling LiDAR adoption in vehicles priced at RMB 150,000–200,000 ($21,064-28,085), according to RoboSense’s CEO Qiu Chunchao.

That’s not to say Chinese LiDAR makers face no competition—quite the opposite. Overseas rivals often achieve breakthroughs earlier. In 2022, US sensor startup Aeva integrated all core LiDAR components into a single silicon photonics chip, further shrinking size and potentially revolutionizing autonomy. Its four-dimensional LiDAR even measures instantaneous velocity, allowing vehicles and robots to make safer, smarter decisions.

Chinese firms face myriad hurdles. Sony and onsemi still dominate SPAD chip technology, while silicon OPA scanning precision, a key process, needs to be improved. These are but some core technical challenges to tackle.

But at the end of the day, scale will prove decisive. As with almost all sectors, abundant engineering talent, supply-chain advantages, and innovative use cases have given Chinese LiDAR manufacturers a head start in the race to democratize the once-expensive tool. This in turn allows them to cut costs sharply and iterate products rapidly. In the intense global LiDAR race, this agility could tip the balance.

So while the grass may not always be greener on China’s lawns, robotic mowers are quietly cutting the path for LiDAR’s next era. And if engineers want to see where the technology is really heading, they should stop staring at the highway and start looking at the garden.