Laser ranging signal-to-noise ratio adaptive control method and system based on AI illumination information

Abstract

The present application relates to the technical field of laser ranging, and more particularly to a laser ranging signal-to-noise ratio adaptive control method and system based on AI lighting information. The method obtains ambient lighting information and generates a signal-to-noise ratio degradation prediction value, searches for a Pareto optimal combination of transmission and reception parameters for adjustment, and then performs feedback correction based on the deviation of the actual echo signal-to-noise ratio from the target until the deviation converges. Finally, the corrected parameters are used to drive the ranging. The present application realizes adaptive optimization control of the signal-to-noise ratio, improves the stability and accuracy of laser ranging in complex lighting environments, and reduces laser transmission power consumption.

Description

Technical Field

[0001] This invention relates to the field of laser ranging technology, and in particular to a laser ranging signal-to-noise ratio adaptive control method and system based on AI illumination information. Background Technology

[0002] Existing laser ranging systems typically employ fixed thresholds or simple empirical adjustments for parameter configuration. The peak power and pulse width of the transmitting unit are preset based on the maximum range or typical operating conditions, while the detection threshold and amplifier gain of the receiving unit are statically set based on a calibrated single echo intensity. This approach can maintain basic ranging requirements in stable and ideal environments, but it cannot adapt to complex and ever-changing real-world scenarios.

[0003] When ambient lighting conditions change drastically, fixed parameter settings reveal significant limitations. Under strong light, suspended particles in the air and the scattering of sunlight by the target surface create high-level background noise, causing a sharp drop in the signal-to-noise ratio of the echo signal. Maintaining a low detection threshold introduces numerous false alarms, while excessively high amplifier gain can saturate the noise, ultimately leading to severe degradation of ranging accuracy or even missed target detection. Conversely, in low-light or nighttime environments, although background noise is reduced, the fixed transmit power may be far higher than actually required. This not only wastes the laser's energy resources but also poses a potential hazard to humans or other sensitive equipment due to excessive peak power. If the receiving parameters are not adjusted accordingly, the detection advantages in low-noise environments cannot be fully utilized.

[0004] Existing technologies include various schemes for dynamic gain control at the receiver, such as automatically adjusting the amplifier gain based on the echo amplitude or switching the detection threshold based on the target distance. However, these methods are usually limited to the independent adjustment of a single component, failing to consider the transmitting and receiving units as a coupled whole. While increasing the transmit power can enhance the echo amplitude, it also exacerbates backscatter noise. If the receiver does not simultaneously optimize the threshold and gain, the improvement in signal-to-noise ratio will be limited, and it may even be counterproductive due to a surge in power consumption. Similarly, while reducing the receiver gain can suppress noise, if the transmit pulse width is too wide, it cannot effectively suppress background light interference, and the synergistic effect between the two is completely ignored.

[0005] Another common characteristic of the aforementioned isolated adjustment strategies is the lack of proactive sensing and prediction of illumination information. Most systems rely on real-time feedback from echo signals for passive response, but abrupt changes in ambient illumination often outpace the convergence speed of feedback correction, resulting in a large amount of invalid or erroneous ranging data during parameter adjustment. Furthermore, most existing adjustment methods do not consider the Pareto balance between power consumption and signal-to-noise ratio, often pushing transmission parameters to high power extremes to ensure reliable detection, which contradicts the requirements of modern equipment for low energy consumption and green operation. Therefore, a control strategy that can comprehensively consider ambient illumination characteristics and achieve joint dynamic optimization of transmission and reception parameters is urgently needed. Summary of the Invention

[0006] This invention provides a laser ranging signal-to-noise ratio adaptive control method and system based on AI illumination information, which can solve the problems in the prior art.

[0007] A first aspect of the present invention provides a laser ranging signal-to-noise ratio adaptive control method based on AI illumination information, comprising: The ambient illumination information of the laser ranging target area is acquired and its features are extracted to generate a signal-to-noise ratio degradation prediction value under the current illumination conditions. The signal-to-noise ratio degradation prediction value is used as input to search for the Pareto optimal combination of transmission parameters and reception parameters in the preset transmission parameter-reception parameter joint adjustment strategy space to maximize the echo signal-to-noise ratio and minimize the laser transmission power consumption. The peak power and pulse width of the laser emitting unit are adjusted according to the emission parameters in the Pareto optimal combination, and the detection threshold and amplifier gain of the echo signal processing unit are adjusted according to the receiving parameters in the Pareto optimal combination. Acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with the preset target SNR, and generate an SNR deviation. When the SNR deviation exceeds a preset tolerance threshold, use the SNR deviation as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold. The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are processed and their arrival time is detected using the corrected reception parameters to calculate the target distance.

[0008] The system acquires ambient illumination information of the laser ranging target area and extracts its features to generate a predicted signal-to-noise ratio degradation value under the current illumination conditions, including: Multi-dimensional illumination feature separation is performed on the ambient illumination information to extract the average illumination intensity feature, illumination spatial contrast feature, and illumination temporal flicker frequency feature. The average light intensity feature, the spatial contrast of light, and the temporal flicker frequency of light are concatenated to generate a multi-dimensional joint feature vector of light. The multi-dimensional illumination joint feature vector is input into a pre-trained cascaded nonlinear mapping network. The cascaded nonlinear mapping network contains multiple cascaded feature compression coding layers and nonlinear activation transformation layers. Through layer-by-layer dimensionality reduction compression and nonlinear mapping, the multi-dimensional illumination joint feature vector is transformed into a signal-to-noise ratio degradation prediction value under the current illumination conditions. The prediction confidence score corresponding to the signal-to-noise ratio degradation prediction value is generated synchronously at the output layer of the cascaded nonlinear mapping network. The prediction confidence score is used to characterize the reliability of the prediction of the signal-to-noise ratio degradation prediction value under different illumination distribution patterns.

[0009] Using the predicted signal-to-noise ratio degradation as input, a Pareto optimal combination of transmission and reception parameters that maximizes the echo signal-to-noise ratio and minimizes laser emission power consumption is searched within a preset joint adjustment strategy space for transmission and reception parameters. This includes: A two-dimensional orthogonal gridded joint adjustment strategy space is constructed, with the peak power and pulse width of the laser emitting unit constituting the emission parameter dimension, and the detection threshold offset and amplifier gain compensation coefficient of the echo signal processing unit constituting the receiving parameter dimension. Each grid intersection node corresponds to a set of pre-calibrated expected echo signal-to-noise ratio and expected laser emission power consumption. Using the predicted signal-to-noise ratio degradation value and the predicted confidence score as search inputs, all grid intersection nodes are traversed and evaluated in the joint adjustment strategy space. With the goal of maximizing the expected value of echo signal-to-noise ratio and minimizing the expected value of laser emission power consumption, the Pareto front grid node set is selected by non-dominated sorting. In the Pareto front grid node set, a weighted comprehensive score is calculated for each candidate grid node using a preset signal-to-noise ratio priority weighting factor. The transmit and receive parameters corresponding to the grid node with the highest weighted comprehensive score are selected as the Pareto optimal combination.

