Best LiDAR Sensors for Quadruped Robots: Spot, Unitree Go2, and Custom Builds Compared

Quadruped robots carry a constraint that most other mobile platforms don't: every gram and every watt counts. A wheeled AGV can bolt on a 1.2 kg sensor without thinking twice. A robot dog carrying a 3 kg payload budget doesn't have that luxury — the LiDAR sitting on its back directly affects how high it can jump, how long the battery lasts, and whether the gait controller stays stable.

This guide breaks down the LiDAR options actually being used on quadruped platforms today — what works, what's overkill, and what the trade-offs look like when you're building your own.

Why Quadrupeds Need LiDAR (and Why Cameras Alone Fall Short)

Legged robots operate in environments wheeled robots can't touch: staircases, rubble, collapsed buildings, dense forests. Navigation in these spaces requires three things cameras struggle to deliver simultaneously — range measurement independent of lighting, centimeter-level geometric precision in 3D, and no sensitivity to reflective or repetitive surfaces.

Cameras handle texture-rich, well-lit spaces well. Throw a quadruped into a dimly lit industrial plant at midnight with polished concrete floors and the visual SLAM pipeline starts drifting. LiDAR doesn't care about lighting conditions, surface reflectivity (within reason), or visual texture. It returns a direct distance measurement to every point it hits.

That said, most production quadruped systems run sensor fusion — LiDAR for geometry and mapping, cameras for object recognition and semantic understanding. The question isn't "LiDAR or camera?" It's "which LiDAR fits the platform, and what can the payload budget actually support?"

Commercial Quadruped Platforms and Their LiDAR Configurations

Boston Dynamics Spot — Velodyne VLP-16

Spot's Enhanced Autonomy Payload (EAP-2) ships with a Velodyne VLP-16 Puck, a 16-channel mechanical spinning LiDAR. The specs are well documented: 100 m range, ±3 cm accuracy, 30° vertical FOV, 360° horizontal. It works, and Spot has navigated nuclear plants, offshore oil rigs, and construction sites with it.

Two things worth noting. The VLP-16 weighs 830 g and draws around 8 W. On a robot that costs $75,000+ and carries 14 kg of payload, those numbers are irrelevant. But if you're sourcing the VLP-16 for a smaller quadruped, that weight and power draw become real constraints.

The 30° vertical FOV is also a limitation on a legged platform. A robot dog standing on a slope or climbing stairs needs to see the ground immediately in front of its feet as well as obstacles at eye level. With only +15° to -15° vertical coverage, the VLP-16 misses the near-ground zone unless tilted — and tilting a 360° sensor creates blind spots behind the robot.

Unitree Go2 — Stock L1 or Aftermarket 3D LiDAR

The standard Unitree Go2 ships with Unitree's own L1, a 4D LiDAR covering 360° × 90°. It's lightweight and cheap (included in the base price), but it's a proximity sensor more than a mapping sensor. Detection range tops out around 10 m with limited angular resolution — fine for avoiding walls in a living room, not sufficient for SLAM-based navigation in unknown environments.

This is where the Edu variant matters. Unitree sells the Go2 Edu specifically as a development platform, and resellers offer it bundled with two LiDAR upgrades:

• Livox Mid-360: 360° × 59° FOV, 40 m detection range (at 10% reflectivity), 200 kHz point rate, 265 g, ~6.5 W. This is the sweet spot for research — enough range for indoor and light outdoor SLAM, fits the Go2's ~3 kg payload with room to spare, and has mature ROS2 drivers.
• Hesai XT16: 16-channel mechanical LiDAR, 120 m range, ±1 cm accuracy, 30° vertical FOV, roughly 920 g and 12 W. Longer range than the Mid-360, but heavier and power-hungrier. Better suited for large outdoor mapping than indoor research.

The Go2 Edu Plus (100 TOPS computing) with a Mid-360 bundle typically runs around $4,000-$5,000 from resellers like RoboStore, Futurology Tech, and DyniBot. That's a complete autonomous navigation research platform — quadruped, compute, and 3D LiDAR — for less than the price of a single industrial Velodyne sensor.

DEEP Robotics X30 — Four Mid-360s

DEEP Robotics took a different approach with the X30. Rather than a single LiDAR trying to cover everything, they mounted four Livox Mid-360 units around the body — front, rear, and both sides. The result is true omnidirectional 3D perception with no blind spots.

