GB/T 45500-2025 “Performance Requirements and Test Methods for Automotive LiDAR” is China's first national technical specification for the performance testing of automotive LiDAR. It systematically specifies the performance indicators and test methods for LiDAR in terms of ranging, field of view, angle, reflection characteristics, and other aspects. This standard aims to unify industry testing benchmarks and enhance the reliability, safety, and consistency of LiDAR products.

This standard applies to the inspection of LiDAR used for external information perception in road vehicles, and other LiDAR systems may refer to it. This article will focus on the most core part of the standard—the detection performance testing—and provide an in-depth interpretation of how it systematically evaluates whether a LiDAR can “see far, measure accurately, distinguish clearly, and operate reliably.”

I. Ranging and Precision
Ranging Capability
Definition: The distance range within which the LiDAR can detect an object with normal laser incidence under specified conditions such as light intensity and detection probability. It is characterized by the maximum and minimum detection capabilities.
Specifications:
Maximum detection distance: Shall not be less than the nominal value.
Minimum detection distance: Shall not be greater than the nominal value.

Distance Precision
Definition: The degree of dispersion among multiple measurement values of a given quantity under specific measurement conditions.
Specifications: The distance precision in each field of view area within the nominal ranging range shall satisfy: $\sigma_d \leq \max(0.1\ \text{m}, 0.25\% \cdot S)$, where S is the current detection distance.

Distance Accuracy
Definition: The degree of deviation between the average of multiple measurement values and the true value of a given quantity under specific measurement conditions.
Specifications: The distance accuracy in each field of view area within the nominal ranging range shall satisfy: $|\delta_d| \leq \max(0.2\ \text{m}, 0.5\% \cdot S)$, where S is the current detection distance.

Planar Accuracy
Definition: The degree of conformity between the virtual plane constructed from LiDAR point cloud data and the true target plane.
Specifications: The planar accuracy $(P)$ in each field of view area of the LiDAR shall satisfy: $P \leqslant 1.5 \times \max(\sigma_{\text{d}}, \left| \delta_{\text{d}} \right|)$
Calculation Formula:
$$P = \sqrt{\frac{\sum_{1}^{n} b_i^2}{n}} \tag{1}$$ Symbol Definitions:
\( P \) is the planar accuracy, in meters (m); \( b_i \) is the distance from the \( i \)-th point to the true plane, in meters (m); \( n \) is the total number of selected valid points.

II. Angle and Field of View
Angular Precision and Accuracy
Definition: Angular precision reflects the degree of dispersion of angular measurement values (i.e., repeatability), while angular accuracy reflects the degree of deviation between the average of angular measurement values and the true value (i.e., correctness).
Specifications: The horizontal and vertical angular precision in each field of view area of the LiDAR shall not exceed 0.1°, and the horizontal and vertical angular accuracy shall not exceed ±0.2°.

Field of View:
Definition: In the point cloud generated by the LiDAR, the maximum angle formed by the lines connecting the outermost valid points on both sides to the ranging center in the horizontal/vertical direction. It is divided into the horizontal field of view and the vertical field of view, determining the “viewing range” of the LiDAR.
Specifications: Both the horizontal and vertical fields of view shall not be less than the nominal values.

Angular Resolution
Definition: In the point cloud generated by the LiDAR, the angle formed by the lines connecting adjacent detection points to the ranging center in the horizontal/vertical direction. It is divided into horizontal angular resolution and vertical angular resolution.
Specifications: The difference between the angular resolution and the nominal value shall not be greater than 0.25 times the nominal value.

Missed Detection Angle
Definition: The maximum angle between adjacent valid points within the LiDAR's field of view where a target cannot be detected. It is divided into the horizontal missed detection angle and the vertical missed detection angle.
Specifications: Shall not be greater than the nominal value.

III. Point Cloud Quality
Detection Probability
Definition: The probability of the LiDAR generating valid points during measurement, calculated as the ratio of the number of valid points in the point cloud to the theoretical number of detection points.

False Point: A point in the point cloud that corresponds to a spatial position in the real world where no actual object exists.
False Point Rate: The ratio of the number of false points to the theoretical number of detection points.

Reflection Characteristics
Definition: The reflectance mapping value of the target surface output by the LiDAR through the measurement of reflected echo amplitude or flux.

Specifications: There are explicit requirements for the measurement accuracy of reflection characteristics for targets with different reflectances (10%, 50%, 90%, Class V retroreflective sheeting) (see the table below).

High-Reflectivity Ghosting
Definition: In the point cloud generated by the LiDAR, false points that are continuous in time and space, corresponding to positions where no real target exists.
High-reflectivity ghosting originates from the LiDAR's own emitted laser generating multiple internal echoes after encountering highly reflective objects (such as traffic signs, traffic cones, triangular warning signs, license plates, taillights, etc.), resulting in the generation of continuous false points at incorrect distances (such as the large red area at the top of the image and the small red blocks beside the track in the figure above).
Specifications: The maximum single-frame false point rate in each field of view area of the LiDAR shall not be greater than 0.1%.

Inter-LiDAR Interference Resistance
Definition: The ability of a LiDAR to resist interference from laser pulses emitted by other LiDARs of the same type in a multi-LiDAR operating environment, while correctly detecting real targets and suppressing false points.
When the inter-LiDAR interference resistance is insufficient, random and discrete false points will be injected into the point cloud, which may obscure real weak targets.

Specifications: The false point rate during the test shall not be greater than 0.1%.

Trailing Points
Definition: False points in the point cloud generated by the LiDAR that appear between the edges of two corresponding objects.
Specifications: Under specified conditions (e.g., target distance > 5m), no trailing points shall exist in the point cloud.

IV. Dynamic Performance and Reliability Assessment
Real-world driving environments are dynamic, and the standard fully takes this characteristic into account.
Startup Time: The time from power-on to the output of a valid point cloud is required to not exceed 40 seconds, meeting the demand for rapid vehicle startup.
Point Rate and Frame Rate: The output stability is required to be no less than 99.9% of the nominal value, ensuring the continuity of the perception system's data stream and preventing decision-making errors caused by data frame drops.

V. Testing Solutions: Yanding has independently developed a standardized testing solution for automotive LiDAR that complies with the GB/T 45500-2025 national standard requirements and covers the entire process from R&D to mass production.
This testing system adopts a modular design, and the core equipment includes: