Introduction
In recent years, the development of intelligent vehicles has significantly increased interest in the LiDAR industry, with an increasing number of companies in China and abroad working in this field. LiDAR is a system that emits light beams and receives echoes to obtain three-dimensional information about targets, with decades of application history. LiDAR systems are complex, have diverse application scenarios, and include multiple technical approaches. Therefore, evaluating LiDAR performance is not simple, and several parameters are easily confused, such as detection precision, detection accuracy, resolution, frame rate, and point frequency.
Because there is no unified standard, different manufacturers may highlight different parameters in their specifications. Not all parameters are core to LiDAR performance, and some parameters are strongly interrelated. Without correctly understanding their meanings, interrelationships, and determining factors, users may be misled by a few high-value specifications and fail to grasp the true performance of a LiDAR unit.
This article selects key performance parameters for automotive LiDAR, defines and summarizes calculation methods, interrelationships, and determining factors to help users better apply LiDAR and guide LiDAR design.
Core Performance Parameters and Classification
LiDAR evaluation typically covers performance, reliability, and application aspects. Common parameters of interest include laser wavelength, detection range, field of view (vertical + horizontal), range precision, angular resolution, point count, beam count, safety class, output parameters, ingress protection, power consumption, supply voltage, laser emission type, and lifetime. Functionally, LiDAR is a ranging system, so its core performance should focus on measurement speed, spatial coverage, resolution, accuracy, and repeatability. By this principle, core parameters usually include frame rate, maximum and minimum detection distance, field of view, range resolution, horizontal and vertical angular resolution, point frequency, and range precision and accuracy.
Parameter Definitions
Referencing existing laser rangefinder standards, the parameter definitions are:
- Repeat frequency: number of range measurements completed per second.
- Range: the farthest/nearest distance that can be detected under specified atmospheric conditions and target characteristics while meeting specified range accuracy.
- Horizontal field of view: maximum azimuth scanning angle in the horizontal direction, θx.
- Vertical field of view: maximum azimuth scanning angle in the vertical direction, θγ.
- Range resolution: minimum distinguishable distance interval between two targets along the beam propagation direction.
- Angular resolution: the ability to resolve two targets located on the beam cross-section at a given range.
- Range precision: deviation of measured target distance values from actual distance values.
- Measurement accuracy: probability that a range measurement meets the specified range precision.
Relationships Between Parameters and Determining Factors
Frame Rate, Rotation Speed, and Scan Frequency
Rotation speed is an intuitive parameter for automotive LiDAR. For mechanical rotating LiDAR, rotation speed refers to the motor rotation speed, typically expressed in revolutions per second or per minute. For example, 20 r/s means the LiDAR motor rotates 20 revolutions per second.
Frame rate and scan frequency are related to rotation speed: one full rotation corresponds to one scan and one output frame of point cloud. A rotation speed of 20 r/s corresponds to a frame rate and scan frequency of 20 Hz.
Faster motor rotation yields a higher scanning speed and a faster output of point clouds.
Maximum and Minimum Detection Distance
Detection distance is a critical performance parameter for LiDAR. The LiDAR range equation provides the theoretical basis for estimating maximum detection distance. As illustrated, the laser emits power P0; after transmission through the optical system and atmospheric attenuation over distance R, light reaches the target, then is reflected back to the receiver.
In the formula, target reflectivity is ρ, the transmit and receive optical throughputs are TE and TR respectively, narrowband filter throughput is TF, atmospheric attenuation is σ, the angle between target surface normal and optical axis is β, the receiver lens aperture area is AR, and target distance is R.
Analysis of the equation shows that maximum range is influenced by both external ranging conditions and the LiDAR system's own performance. To extend range, designers may increase emitted laser power, improve optical throughput, enlarge receiver aperture, or reduce the minimum detectable receiver power PRmin.
Coefficients ε and γ correlate with spot size, target effective reflective area Am, far-field beam divergence angle θt, and receiver field of view θr.
When the target effective reflecting area is smaller than the beam spot area, the far-field divergence angle increases, causing the beam to spread more rapidly and reducing received optical power. To reduce beam divergence and spot size, the transmitter must be optimized through collimation and beam shaping. Besides maximum range, the separation between the transmitter and receiver optical paths inside the LiDAR can create a blind zone. When range is below the minimum detectable distance, the LiDAR cannot detect the target.
The minimum detection distance Lmin is calculated as shown in the article image.
In the formula, D is the perpendicular distance between the transmit collimation axis center and the receive convergence axis center; d is the detector active area diameter; f is the focal length of the chosen lens.
Field of View (FOV)
Field of view is an important metric for LiDAR sensing range. A larger FOV allows wider perception, which benefits vehicle safety. However, a larger FOV is not always better: for the same number of vertical lines, increasing the FOV reduces vertical angular resolution. FOV selection should balance the application scenario and requirements. Key factors determining FOV include the scanning mechanism and optical design. Mechanical rotation can readily achieve 360° horizontal FOV since the entire transmit-receive assembly rotates. Scanning methods using gimbals, prisms, or micro-electromechanical systems (MEMS) have FOV limits determined by mirror reflection angles. MEMS scanning is limited by micro-mirror aperture; to expand FOV, optical beam expansion using convex-concave lens assemblies can be employed.
Range Resolution
Range resolution ΔR is the minimum distinguishable distance between two targets at the same azimuth.
The transmitted pulse width is τ. If the time interval between the leading edges of two echo pulses Δt = 2ΔR / c ≥ τ, the pulses do not overlap and the LiDAR can resolve the two echoes; otherwise, the echoes overlap and the LiDAR cannot distinguish the two objects.
Therefore, for pulsed LiDAR the minimum range resolution satisfies ΔR ≥ cτ / 2. Reducing pulse width improves range resolution.
Horizontal and Vertical Angular Resolution
Angular resolution determines the smallest object size LiDAR can resolve at a given distance. For example, if vertical angular resolution is 0.08°, the angular separation between two beams is 0.08°. At 200 m, the beam spacing is 200 m × tan(0.08°) ≈ 0.28 m. Thus beyond 200 m, objects smaller than about 28 cm may be missed.
Vertical angular resolution is directly determined by the number of beams. With uniformly distributed beams:
Vertical angular resolution = vertical field of view / number of beams
To increase beam count, one-dimensional scanning typically stacks emitters in the central area, while two-dimensional scanning can vary beam distribution to adjust the angular range and vertical resolution for regions of interest.
Horizontal angular resolution is given by:
Horizontal angular resolution = field of view × rotation speed / sampling rate
Horizontal angular resolution depends on field of view and frame rate and cannot be considered independently. For the same FOV and sampling rate, lower rotation speed yields higher horizontal resolution.
Point Frequency (Points per Second)
Point frequency, also called point count, is the total number of detected points the LiDAR acquires per second. It is calculated as:
Point frequency = average horizontal points per frame × average vertical points per frame × frame rate
For one-dimensional scanning, average vertical points per scan equals the number of lines, so:
Point frequency = (horizontal FOV / horizontal resolution) × number of lines × frame rate = sampling rate × number of lines × frame rate / rotation speed = sampling rate × number of lines
Thus point frequency is determined by sampling rate and number of lines, and is independent of frame rate. Increasing frame rate reduces horizontal resolution while point frequency remains constant. Using point frequency avoids artificially improving horizontal resolution by lowering frame rate. Point frequency is a core performance parameter: a higher point frequency implies stronger perception capability.
Range Precision and Range Accuracy
Range precision and range accuracy are easily confused. Precision measures repeatability: high precision means repeated measurements of the same target are very close, while low precision indicates larger dispersion around the mean. Range accuracy reflects the closeness of measured distance to true distance.
Range precision is closely related to signal-to-noise ratio (SNR). When SNR exceeds a threshold, range data distribution approximates a normal distribution; as SNR decreases, measurement dispersion increases and precision degrades.
Range accuracy is directly proportional to the accuracy of time-of-flight measurement. LiDAR calculates time-of-flight by comparing transmit and receive pulse timing differences.
Accuracy of time-of-flight depends on the accuracy of transmit and receive pulse trigger times and on the frequency stability of internal clock oscillators. Transmit pulse timing and clock stability are determined by internal circuitry, while receive pulse trigger accuracy is influenced by pulse shape, target reflectivity characteristics, and background noise.
Summary
With the rapid development of driver-assistance applications, LiDAR technology continues to evolve. Multiple factors make LiDAR performance evaluation complex. This article analyzed several key automotive LiDAR performance parameters, their calculation methods, relationships, and determining factors to aid understanding and comparison of LiDAR systems.
