Do I need 360° LiDAR, or is forward-facing solid-state enough?

If you're doing SLAM or need situational awareness around the whole robot, get 360°. If you only need forward obstacle avoidance on a platform that turns to look, a solid-state forward block (E1R class) is lighter, more rugged and often cheaper to integrate.

Is the range figure comparable between brands?

Only at the same reflectivity. A "70 m" spec at 80% reflectivity and a "40 m @10%" spec can describe the same sensor. Always normalise to the @10% number for dark real-world targets.

Can Chinese robotics LiDAR replace my safety laser scanner?

No. SLAM/navigation LiDAR and a certified safety-rated scanner are different devices with different certifications. Keep them separate and confirm what your local safety standard requires with the relevant authority.

Why is the same sensor sometimes half the price at street versus quoted?

Consumer-channel street pricing (single units, no support wrapper) differs from robotics-grade volume contracts that include validated firmware, warranty and support. The POA units aren't overpriced — you're buying the support layer.

What compute do I need behind it?

Budget for LiDAR-inertial odometry on a Jetson-class board or better; a 200k pts/s unit produces more data to register than a 64k pts/s one, so higher point rate is a downstream compute cost, not a free upgrade.

Will these ship with a ROS/ROS 2 driver?

The mainstream modules do, but verify version support and time-sync (PTP) for your ROS distro before buying a fleet — driver maturity varies more than the hardware specs.

LiDAR & Perception Modules from China for Robotics

Why robotics LiDAR is a different buy from automotive

Most LiDAR marketing is written for autonomous driving: 200 m range, ASIL-grade functional safety, roof-mount packaging. Almost none of that matters on a 25 kg quadruped or a warehouse AMR. A legged robot needs to see the ground 1–5 m ahead and the ceiling above it; an indoor AMR needs a clean 2D-ish slice for costmaps; a humanoid needs a light head sensor that won't wreck its mass budget. So the priorities invert. Coverage geometry, weight and power come first; maximum range comes last. This guide is organised that way.

Chinese makers dominate this specific corner because the volume is here — legged robots and AMRs ship in far greater numbers than robotaxi fleets. That volume is what pushed a usable 360° LiDAR under $1k.

The three form factors

Mechanical spinning. A motor rotates the emitter/receiver assembly for a full 360° horizontal sweep. Mature, well-understood, excellent horizontal coverage — the traditional choice for 2D/3D SLAM on rolling platforms. The trade-off is a moving part (bearing wear over thousands of hours) and, historically, cost and weight. Most classic AMR LiDARs are this type.

Solid-state / semi-solid. No spinning housing. Fully solid-state units (flash or SPAD-on-chip, e.g. optical-phased-array and MEMS-adjacent designs) steer the beam electronically or with a tiny internal mirror; "semi-solid" designs use a small oscillating mirror behind a fixed window. These give a forward-facing FOV rather than 360°, higher reliability (no big bearing), and a compact rugged package — ideal for fixed-mount obstacle avoidance on the front of an AMR or the chest of a humanoid. The Livox Mid/Avia family uses a non-repetitive "risley-prism" scan that behaves like a hybrid: mechanical internals, solid outer housing, and coverage that fills in over time.

Compact 360° modules for legged robots / AMRs. The category that changed the market: small, light, low-power pucks that still deliver a full 360° horizontal sweep plus a wide vertical FOV, priced for volume robotics. The Livox Mid-360 and Unitree L2 are the reference points here — both sit in the sub-$1k class and are small enough to mount on a dog's back or a humanoid's shoulder without dominating the power budget.

The specs that decide fit

Read a robotics LiDAR datasheet in this order:

FOV / coverage geometry. Horizontal and vertical. 360° horizontal matters for SLAM and situational awareness; a wide vertical FOV (roughly 60–96° across current pucks) is what lets a legged robot see both the step in front and the doorway above. A 120° × 90° forward block is fine for obstacle avoidance but won't map behind you.
Range — and at what reflectivity. Range figures are quoted at a target reflectivity (commonly @10% for dark objects, @80% for bright). Always compare like-for-like: a "70 m" number at 80% and a "40 m" number at 10% can be the same sensor. For indoor AMRs and legged robots, 30–40 m @10% is plenty.
Point rate. Points per second sets how fast your map densifies and how much detail you get per frame. Higher is better for mapping fidelity but costs compute downstream — a Jetson-class board has to ingest and register every point.
Size / weight / power. The spec that quietly kills projects. On a battery-powered legged robot, single-digit versus double-digit watts, and 100 g versus near-1 kg, change your runtime and your gait. Confirm operating voltage matches your platform bus.
Interface & driver. Ethernet (with PTP time-sync) dominates; some units offer a native ROS/ROS 2 driver and a point-cloud SDK. The documentation test applies here as with actuators — buy one, integrate it, and confirm the ROS driver and time-sync work before you commit to a fleet. A sensor without a clean driver is a month of your life.
Ingress & environment. Outdoor inspection robots need IP-rated, temperature-qualified units; indoor AMRs can relax this.
Price band. Robotics LiDAR now spans <$1k for volume pucks up to POA for robotics-grade solid-state supplied on volume contracts.

Comparison: three representative Chinese modules

Figures below are from supplier datasheets; verify current specs and street pricing at RFQ.

Model	Type	FOV (H × V)	Range	Point rate	Price band (single unit)	Lead
Livox Mid-360	Compact 360° (prism-scan hybrid)	360° × 59°	40 m @10% (70 m @80%)	200k pts/s	<$1k (≈$750 street)	1–2 wks
Unitree L2	Compact 360°	360° × 96°	30 m (±2 cm)	64k pts/s effective	<$1k (confirm street price via RFQ)	1–2 wks
RoboSense E1R	Fully solid-state (SPAD-SoC)	120° × 90°	30 m @10% (max 75 m)	(via RFQ)	POA	2–4 wks

Read it this way: the Mid-360 buys you range and point density for mapping-heavy SLAM; the L2 trades point rate for the widest vertical FOV (96°), which suits a legged robot that must see high and low from a low mount; the E1R is a forward-facing solid-state block for rugged fixed-mount obstacle avoidance where you don't need to see behind the robot. RoboSense also supplies mechanical spinning and mechanical-hybrid families across a broader catalog — the E1R is just the fully-solid-state entry.

Matching the sensor to the platform

Quadrupeds (SLAM + terrain). A back-mounted compact 360° puck (Mid-360 / L2 class) is the near-universal answer — full horizontal coverage, wide vertical FOV to catch both ground and overhead, light enough not to eat the payload. Most off-the-shelf quadrupeds from China either ship with or bolt straight onto this class of sensor.

Humanoids (head/chest perception). Weight and power are brutal here. A compact 360° unit at the shoulder or a forward solid-state block on the chest is typical, often fused with a depth camera for close manipulation. Range matters less than mass and a clean time-synced feed into the perception stack. See humanoid platforms for what integrators are mounting.

Indoor AMRs (navigation + safety). A 360° unit for the costmap, frequently paired with a forward solid-state sensor for obstacle avoidance and a certified safety scanner where standards require it (LiDAR SLAM does not replace a rated safety laser scanner — that is a separate, certified device). Confirm what your safety standard demands.

Outdoor / inspection. Prioritise IP rating and thermal spec; expect POA robotics-grade units rather than the sub-$1k consumer-adjacent pucks.

In every case the sensor is one input to a stack that usually also includes IMU, wheel/leg odometry and often a depth camera. LiDAR-inertial odometry (LIO) is the common backbone; budget compute for it.

What Livox changed

Before roughly 2020, a robotics-usable 3D LiDAR was a five-figure line item, and 360° units were heavy spinning boxes. Livox's prism-scan approach and, later, the Mid-360, pushed a genuinely robotics-grade sensor under $1k at single-unit street pricing. That price move is why SLAM demos that used to run on five-figure sensors now ship on hobby-budget quadrupeds, and why RoboSense, Unitree and others now compete hard in the same band. The compression is the same volume story seen in joint modules and actuators: domestic robot volume drives the component price down, which drives more robot volume.

Sourcing notes

MOQ is 1 across these modules — you can buy a single unit to evaluate, which you should, before any fleet commitment.
Lead times are short (1–4 weeks) because these are catalog parts, not configured builds.
Prices are EXW/street indicative as of 2026. Landed cost adds freight, duty and VAT; compare on landed cost, not sticker. See the import playbook.
Confirm the driver and time-sync before volume. The datasheet number is easy; a clean ROS 2 driver, PTP sync and a maintained SDK are what actually determine integration time.
Send the platform, mount point and target FOV/range via RFQ and we'll return matched options with landed-cost framing.