[0010] The set of Pareto front grid nodes was selected through non-dominated sorting, including: Using the expected echo signal-to-noise ratio as the first optimization objective dimension and the expected laser emission power consumption as the second optimization objective dimension, the pairwise dominance relationship of all grid intersection nodes in the joint adjustment strategy space is determined: if the first node is not inferior to the second node in both the expected echo signal-to-noise ratio and the expected laser emission power consumption, and is strictly superior to the second node in at least one dimension, then the first node is determined to dominate the second node. Extract all grid nodes in the joint regulation strategy space that no other grid node can dominate, and form the Pareto front grid node set; In the Pareto front grid node set, the nodes are sorted from high to low according to the expected echo signal-to-noise ratio. The ratio of the decrease in the expected echo signal-to-noise ratio between adjacent grid nodes in the sorted sequence to the decrease in the expected laser emission power consumption is calculated as the local marginal substitution rate. Grid nodes with a local marginal substitution rate lower than a preset marginal efficiency threshold are removed to eliminate redundant nodes in the Pareto front that have limited power consumption reduction but significant signal-to-noise ratio loss.

[0011] Acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with a preset target SNR, and generate an SNR deviation, including: The time-domain waveform of the acquired echo pulse signal is subjected to continuous non-overlapping segmented windowing processing according to a preset time window length to generate multiple time-domain signal windows of equal length; For each time-domain signal window, the noise floor segment before the arrival of the echo pulse and the signal superposition segment after the arrival of the echo pulse are extracted. The mean of the squares of the sampling point amplitudes in the signal superposition segment is calculated as the mean signal power, and the mean of the squares of the sampling point amplitudes in the noise floor segment is calculated as the mean noise power. The ratio of the mean signal power to the mean noise power is used as the signal-to-noise ratio estimate for that window. The signal-to-noise ratio (SNR) estimates of all time-domain signal windows are sorted in ascending order, and the median of the sorted values ​​is taken as the actual echo SNR, so as to eliminate abnormal deviations in the SNR estimates of individual windows caused by instantaneous illumination interference. The difference between the actual echo signal-to-noise ratio and the preset target signal-to-noise ratio is calculated, and the absolute value of the difference is taken as the signal-to-noise ratio deviation.

[0012] When the signal-to-noise ratio (SNR) deviation exceeds a preset tolerance threshold, the SNR deviation is used as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold, including: Taking the grid node where the currently effective transmission parameters and reception parameters are located in the joint adjustment strategy space as the current node, obtain the expected echo signal-to-noise ratio (SNR) of the current node and all adjacent grid nodes, and add the SNR deviation to the expected echo SNR of the current node to obtain the target SNR expected value. Among the neighboring grid nodes of the current node, the grid node whose expected echo signal-to-noise ratio is closest to the expected target signal-to-noise ratio is selected as the target correction node, and the transmission parameters and reception parameters corresponding to the target correction node are used as the parameters after one correction. The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are re-acquired to calculate the signal-to-noise ratio. The deviation between the newly obtained signal-to-noise ratio and the preset target signal-to-noise ratio is then calculated. Repeat the target node screening and parameter correction loop until the signal-to-noise ratio deviation converges to within the preset tolerance threshold.

[0013] The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are processed and their arrival time detected using the corrected reception parameters to calculate the target distance, including: The time-domain sampling sequence of the echo pulse is subjected to matched filtering with a preset transmitted pulse reference waveform. The signal-to-noise ratio of the echo pulse response under noise background is enhanced by sliding cross-correlation operation, and the enhanced echo pulse envelope is extracted from the cross-correlation result. A trigger amplitude discrimination level is set on the rising edge of the enhanced echo pulse envelope. The trigger amplitude discrimination level is the product of the current noise floor root mean square value and a preset threshold multiple. The arrival time of the echo pulse is determined by the intersection of the trigger amplitude discrimination level and the rising edge of the envelope. The emission time of the laser pulse is obtained and the time-of-flight difference between the arrival time of the echo pulse and the emission time is calculated. A timing compensation amount associated with the constant false alarm detection threshold offset is applied to the time-of-flight difference to compensate for the arrival time detection drift caused by the threshold offset. The product of the compensated time difference and the speed of light constant is taken as the target distance between the laser ranging system and the target.

[0014] A second aspect of the present invention provides a laser ranging signal-to-noise ratio adaptive control system based on AI illumination information, comprising: The parameter optimization unit is used to acquire ambient lighting information of the laser ranging target area and extract features to generate a signal-to-noise ratio degradation prediction value under the current lighting conditions. Using the signal-to-noise ratio degradation prediction value as input, it searches in the preset joint adjustment strategy space of transmission parameters and receiving parameters to maximize the echo signal-to-noise ratio and minimize the laser transmission power consumption. The parameter adjustment unit is used to adjust the peak power and pulse width of the laser emitting unit according to the emission parameters in the Pareto optimal combination, and to adjust the detection threshold and amplifier gain of the echo signal processing unit according to the receiving parameters in the Pareto optimal combination. The feedback correction unit is used to acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with a preset target SNR, generate an SNR deviation, and when the SNR deviation exceeds a preset tolerance threshold, use the SNR deviation as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold. The ranging calculation unit is used to drive the laser to emit ranging pulses with the corrected emission parameters, and to perform signal processing and arrival time detection on the echo pulses with the corrected reception parameters to calculate the target distance.

[0015] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0016] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0017] By acquiring ambient lighting information in real time and predicting signal-to-noise ratio (SNR) degradation, the system can proactively anticipate SNR decline trends under varying lighting conditions, avoiding performance failures of traditional fixed-parameter solutions in strong or weak light environments. A joint transmit and receive parameter adjustment strategy based on Pareto optimal search maximizes the echo SNR while minimizing laser emission power consumption, achieving a balance between high SNR and low energy consumption. This predictive optimization significantly reduces unnecessary power waste, enabling the system to maintain optimal operating conditions even during sudden changes in lighting.