Four sensors means four times the data bandwidth, four times the power draw (~26 W total for the LiDARs alone), and significantly more compute needed for SLAM. The X30 is an industrial-grade platform priced well above consumer quadrupeds, and the multi-LiDAR configuration reflects that. It's designed for scenarios where the robot needs to detect obstacles approaching from any direction — underground mining, disaster response, perimeter patrol.

For a custom build, the four-LiDAR approach is expensive but addresses a genuine problem. A single rear-mounted 360° LiDAR misses obstacles directly under the robot's body and has degraded coverage on steep descents. Adding a second sensor facing forward or downward fills those gaps without going to the extreme of four units.

Xiaomi CyberDog 2 — Integrated YDLIDAR TG30

The CyberDog 2 uses a YDLIDAR TG30 at the neck position — a single-point d-ToF (direct time-of-flight) LiDAR using a silicon photomultiplier. It's small and low-power, designed for close-range obstacle avoidance rather than mapping. The TG30 is not user-replaceable, and at roughly $1,300 the CyberDog 2 is positioned as a consumer gadget, not a research platform.

If you're looking at the CyberDog 2 for LiDAR-based SLAM work, you'll be replacing the stock sensor. The platform itself runs ROS2 on an NVIDIA Jetson Xavier NX, which is decent compute — but the chassis isn't designed to carry significant additional payload.

LiDAR Specs That Actually Matter on a Quadruped

When you filter the spec sheet through a quadruped's constraints, the priority order shifts compared to AGV or handheld scanning applications.

Weight and Payload Budget

On a small quadruped like the Go2, the difference between 265 g and 408 g is 143 g — roughly the weight of a small servo. Not trivial when you're also carrying a battery, compute module, and possibly a camera. On Spot's 14 kg payload, neither number matters.

Power Draw and Battery Life

Power matters more than most people expect. A quadruped's legs are already the biggest power consumer on the platform. Adding a sensor that draws 12 W (like the Hesai XT16) directly cuts into runtime.

The M360's <4.5 W draw and 12-32 V input range are relevant here. A custom quadruped running a 24 V battery system (common for larger BLDC motor setups) can power the M360 directly without a voltage converter. The Mid-360's tighter voltage range might require additional regulation.

Vertical FOV and Ground Coverage

This is the spec that kills more quadruped LiDAR setups than any other. A robot climbing a 30° slope needs to see the ground at its feet. A sensor with ±15° vertical coverage tilted level will see nothing closer than about 1.5 m on that slope.

The M360's asymmetric vertical FOV (-10° to +60°) is well-suited for quadruped mounting. Most of the sensing range points upward and forward where obstacles live, while the -10° lower bound still provides some near-ground coverage. Tilted slightly forward, it can see obstacles on stairs and slopes without losing rear coverage.

The Mid-360's 59° is usable but leaves less margin. The 30° FOV on the VLP-16 and XT16 is the real problem — you're forced to choose between seeing the ground or seeing obstacles at eye level.

IP Rating and Outdoor Use

Quadrupeds go outside. Rain, dust, mud splatter — the sensor needs to survive.

Both the M360 and Mid-360 carry IP67 ratings. For a robot operating in construction sites, agricultural fields, or disaster zones, this matters. No need for an enclosure or rain cover — just mount and go.

Custom Build: LiDAR Options by Budget Tier

Not everyone starts with a $4,000 Go2 Edu. Here's a realistic breakdown of what's achievable at different budget levels.

Under $1,500 — Entry Level

Start with a Unitree Go1 (discontinued but still available second-hand) or similar budget quadruped. Pair it with an RPLidar A1 or A3 — a 2D LiDAR scanning a single horizontal plane. Cost: ~$500 for the robot, $100-$300 for the LiDAR.

You get basic obstacle avoidance and 2D SLAM. No 3D mapping, no stair detection, but enough to learn ROS2 navigation stacks on a real walking platform. This is the educational entry point.

$2,000-$4,000 — Research Standard

Unitree Go2 Edu + Livox Mid-360. This is the configuration most university robotics labs are running. The bundle gives you a walking robot with 3D perception, 100 TOPS of onboard compute, and a well-documented ROS2 integration path.

Total cost with accessories (cables, mounting bracket, battery): $3,500-$4,500 depending on the reseller.

An alternative at this tier: a custom quadruped frame (open-source designs like Stanford DogGo or MIT Mini Cheetah derivatives) paired with the M360. At 408 g and <4.5 W, the M360 fits within tighter payload budgets while offering the same 200 kHz point rate and a wider 70° vertical FOV than the Mid-360. The built-in 6-axis IMU is a bonus — quadruped motion produces significant vibration and acceleration that an external IMU would need to handle separately. With the M360, that IMU data arrives time-synchronized with the point cloud through PTP (IEEE 1588-2008), which simplifies motion compensation in SLAM.

$5,000-$15,000 — Industrial

Custom or semi-custom quadruped with dual LiDAR setup: one 360° unit for horizontal mapping and navigation, one forward-facing unit for detailed obstacle detection. At this budget, you're looking at platforms like the DEEP Robotics Lite3 or custom builds using industrial-grade actuators.

$15,000+ — Multi-Sensor Fusion

The DEEP Robotics X30 approach. Multiple LiDAR units, thermal cameras, RGB cameras, RTK-GNSS. This tier is for industrial inspection, mining, and defense applications where no single sensor covers all requirements.

Where to Mount the LiDAR on a Quadruped

Mounting position matters more than most people realize. The same sensor produces very different results depending on where it sits.

Back Mount (Default for Spot and Go2)

The rear body panel is the default location on most commercial quadrupeds. It's structurally rigid, relatively high off the ground (good forward visibility), and keeps the sensor away from kicking legs.

The blind spots: directly underneath the body and immediately behind the robot. For a quadruped backing up or navigating in reverse, this matters. On the X30's four-sensor setup, the rear unit specifically addresses this gap.

Front Mount (Head)

A front-mounted LiDAR sees what the robot is walking into — stairs, gaps, low-hanging obstacles. It's the intuitive choice for inspection and reconnaissance tasks. The downside: the head moves with the gait, introducing vibration and point cloud distortion. Mechanical isolation or software-based motion compensation (where the built-in IMU helps) becomes necessary.

Belly Mount

Under-body mounting provides excellent near-ground coverage — detecting curbs, steps, and small obstacles the back-mounted sensor misses. The challenge is ground debris. Rocks, mud, and vegetation can hit a sensor mounted only 10-15 cm off the ground. An IP67 rating helps, but physical impacts are still a risk.

Dual Mount (Back + Front or Back + Belly)

For custom builds, two sensors on a quadruped is usually the sweet spot. A back-mounted 360° unit handles general SLAM and mapping, while a secondary unit (either a smaller LiDAR or a depth camera) fills the specific blind spots of the primary sensor. Total additional weight: 400-600 g for two compact units.

Quick Recommendation Matrix

The Weight-Power-FOV Triangle

Picking a LiDAR for a quadruped always comes back to three competing constraints:

1. Weight — every gram reduces jump height and payload capacity
2. Power — every watt reduces operating time
3. Coverage — narrow FOV means blind spots, especially near the ground

The M360 sits in an interesting position here. It's 143 g heavier than the Mid-360, which is noticeable on a Go2. But it draws significantly less power (<4.5 W vs ~6.5 W), has a wider vertical FOV (70° vs 59°), and includes a built-in IMU that eliminates the need for a separate inertial sensor. On a custom build where you're designing the power system from scratch, the 12-32 V input range removes a DC-DC converter from the BOM.

For teams already in the Mid-360 ecosystem, switching isn't automatic — the ROS2 drivers, mounting brackets, and community support around the Mid-360 are established. But for new builds where the sensor choice isn't locked in yet, the M360's spec sheet is worth a close look, especially if the robot will operate outdoors or in environments where power efficiency matters.

What's Next

If you're evaluating LiDAR options for a quadruped project, start with two questions: what's the payload budget, and what's the operating environment? Those two constraints eliminate most options before you even look at range or point rate.

For a detailed sensor-to-sensor comparison including the M360's dual-echo mode and how it handles multi-sensor interference in close-quarters operation, see our M360 vs Mid-360 comparison.

Need help matching a LiDAR to your specific quadruped platform? Contact us — we've tested these sensors on walking robots and can point out the gotchas before you buy.

© 2026 SmartBotParts. All rights reserved.

Product specifications, weights, and pricing referenced in this article are based on publicly available data from manufacturers and authorized resellers as of July 2026. Actual performance may vary depending on mounting configuration, environmental conditions, and firmware version. SmartBotParts is not affiliated with Boston Dynamics, Unitree Robotics, DEEP Robotics, Xiaomi, Velodyne, Hesai, or Livox. All product names are trademarks of their respective owners. This article is provided for informational purposes only and does not constitute an endorsement or recommendation of any specific product.