Blogs

Benewake Unveils the NOVA20 Solid-State Near-Field Linear LiDAR: Eliminating Robotic Close-Proximity Perception Blind Spots with a 20-Point Array

Recently, Benewake officially launched the NOVA20, a solid-state near-field linear LiDAR designed specifically for embodied AI and mobile robots. Featuring 20 detection channels that form a 40° fan-shaped field of view, the NOVA20 covers the close-proximity zone from 5 centimeters to 1 meter in front of the robot. Working in tandem with 360° rotating LiDARs responsible for long-range scanning, it provides robots with a low-power, glare-resistant, and flexible near-field safety perception solution that can be distributed along the robot's body.

▲ Benewake NOVA20 solid-state near-field linear LiDAR (right), comparable in size to a beer bottle cap

Adopting a pure solid-state design with no moving parts, the NOVA20 is roughly the size of a beer bottle cap (25 × 22 × 13 mm typical), weighs less than 5 grams, consumes an average power of under 1 watt, and operates reliably under 100 Klux of intense ambient light.

The Dilemma of Embodied Robots: "Seeing the Distance, Missing the Meter"

As embodied AI accelerates toward commercial deployment, a fundamental issue has become increasingly prominent: the "close-range safety zone" within 1 meter in front of the robot remains a blind spot in existing sensing setups. While 360° rotating LiDARs handle long-range scanning, they leave a blind zone right next to the robot. Meanwhile, head-mounted cameras suffer from a short-range blind lower limit and are easily occluded by the robot's own limbs. Traditional solutions also fall short: ultrasonic and infrared proximity switches offer too few measurement points with large blind spots; triangulation-based linear lasers fail completely under outdoor bright light; and common 3D array solutions struggle with high-reflectivity crosstalk, bulkiness, and high costs.

▲ Stair edges, side obstacles, and other close-range hazards often fall outside the field of view of primary head-mounted cameras

This is not a niche problem. According to IDC data, global shipments of humanoid robots reached 18,000 units in 2025, and TrendForce projects global shipments to exceed 50,000 units in 2026. As robots enter real-world environments such as factories and commercial facilities at scale, physical guardrails and safety supervisors cannot always be present. Promptly detecting people and obstacles and slowing down or stopping before physical contact becomes an essential capability.

Furthermore, safety should not rely solely on high-level models prone to misinterpretation. Even the most advanced end-to-end models cannot deduce spaces occluded by the robot's own body. Continuously offloading high-frequency judgments—such as whether there is an object nearby and how far it is—to models drains the robot's limited computing power, battery, and thermal space. The industry urgently needs a deterministic ranging bypass independent of high-level models that can be independently verified: models handle understanding and planning, while underlying control systems rely on this data to slow down, detour, or stop. The updated ISO 10218 series of industrial robot safety standards (2025 edition) has incorporated collaborative safety capabilities such as "Speed and Separation Monitoring (SSM)", and the upcoming ISO 25785-1 standard for dynamically stable mobile robots is also under development—using deterministic ranging to guard human-robot separation is becoming a clear industry direction.

Designed precisely to bridge this gap, the NOVA20 integrates solid-state device architecture and array design into a systematic near-field safety solution, combining five major characteristics into a single LiDAR: pure solid-state architecture, linear array detection, compact size, high cost-effectiveness, and outdoor usability.

▲ 360° rotating LiDAR looks far, while the NOVA20 clears the blind spots right ahead: long-range and near-field sensors perform their respective duties

Highlight 1: A 20-Point Array Opening a 40° Fan-Shaped Sector, Clearing the Near-Field with a Single Module

The NOVA20 deploys 20 detection channels within a 40° field of view (evenly distributed across ±20°), measuring distances collaboratively at a 10 Hz frame rate with centimeter-level precision. Unlike single-point solutions that only register a sudden change, 20 channels can simultaneously observe distance variations across different directions within a fan-shaped zone. For instance, when a robot approaches the edge of a step, readings in a specific direction suddenly increase, enabling the system to react rapidly and providing more comprehensive near-field data for fall prevention and lateral collision avoidance.

Near-field detection requires not just visibility, but complete visibility. With a 10% target reflectivity and 100 Klux ambient light, the NOVA20 achieves a range of up to 1 meter with a blind zone of less than 5 centimeters. From the far end of its range down to the 5-centimeter close-proximity end, distance data remains continuous—ensuring that as obstacles get closer, readings stay online, allowing robots to "get close while seeing clearly".

Highlight 2: 100 Klux Glare Resistance and Zero Crosstalk, Reliable Under Intense Outdoor Sunlight

While triangulation linear lasers go blind outdoors, the NOVA20 resists up to 100 Klux of ambient light—equivalent to direct midday sunlight. This means the exact same near-field safety solution can scale seamlessly from indoor environments to inspection, logistics, and outdoor operations. When a robot transitions from a warehouse to an open-air loading dock where lighting changes drastically within seconds, the NOVA20 continues outputting reliable data without data confusion caused by lighting shifts.

To solve "crosstalk"—a long-standing industry headache in array LiDARs where a high-intensity echo "bleeds" into adjacent channels and mixes with real scenarios—the NOVA20 adopts a time-division multiplexing, sequential readout design. This ensures that each echo falls exclusively into its designated channel, translating real environments into clear, distinct points. By eliminating false signals, robots avoid emergency stops or detours caused by non-existent obstacles, making zero-margin edge-following operations far more reliable.

Highlight 3: Beer Bottle Cap Size, Under 5 Grams, Under 1 Watt—Distributed Along the Body and Deployed On Demand

Adopting a pure solid-state architecture with no moving parts, the NOVA20 measures only 25 × 22 × 13 mm (typical), comparable in size to a beer bottle cap. Weighing less than 5 grams (lighter than a standard one-yuan coin) and consuming an average power of less than 1 watt, it is small enough to mount directly onto robot bodies, robotic arms, and limbs for close-range anti-collision protection. Low power consumption also means multi-point deployment adds no extra thermal or stability burden to the system, ensuring long life and outdoor durability.

The NOVA20 does not output images; instead, it provides direct distance data to the underlying control system. This shortens the data pipeline, lowers latency, and preserves precious computing power for semantic understanding, task planning, and fine manipulation—much like how a human doesn't need to recognize a table's category and shape every time they check how far their hand is from a table corner. It acts like a circle of non-contact neural endings outside the robot's "skin," detecting distance changes before physical impact occurs rather than sensing force after a collision.

This design clearly defines product boundaries: the NOVA20 does not replace vision and models, nor does it handle fine force control post-contact; instead, it provides a low-burden, task-specific distance input stream. Paired with a UART interface and ROS/Linux/Windows support, original equipment manufacturers (OEMs) can flexibly deploy sensors based on risk points—covering critical locations first and expanding points according to robot morphology and application scenarios. With its compact size, low power, zero compute overhead, and controllable cost, the NOVA20 integrates directly into mass-production robot BOMs to support large-scale batch deployment.

A Single LiDAR Resolving Three Major Near-Field Challenges for Robots

Focusing on the most typical near-field risks for robots, the NOVA20 underpins three core capabilities: edge-following, fall prevention, and collision avoidance.

Edge-Following: When maintaining a centimeter-level distance from a wall, the robot stays close, moves straight, and avoids scraping.

Fall Prevention: Beams constantly scan the ground ahead, combining 20 points into an area scan to detect slopes and drop-offs in advance rather than waiting for single-point ranging to catch a sudden change.

Collision Avoidance: Beams target obstacles ahead, triggering alerts and stops upon approach to protect the system before physical contact occurs.

▲ Edge-following, fall prevention, and collision avoidance: Three core near-field safety capabilities supported by the NOVA20

Regarding installation, the NOVA20 follows an "on-demand deployment" design: humanoids on the chest, wrists, and ankles; quadrupeds on the jaw and feet; mobile chassis at the four corners—mounted wherever blind spots exist to piece together a localized safety perimeter without stacking bulky, expensive sensors all over the body. The blind spots of embodied AI are precisely where the NOVA20 was born to serve.

▲ Chest, wrists, ankles, and chassis corners—mount wherever blind spots exist

(Specific mounting locations, detection performance, and control strategies should be verified against specific robot structures and actual environments)

Key Parameters

About Benewake

Founded in 2015, Benewake is a LiDAR enterprise dedicated to AI perception for robots and the physical world. Grounded in full-stack technical capabilities spanning "optics, mechanics, electronics, software, algorithms, and chips," the company builds a full-scenario "Eyes of the Robot" product matrix, striving to create perception infrastructure for the physical intelligence era.

Solid-state LiDAR represents Benewake's core technical accumulation: from the first-generation solid-state flash LiDAR CE30 in 2017, to the solid-state small-angle linear LiDAR NOVA in 2024, and now the newly released pure solid-state near-field linear LiDAR NOVA20, nine years of continuous evolution across three generations have established a secure and reliable perception foundation for physical intelligence.