[0018] A closed-loop feedback mechanism is introduced to address the deviation between the actual echo signal-to-noise ratio (SNR) and the target SNR. Transmit and receive parameters are iteratively optimized in the joint adjustment strategy space using gradient direction correction. When ambient lighting or target reflection characteristics dynamically change, this method can quickly track and automatically adjust parameters to bring the SNR deviation within the tolerance range. Adaptive closed-loop control effectively suppresses SNR fluctuations caused by external interference, ensuring stable reception and processing of ranging pulses and avoiding the loss of ranging points or false echoes caused by parameter mismatch in traditional open-loop control. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the adaptive control method for laser ranging signal-to-noise ratio based on AI illumination information. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] Figure 1 This is a flowchart illustrating the adaptive control method for laser ranging signal-to-noise ratio based on AI illumination information according to an embodiment of the present invention. The adaptive control method for laser ranging signal-to-noise ratio based on AI illumination information includes: The ambient illumination information of the laser ranging target area is acquired and its features are extracted to generate a signal-to-noise ratio degradation prediction value under the current illumination conditions. The signal-to-noise ratio degradation prediction value is used as input to search for the Pareto optimal combination of transmission parameters and reception parameters in the preset transmission parameter-reception parameter joint adjustment strategy space to maximize the echo signal-to-noise ratio and minimize the laser transmission power consumption. The peak power and pulse width of the laser emitting unit are adjusted according to the emission parameters in the Pareto optimal combination, and the detection threshold and amplifier gain of the echo signal processing unit are adjusted according to the receiving parameters in the Pareto optimal combination. Acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with the preset target SNR, and generate an SNR deviation. When the SNR deviation exceeds a preset tolerance threshold, use the SNR deviation as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold. The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are processed and their arrival time is detected using the corrected reception parameters to calculate the target distance.

[0023] The system acquires ambient illumination information of the laser ranging target area and extracts its features to generate a predicted signal-to-noise ratio degradation value under the current illumination conditions, including: Multi-dimensional illumination feature separation is performed on the ambient illumination information to extract the average illumination intensity feature, illumination spatial contrast feature, and illumination temporal flicker frequency feature. The average light intensity feature, the spatial contrast of light, and the temporal flicker frequency of light are concatenated to generate a multi-dimensional joint feature vector of light. The multi-dimensional illumination joint feature vector is input into a pre-trained cascaded nonlinear mapping network. The cascaded nonlinear mapping network contains multiple cascaded feature compression coding layers and nonlinear activation transformation layers. Through layer-by-layer dimensionality reduction compression and nonlinear mapping, the multi-dimensional illumination joint feature vector is transformed into a signal-to-noise ratio degradation prediction value under the current illumination conditions. The prediction confidence score corresponding to the signal-to-noise ratio degradation prediction value is generated synchronously at the output layer of the cascaded nonlinear mapping network. The prediction confidence score is used to characterize the reliability of the prediction of the signal-to-noise ratio degradation prediction value under different illumination distribution patterns.

[0024] In the stage of acquiring ambient illumination information of the target area for laser ranging, a wide dynamic range photoelectric sensor array deployed at the front end of the laser ranging system continuously collects ambient illumination data of the target area at a sampling frequency of no less than 100Hz. The collected raw illumination data contains three basic data dimensions: spatial distribution information, temporal variation information, and intensity amplitude information. Before entering the feature extraction process, the raw illumination data needs to be preprocessed, including removing sensor dark current noise, performing pixel-level gain correction, and filtering out abnormal frames, to ensure that the illumination data input to the feature extraction stage has sufficient signal-to-noise quality.

[0025] Multi-dimensional illumination feature separation is a crucial preliminary step in generating signal-to-noise ratio (SNR) degradation predictions. The mean illumination intensity feature is obtained by taking the spatial mean of the irradiance readings of all valid pixels within the acquisition frame, reflecting the overall background illumination level of the target area. When the background illumination intensity is high, the competition between echo photons and background photons intensifies, and the detector's SNR will significantly decrease with increasing background irradiance. Therefore, the mean illumination intensity feature is one of the core inputs for predicting the degree of SNR degradation. The spatial contrast illumination feature is obtained by calculating the ratio of the standard deviation to the mean of the irradiance in a local area within the same acquisition frame, i.e., the normalized spatial variation coefficient. This feature quantifies the non-uniformity of illumination distribution within the target area. In scenes with high spatial contrast illumination, there are obvious bright patches or alternating shadows within the target area, leading to a non-uniform SNR distribution in the spatial dimension of the echo signal, thus affecting the accuracy of arrival time detection. The temporal flicker frequency characteristics of illumination are obtained by performing a Fast Fourier Transform on a sequence of pixel mean values ​​from multiple consecutive frames to extract the dominant frequency component and its power spectral density. This characteristic is used to characterize the periodic intensity modulation introduced by ambient light sources (such as fluorescent lamps and LED streetlights). Temporal flicker introduces interference components coupled with the pulse repetition frequency within the echo acquisition window, making it more difficult to set the detection threshold for the echo signal processing unit. Therefore, the temporal flicker frequency characteristics make a significant contribution to signal-to-noise ratio degradation prediction.

[0026] After independently extracting the three types of illumination features, the mean illumination intensity feature, spatial contrast illumination feature, and temporal flicker frequency illumination feature are concatenated along the feature dimension to form a multi-dimensional joint illumination feature vector. Let the dimension of the mean illumination intensity feature be... The dimension of the spatial contrast feature of illumination is The dimension of the illumination time-domain scintillation frequency characteristic is The dimension of the spliced ​​multi-dimensional illumination joint feature vector is then... The channel splicing operation preserves the physical semantics of each of the three types of features, while providing complete multi-source illumination information input for subsequent cascaded nonlinear mapping networks, thus avoiding information loss during feature fusion. In practical deployment, It is usually taken as 1 (i.e., the scalar mean). Take 1 (i.e., scalar coefficient of variation). Take several dimensions related to the spectral resolution, for example, take the amplitude of the main frequency component and its corresponding frequency, a total of two values. This forms a compact four-dimensional joint feature vector, which is convenient for real-time operation on embedded processing platforms.

[0027] The multi-dimensional illumination joint feature vector is then fed into a pre-trained cascaded nonlinear mapping network. This network employs a structure of alternating stacked cascaded feature compression coding layers and nonlinear activation transformation layers. The feature compression coding layer is essentially a fully connected linear transformation layer, where the number of rows in its weight matrix is ​​less than the number of columns, thereby achieving layer-by-layer compression of the input feature dimensions and gradually mapping the high-dimensional joint feature vector to a low-dimensional implicit representation space. The nonlinear activation transformation layer follows each feature compression coding layer, using a parameterized rectified linear unit (PReLU) as the activation function. Its negative half-axis slope is automatically optimized during training as a learnable parameter, providing stronger feature representation capabilities compared to the standard ReLU activation function and better fitting the complex nonlinear relationship between illumination conditions and signal-to-noise ratio degradation. The network's final output layer is a single-neuron linear output, directly outputting the predicted signal-to-noise ratio degradation value. This value physically represents the expected decrease in the system's echo signal-to-noise ratio under the current illumination conditions relative to a standard background-free illumination reference condition, expressed in dB.

[0028] The training dataset for the cascaded nonlinear mapping network was obtained through field measurements of the laser ranging system under various typical lighting scenarios (including strong direct sunlight, overcast diffuse light, indoor fluorescent lights, LED streetlights, and mixed light sources). For each scenario, the corresponding joint illumination feature vector and the measured signal-to-noise ratio (SNR) degradation were recorded, forming supervised training sample pairs. The training process employed a mean squared error loss function, combined with L2 regularization to prevent overfitting, and used the Adam optimizer for parameter updates. After training, the network weights were fixed and deployed to the embedded processor of the laser ranging control unit, outputting the predicted SNR degradation value in real time during system operation via forward inference.

[0029] In the output layer of the cascaded nonlinear mapping network, in addition to the signal-to-noise ratio (SNR) degradation prediction value, a corresponding prediction confidence score is generated simultaneously. The generation mechanism of the prediction confidence score is based on the Monte Carlo Dropout method: during the inference phase, each Dropout layer in the network maintains a fixed probability of randomly deactivating some neurons, performs multiple forward inferences on the same input joint feature vector, obtains sampling results of multiple SNR degradation prediction values, calculates the variance of these sampling results as an uncertainty measure, and then converts the variance into a confidence score between 0 and 1 through a monotonically decreasing mapping function. When the confidence score is close to 1, it indicates that the network has high reliability in predicting the SNR degradation under the current illumination conditions, and the predicted value can be directly used for subsequent joint adjustment of transmission and reception parameters; when the confidence score is low, it indicates that there is a large deviation between the current illumination distribution and the training sample distribution, and the network prediction has high uncertainty. At this time, the system will trigger a conservative adjustment strategy, that is, in the joint adjustment strategy space, it will prioritize the combination of reception parameters with a larger margin to cope with the risk of insufficient SNR caused by prediction errors. The prediction confidence score and the signal-to-noise ratio degradation prediction value together constitute the input information for the subsequent Pareto optimal search, enabling the entire adaptive control process to maintain robust control performance under different illumination distribution patterns.

[0030] Using the predicted signal-to-noise ratio degradation as input, a Pareto optimal combination of transmission and reception parameters that maximizes the echo signal-to-noise ratio and minimizes laser emission power consumption is searched within a preset joint adjustment strategy space for transmission and reception parameters. This includes: A two-dimensional orthogonal gridded joint adjustment strategy space is constructed, with the peak power and pulse width of the laser emitting unit constituting the emission parameter dimension, and the detection threshold offset and amplifier gain compensation coefficient of the echo signal processing unit constituting the receiving parameter dimension. Each grid intersection node corresponds to a set of pre-calibrated expected echo signal-to-noise ratio and expected laser emission power consumption. Using the predicted signal-to-noise ratio degradation value and the predicted confidence score as search inputs, all grid intersection nodes are traversed and evaluated in the joint adjustment strategy space. With the goal of maximizing the expected value of echo signal-to-noise ratio and minimizing the expected value of laser emission power consumption, the Pareto front grid node set is selected by non-dominated sorting. In the Pareto front grid node set, a weighted comprehensive score is calculated for each candidate grid node using a preset signal-to-noise ratio priority weighting factor. The transmit and receive parameters corresponding to the grid node with the highest weighted comprehensive score are selected as the Pareto optimal combination.

[0031] After obtaining the predicted signal-to-noise ratio (SNR) degradation value, a multi-objective optimization search needs to be performed in a pre-constructed joint adjustment strategy space to find the optimal parameter combination that simultaneously satisfies the two competing objectives of maximizing the echo SNR and minimizing the laser emission power consumption. The joint adjustment strategy space is organized using a two-dimensional orthogonal grid structure. One dimension consists of the peak power and pulse width of the laser emission unit, while the other dimension consists of the detection threshold offset and amplifier gain compensation coefficient of the echo signal processing unit. These four parameters are combined in pairs and discretized at uniform intervals within their respective value ranges, forming a regular grid intersection node array. Each grid intersection node corresponds to a set of pre-calibrated expected echo SNR and expected laser emission power consumption values. These two expected values ​​are obtained through extensive experimental measurements and simulation calculations before the system is manufactured or deployed, and are stored in the controller's non-volatile memory in the form of a lookup table, allowing direct indexing during real-time operation without recalculation.

[0032] The discretization range of the peak power dimension is determined based on the laser's rated maximum output power and minimum available power, while the discretization range of the pulse width dimension is determined based on the ranging accuracy requirements and the laser's allowable pulse duration. The detection threshold offset characterizes the relative offset based on the reference detection threshold, and its value can be positive (increasing the threshold to suppress noise) or negative (decreasing the threshold to improve weak signal detection sensitivity). The amplifier gain compensation coefficient characterizes the multiplicative adjustment factor based on the reference gain, and its value range covers the system's allowable minimum gain to maximum gain range. Through the orthogonal combination of the above four parameters, the joint adjustment strategy space constitutes a four-dimensional discrete parameter grid, but at the implementation level, it is projected as a two-dimensional plane for visualization and index management. The transmit parameter pair (peak power, pulse width) serves as the row index, and the receive parameter pair (detection threshold offset, amplifier gain compensation coefficient) serves as the column index. Each row and column intersection is a grid node.

[0033] The search phase uses the predicted signal-to-noise ratio (SNR) degradation value and the prediction confidence score as dual inputs to traverse and evaluate all grid intersection nodes in the joint adjustment strategy space. During traversal, for each node, the corresponding expected echo SNR value and expected laser emission power consumption value are read from the lookup table to form a dual-objective evaluation vector for that node. When the prediction confidence score is high, the traversal evaluation directly uses the calibration value from the lookup table; when the prediction confidence score is low (e.g., the current lighting scene differs significantly from the calibration scene), a confidence-weighted uncertainty correction is applied to the expected value in the lookup table, conservatively lowering the estimated echo SNR value to avoid insufficient actual SNR due to prediction bias. This confidence-aware evaluation mechanism makes the search process robust to prediction errors.

[0034] After completing the bi-objective evaluation of all nodes, a non-dominated sorting method is used to select the Pareto front node set. For any two nodes in the node set... and If node It is not inferior to the expected echo signal-to-noise ratio of [previous value]. Furthermore, it is significantly superior to the expected value of laser emission power consumption. Or, it is significantly better than the expected echo signal-to-noise ratio. Furthermore, it is not inferior to the expected value of laser emission power consumption. Then it is called Dominate The Pareto front set consists of all nodes that are not dominated by any other node, denoted as set A. .gather The nodes in the diagram represent the optimal trade-off boundary between signal-to-noise ratio (SNR) performance and power consumption under the current SNR degradation prediction conditions. Any node not included in this boundary represents the optimal trade-off boundary between SNR performance and power consumption. Each node in the algorithm has at least one alternative node that is no worse than either objective and is better than at least one objective, and therefore can be directly eliminated.

[0035] Pareto front node set Once determined, a unique optimal combination needs to be selected as the final output. Because... The trade-off between signal-to-noise ratio and power consumption varies among nodes, necessitating the introduction of decision preferences for final selection. A preset signal-to-noise ratio priority weighting factor is used. (satisfy Calculate a weighted composite score for each candidate node. For the set... The first in Each node has a weighted composite score. The calculation method is as follows: the expected value of the echo signal-to-noise ratio of this node is normalized across all Pareto front nodes and then multiplied by... In addition, the expected laser emission power consumption at this node is normalized across all Pareto front nodes, inverted (i.e., 1 minus the normalized power consumption value), and multiplied by... ,Right now ; in For the first The expected value of the normalized echo signal-to-noise ratio of each node. For the first The expected normalized laser emission power consumption of each node. This is the signal-to-noise ratio priority weighting factor. The normalization operation uses max-min normalization, linearly mapping the original expected value of each node to... The interval allows nodes with higher signal-to-noise ratios or lower power consumption to have normalized values ​​for their corresponding components that are closer to 1.

[0036] Signal-to-noise ratio priority weighting factor The value of reflects the system's preference for a balance between signal-to-noise ratio performance and power saving. In scenarios with extremely high ranging accuracy requirements or strong ambient light interference, Taking values ​​that are too large (e.g., 0.8 to 0.9) tends to result in the selection of nodes with higher signal-to-noise ratios, even if their power consumption is relatively high; in power-sensitive low-power application scenarios, The value is set to a smaller value (e.g., 0.4 to 0.6) so that the power consumption is reduced as much as possible while meeting the basic signal-to-noise ratio requirements. The specific values ​​are preset according to the application scenario requirements before system deployment, or can be dynamically adjusted according to the runtime status such as remaining battery power and task priority.

[0037] For sets Calculation of all nodes Then, the node with the highest score is selected, and its corresponding peak power, pulse width, detection threshold offset, and amplifier gain compensation coefficient constitute the Pareto optimal combination. This combination is then passed to the laser emission unit and the echo signal processing unit, respectively, to drive the hardware adjustment of the corresponding parameters. The entire search and decision-making process is implemented in the controller using a lookup table and sorting method. The computational complexity is linearly related to the total number of grid nodes. When the number of nodes is moderate (e.g., the emission parameter dimension and the receiving parameter dimension each have 16 to 32 discrete levels, and the total number of nodes does not exceed several thousand), it can be completed in the microsecond to millisecond range, meeting the real-time requirements of laser ranging systems.

[0038] The set of Pareto front grid nodes was selected through non-dominated sorting, including: Using the expected echo signal-to-noise ratio as the first optimization objective dimension and the expected laser emission power consumption as the second optimization objective dimension, the pairwise dominance relationship of all grid intersection nodes in the joint adjustment strategy space is determined: If the first node is not inferior to the second node in both the expected value of echo signal-to-noise ratio and the expected value of laser emission power consumption, and is strictly superior to the second node in at least one dimension, then the first node is determined to dominate the second node. Extract all grid nodes in the joint regulation strategy space that no other grid node can dominate, and form the Pareto front grid node set; In the Pareto front grid node set, the nodes are sorted from high to low according to the expected echo signal-to-noise ratio. The ratio of the decrease in the expected echo signal-to-noise ratio between adjacent grid nodes in the sorted sequence to the decrease in the expected laser emission power consumption is calculated as the local marginal substitution rate. Grid nodes with a local marginal substitution rate lower than a preset marginal efficiency threshold are removed to eliminate redundant nodes in the Pareto front that have limited power consumption reduction but significant signal-to-noise ratio loss.

[0039] After constructing the joint adjustment strategy space and evaluating the expected echo signal-to-noise ratio and laser emission power consumption of each grid node, it is necessary to screen out the true Pareto front from all candidate grid nodes to support the subsequent selection of the optimal parameter combination. The core of this screening process lies in establishing a strict criterion for determining the dominance relationship between the two objectives, and on this basis, completing the non-dominated sorting. Finally, redundant nodes in the Pareto front are further eliminated by calculating the local marginal substitution rate.

[0040] When determining dominance, the expected echo signal-to-noise ratio (SNR) is defined as the first optimization objective dimension, and the expected laser emission power consumption is defined as the second optimization objective dimension. For any two grid nodes in the joint adjustment strategy space, denoted as node 1 and node 2, the criteria for determining whether node 1 dominates node 2 are as follows: if the expected echo SNR of node 1 is not lower than that of node 2, and the expected laser emission power consumption of node 1 is not higher than that of node 2 (i.e., "not inferior" in the power consumption dimension means that the power consumption is not greater), and at least one of the above two dimensions is strictly superior to node 2 (i.e., the echo SNR is strictly higher, or the laser emission power consumption is strictly lower), then node 1 is determined to dominate node 2. This determination logic reflects the standard definition of Pareto dominance in multi-objective optimization: not inferior to the other in all objective dimensions, and strictly dominant in at least one objective dimension.

[0041] It is particularly important to note that laser emission power consumption is a minimization objective. Therefore, in terms of power consumption, "not inferior to" means the power consumption value is not higher, while "strictly superior to" means the power consumption value is strictly lower. Echo signal-to-noise ratio (SNR) is a maximization objective. "Not inferior to" means the SNR is not lower, while "strictly superior to" means the SNR is strictly higher. In actual implementation, traversing and determining the pairwise dominance relationships of all grid nodes in the joint regulation strategy space has a time complexity proportional to the square of the total number of nodes. When the grid partitioning of the strategy space is relatively fine and the number of nodes is large, an acceleration method based on sorting preprocessing can be used: first, all nodes are sorted in descending order according to the expected echo SNR value; then, only the nodes at the bottom of the sort are checked one by one to see if they are dominated by a node at the top of the sort, thus significantly reducing the number of invalid comparisons.

[0042] After determining the pairwise dominance relationships of all nodes, all grid nodes in the joint adjustment strategy space that cannot be dominated by any other grid node are extracted, forming the Pareto front grid node set. These nodes represent the optimal trade-off boundary between the echo signal-to-noise ratio and laser emission power consumption, which cannot be simultaneously improved under the current strategy space discretization conditions. Each node on the Pareto front corresponds to a position in both target spaces that cannot be dominated by any other node; that is, there is no other node that can reduce power consumption without reducing the signal-to-noise ratio, or increase the signal-to-noise ratio without increasing power consumption.

[0043] After obtaining the Pareto front grid node set, a local marginal substitution rate analysis is performed to remove redundant nodes with limited power consumption reduction but significant signal-to-noise ratio loss. Specifically, all nodes in the Pareto front grid node set are first sorted from highest to lowest according to their expected echo signal-to-noise ratio, forming an ordered sequence. Let the node in this sequence be... The expected echo signal-to-noise ratio of each node is The expected power consumption of laser emission is Then the two adjacent nodes (the first one) The and the first Local marginal rate of substitution between ( ) Defined as: ; in, Indicates from the first The node moves to the first The expected increase in echo signal-to-noise ratio (SNR) at each node (i.e., the SNR loss, which is the decrease from the high SNR node to the low SNR node). This represents the expected decrease in the corresponding laser emission power consumption (since the sequence is arranged in descending order of signal-to-noise ratio, nodes with lower signal-to-noise ratios typically correspond to lower power consumption, therefore...). (Typically negative or close to zero).

[0044] In actual calculations, The physical meaning is: at the Pareto front, the amount of signal-to-noise ratio loss per unit of power consumption corresponding to the transition from a node with a higher signal-to-noise ratio to an adjacent node with a lower signal-to-noise ratio. If The value is lower than the preset marginal efficiency threshold. This indicates that compared to its neighboring high signal-to-noise ratio (SNR) nodes, the power consumption reduction of this node is extremely limited, while the SNR loss is relatively significant. In other words, this node is in a region of low "cost-effectiveness" on the Pareto front and is considered a redundant node, which should be removed. (Preset marginal efficiency threshold) The value of needs to be set according to the relative importance of signal-to-noise ratio and power consumption in the specific application scenario, and its dimensions are similar to They are consistent, both being the change in signal-to-noise ratio per unit power consumption.

[0045] The operation of removing redundant nodes is performed pair by pair along the sorted sequence, calculating for each pair of adjacent nodes. , all In this case, the node with the lowest signal-to-noise ratio in the sequence (i.e., the first node) will be selected. (Number of nodes) are marked as redundant and removed from the Pareto front grid node set. It is important to note that after the removal operation, the adjacency relationships of the remaining nodes in the sequence change. Therefore, it is necessary to recalculate the local marginal substitution rate between adjacent nodes in the updated sequence and perform the removal judgment again until the local marginal substitution rate of all adjacent node pairs in the sequence is not less than [amount missing]. Until, or until only one node remains in the sequence.

[0046] After the aforementioned local marginal substitution rate screening, the nodes retained in the Pareto front grid node set all meet the following condition: the ratio of the power reduction to the signal-to-noise ratio loss between adjacent nodes is not less than a preset marginal efficiency threshold. In other words, each step of power reduction is accompanied by a sufficiently significant power gain, rather than exchanging a large signal-to-noise ratio loss for a small power saving. This screening mechanism effectively removes nodes with low practical engineering value from the Pareto front, making the final retained Pareto front more concise and effective, facilitating the selection operation during subsequent weighted comprehensive score calculation, and also reducing the risk of misleading selection due to redundant nodes.

[0047] In practical engineering implementation, since the joint regulation strategy space is a discrete grid structure, the number of nodes on the Pareto front is usually limited, the computation of the local marginal substitution rate is small, and the entire non-dominated sorting and redundant node removal process can be completed in real time on the embedded controller without significantly affecting the response speed of the laser ranging system.

[0048] Acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with a preset target SNR, and generate an SNR deviation, including: The time-domain waveform of the acquired echo pulse signal is subjected to continuous non-overlapping segmented windowing processing according to a preset time window length to generate multiple time-domain signal windows of equal length; For each time-domain signal window, the noise floor segment before the arrival of the echo pulse and the signal superposition segment after the arrival of the echo pulse are extracted. The mean of the squares of the sampling point amplitudes in the signal superposition segment is calculated as the mean signal power, and the mean of the squares of the sampling point amplitudes in the noise floor segment is calculated as the mean noise power. The ratio of the mean signal power to the mean noise power is used as the signal-to-noise ratio estimate for that window. The signal-to-noise ratio (SNR) estimates of all time-domain signal windows are sorted in ascending order, and the median of the sorted values ​​is taken as the actual echo SNR, so as to eliminate abnormal deviations in the SNR estimates of individual windows caused by instantaneous illumination interference. The difference between the actual echo signal-to-noise ratio and the preset target signal-to-noise ratio is calculated, and the absolute value of the difference is taken as the signal-to-noise ratio deviation.

[0049] After selecting the Pareto optimal parameter combination and driving the laser emitting unit and echo signal processing unit to complete the initial parameter adjustment, it is necessary to accurately estimate the echo signal-to-noise ratio (SNR) under actual operating conditions to provide reliable feedback for subsequent closed-loop correction. Due to the complex and variable ambient lighting conditions in real-world applications, SNR estimates obtained from single sampling or short-term observations are often affected by instantaneous strong light interference, background radiation fluctuations, or sudden changes in circuit noise, leading to significant deviations in the estimation results. Therefore, a robust SNR estimation method based on multi-window segmented statistics is adopted to process the acquired echo pulse signals.

[0050] The time-domain waveform of the acquired echo pulse signal is processed according to a preset time window length. (Unit: number of sampling points) Perform continuous, non-overlapping segmented windowing processing. Specifically, assume the acquired complete time-domain waveform contains a total of With a sampling point, it can be divided into Each window contains a time-domain signal window of equal length. Each window contains consecutive sampling points with no overlap between adjacent windows, ensuring that the statistical information contained in each window is independent. Window length The selection needs to take into account both the laser pulse repetition frequency and the target distance range: It should be long enough to simultaneously include the noise floor section and the signal superposition section within a single window, but it should not be too long to avoid mixing multiple independent pulse echoes into the same window. It is usually determined by rounding down the number of sampling points after converting the time delay corresponding to the nominal maximum ranging range of the system.

[0051] For each time-domain signal window, the noise floor segment before the arrival of the echo pulse and the signal superposition segment after the arrival of the echo pulse are extracted. The noise floor segment refers to the pure background noise signal segment collected by the receiver before the laser pulse reaches the target and forms an echo. The amplitude of the sampling points in this segment is mainly composed of the superposition of ambient background light, detector dark current, and preamplifier thermal noise. The signal superposition segment refers to the sampling point segment after the laser echo pulse arrives and is superimposed with the background noise. For the first... Time-domain signal window Assume there are a total of [number] noise base sections. The sampling point, its first The amplitude of each sampling point is There are a total of ; signals superimposed within the section The sampling point, its first The amplitude of each sampling point is .

[0052] Mean signal power Defined as the arithmetic mean of the squares of the amplitudes of all sampling points within the signal superposition section, i.e.: ; noise power mean Defined as the arithmetic mean of the squares of the amplitudes of all sampling points within the noise floor region, i.e.: ; The ratio of the mean signal power to the mean noise power is used as the first... Signal-to-noise ratio estimate for each window ,Right now: ; The above calculation method defines the signal-to-noise ratio (SNR) based on the power ratio, which is consistent with the link budget analysis framework commonly used in laser ranging systems, facilitating direct comparison with the preset target SNR. It is important to note that the amplitude in the signal superposition section includes both echo signal components and noise components; therefore... In essence, it is the average total power of the signal and noise superimposed, rather than the average power of the pure signal. Under conditions of high signal-to-noise ratio, the noise power is negligible relative to the signal power, and the above ratio is approximately equal to the true signal-to-noise ratio. When the signal-to-noise ratio is low, there is a certain overestimation, but this deviation does not have a substantial impact on the direction judgment of closed-loop control. Moreover, this estimation method has the advantages of simple calculation and strong real-time performance in engineering implementation.

[0053] After getting all The signal-to-noise ratio estimates for each time-domain signal window After that, The estimated values ​​are sorted in ascending order to form an ordered sequence. The median of this ordered sequence is taken as the final actual echo signal-to-noise ratio. .when When the number is odd, the median is the th number after sorting. One element; when When the median is even, the median is the second-order number after sorting. The and the first The median is used as the arithmetic mean of the elements. Instead of the arithmetic mean, the median is used as the summary statistic because it has a natural ability to resist abnormal deviations in the signal-to-noise ratio (SNR) estimate of individual windows caused by transient illumination interference. Under the influence of strong direct sunlight, flickering vehicle headlights, or other instantaneous strong light sources, the signal superposition section of a few windows may experience saturation or abnormal amplitude increases, resulting in an SNR estimate for that window that is much higher than normal. Similarly, occasional circuit interference pulses may cause an abnormal increase in the amplitude of the noise floor section of individual windows, resulting in an abnormally low SNR estimate for that window. The median statistic has good robustness to both types of outliers. As long as the number of outlier windows does not exceed 50% of the total number of windows, the median estimate is unaffected, thus ensuring the robustness of the actual echo SNR estimate.

[0054] To obtain the actual echo signal-to-noise ratio Then, it is compared with the preset target signal-to-noise ratio. By comparing the two, calculating the difference and taking the absolute value, the signal-to-noise ratio deviation is obtained. ,Right now: ; Preset target signal-to-noise ratio It is a design benchmark value determined comprehensively based on the system ranging accuracy index, false alarm probability requirements and detection probability requirements, and is usually written into the controller's configuration register during the system initialization phase. This reflects the degree of deviation between the current actual working state and the design goals: when When the preset tolerance threshold is exceeded, it indicates that the current parameter combination can no longer meet the system performance requirements, and a closed-loop correction process needs to be triggered; when When convergence reaches within the tolerance threshold, it indicates that the system has reached a stable optimal operating state, and the ranging task can continue to be performed while maintaining the current parameters. Taking the absolute value ensures that the deviation is always non-negative, facilitating a simple comparison with the tolerance threshold, and also allowing subsequent gradient direction correction steps to be based on... and The magnitude relationship determines the direction of parameter adjustment, that is, when It is sometimes necessary to increase the transmit power or increase the amplifier gain. In such cases, the power consumption of the transmitter can be appropriately reduced to achieve energy-saving optimization.

[0055] In the entire signal-to-noise ratio estimation and bias calculation process, the segmented window length Length of noise floor section Length of the segment superimposed with the signal All parameters must be pre-calibrated based on the specific system's laser pulse width, sampling rate, and maximum ranging range, and kept fixed during system operation to ensure consistency in statistical caliber across windows. Furthermore, This must be strictly satisfied to prevent the noise base section and the signal superposition section from overlapping in the time domain, which would affect the estimation accuracy.

[0056] When the signal-to-noise ratio (SNR) deviation exceeds a preset tolerance threshold, the SNR deviation is used as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold, including: Taking the grid node where the currently effective transmission parameters and reception parameters are located in the joint adjustment strategy space as the current node, obtain the expected echo signal-to-noise ratio (SNR) of the current node and all adjacent grid nodes, and add the SNR deviation to the expected echo SNR of the current node to obtain the target SNR expected value. Among the neighboring grid nodes of the current node, the grid node whose expected echo signal-to-noise ratio is closest to the expected target signal-to-noise ratio is selected as the target correction node, and the transmission parameters and reception parameters corresponding to the target correction node are used as the parameters after one correction. The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are re-acquired to calculate the signal-to-noise ratio. The deviation between the newly obtained signal-to-noise ratio and the preset target signal-to-noise ratio is then calculated. Repeat the target node screening and parameter correction loop until the signal-to-noise ratio deviation converges to within the preset tolerance threshold.

[0057] When the signal-to-noise ratio (SNR) deviation exceeds a preset tolerance threshold, iterative corrections need to be made to the currently effective transmit and receive parameters in the joint adjustment strategy space, so that the actual echo SNR gradually approaches the preset target SNR. The entire correction process uses grid nodes as the basic operation unit, and achieves closed-loop adaptive adjustment of parameters by gradually jumping along the SNR gradient direction in the discretized strategy space.

[0058] The current node is defined as the grid node corresponding to the currently effective transmit parameters (peak power, pulse width) and receive parameters (detection threshold, amplifier gain) in the joint adjustment strategy space. In the joint regulation policy space, each grid node and its neighbors in each dimension form a neighborhood set, denoted as . This set contains all neighboring nodes that differ by one step in both the transmit parameter dimension and the receive parameter dimension. Get the current node. and The expected echo signal-to-noise ratio of each of the adjacent grid nodes is pre-stored. These expected values ​​have been filled in by simulation modeling or historical statistical data during the strategy space construction stage and can be directly read from the table without real-time calculation, thus ensuring the real-time performance of the correction response.

[0059] The signal-to-noise ratio deviation obtained from the current measurement With the current node The corresponding expected echo signal-to-noise ratio Adding them together, we obtain the expected signal-to-noise ratio value required for this round of correction. ,Right now: in The expected echo signal-to-noise ratio (SNR) of the current node is pre-stored in the policy space. This represents the difference between the actual echo signal-to-noise ratio and the preset target signal-to-noise ratio. The physical meaning of this calculation method is that when the current actual signal-to-noise ratio is low, If the value is positive, the target signal-to-noise ratio (SNR) expectation is higher than the current node's expectation, driving the parameter to jump to a neighboring node with a higher SNR; when the actual SNR is higher, When the value is negative, the driving parameters are moderately adjusted towards a more efficient power consumption direction, thereby ensuring ranging performance while taking into account energy consumption constraints.

[0060] In the neighborhood set In the process, traverse all adjacent nodes and calculate the expected echo signal-to-noise ratio (SNR) of each adjacent node and the expected target SNR. The absolute difference between them is used to select the node with the smallest absolute difference as the target correction node, denoted as . If there are multiple nodes and If the differences are the same, the node with the lower expected laser emission power consumption is selected first to achieve the constraint of minimizing power consumption. Target correction node. The corresponding combination of transmission parameters (peak power) Pulse width ) and the combination of received parameters (detection threshold) Amplifier gain This parameter output, after correction in this round, is written into the control registers of the laser emitting unit and the echo signal processing unit respectively, completing one parameter jump.

[0061] The laser is driven to emit a new round of ranging pulses using the corrected parameters, and the acquired echo pulses are reprocessed according to the corrected detection threshold and amplifier gain. Following the time-domain windowing method consistent with the aforementioned implementation, the newly acquired echo pulse waveform is windowed, the signal-to-noise ratio (SNR) estimate for each window is calculated, and the new actual echo SNR is obtained using the median statistic. .Will Signal-to-noise ratio compared to preset target The difference is calculated again to obtain the updated signal-to-noise ratio bias. .

[0062] judge Has convergence been achieved within the preset tolerance threshold? If convergence has occurred, stop iterating and proceed to the next node. The corresponding parameter combination is used as the final effective parameters and enters the formal ranging process. If convergence has not yet occurred, then... Update to the new current node ,by For the new feedback, the complete loop process of calculating the target SNR expectation, selecting neighboring nodes, and parameter switching is repeated. Each iteration starts from the current node and performs a single-step jump within its first-order neighborhood, rather than a global re-search. This makes the computational cost of each round of correction fixed and extremely low, and the convergence speed of the iteration mainly depends on the grid resolution of the policy space and the steepness of the SNR gradient.

[0063] In practical engineering applications, to prevent the correction process from diverging due to drastic changes in ambient light or abnormal target reflectivity, a maximum limit on the number of iterations is set. When the number of iterations reaches If the signal-to-noise ratio deviation has not yet converged, the iteration loop is forcibly terminated and an abnormal state flag is triggered. The parameter combination corresponding to the current optimal node (i.e. the node with the smallest absolute value of the deviation between the actual echo signal-to-noise ratio and the preset target signal-to-noise ratio in each iteration) is output as the emergency effective parameter. At the same time, the convergence failure event is reported to the upper-level control logic so as to perform system-level abnormal handling or re-trigger the initial parameter search process.

[0064] The aforementioned gradient direction correction mechanism based on grid neighborhood jumping essentially performs a constrained local greedy search in the discretized parameter space, moving one grid step in each step along the direction of the maximum gradient of the expected signal-to-noise ratio. Since the Pareto front has been pre-calculated in the joint adjustment strategy space during the initialization phase, and the expected values ​​of signal-to-noise ratio and power consumption at the grid nodes have been filled, each iteration only needs to look up the table to compare the expected values ​​of the neighboring nodes, without having to re-execute the multi-objective optimization search. This greatly reduces the computational burden of real-time control, enabling the method to run stably at a high refresh rate on embedded laser rangefinders, meeting the requirements for adaptive parameter adjustment response speed in dynamic scenarios.

[0065] The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are processed and their arrival time detected using the corrected reception parameters to calculate the target distance, including: The time-domain sampling sequence of the echo pulse is subjected to matched filtering with a preset transmitted pulse reference waveform. The signal-to-noise ratio of the echo pulse response under noise background is enhanced by sliding cross-correlation operation, and the enhanced echo pulse envelope is extracted from the cross-correlation result. A trigger amplitude discrimination level is set on the rising edge of the enhanced echo pulse envelope. The trigger amplitude discrimination level is the product of the current noise floor root mean square value and a preset threshold multiple. The arrival time of the echo pulse is determined by the intersection of the trigger amplitude discrimination level and the rising edge of the envelope. The emission time of the laser pulse is obtained and the time-of-flight difference between the arrival time of the echo pulse and the emission time is calculated. A timing compensation amount associated with the constant false alarm detection threshold offset is applied to the time-of-flight difference to compensate for the arrival time detection drift caused by the threshold offset. The product of the compensated time difference and the speed of light constant is taken as the target distance between the laser ranging system and the target.

[0066] After the parameter iterations in the aforementioned stages have converged, the laser is driven to output ranging pulses using the corrected emission parameters. Simultaneously, the echo signal processing link is configured using the corrected reception parameters, entering the range calculation stage. The echo pulses are pre-amplified and converted from analog to digital to form a time-domain sampling sequence. This sequence contains the superposition of target reflected echoes, background light noise, and circuit thermal noise. To effectively suppress noise interference and accurately locate the arrival time of the echo pulses, a matched filtering method is used to process the time-domain sampling sequence.

[0067] The core of matched filtering lies in performing a sliding cross-correlation operation between the time-domain sampled sequence of the echo pulse and a preset transmitted pulse reference waveform. The transmitted pulse reference waveform is calibrated and acquired during system initialization and stored in a parameter library; its shape reflects the typical output waveform characteristics of the laser under the current pulse width and peak power settings. The physical meaning of the sliding cross-correlation operation is to calculate the similarity between the echo sampled sequence and the reference waveform at each time shift position. When the two are aligned on the time axis, the cross-correlation output reaches its peak. Let the echo time-domain sampled sequence be... The preset transmit pulse reference waveform is Then the sliding cross-correlation result Represented as ,in For time-shift index, This is the sampling point index. The cross-correlation operation in the frequency domain is equivalent to the inverse transform of the conjugate product of the echo signal spectrum and the reference waveform spectrum. In practice, a Fast Fourier Transform is used to accelerate the calculation and reduce computational latency.

[0068] After the cross-correlation operation is completed, the cross-correlation results are analyzed. Envelope detection was performed to extract the enhanced echo pulse envelope. Envelope detection can be achieved by taking the absolute value of the cross-correlation result and then low-pass filtering, or by obtaining the analytic signal using Hilbert transform and then taking the modulus value. After matched filtering, the signal-to-noise ratio of the echo pulse response in a noisy background is significantly improved compared to the original time-domain signal. The improvement is theoretically proportional to the time-bandwidth product of the reference waveform, which significantly improves the accuracy and reliability of subsequent arrival time detection.

[0069] Enhanced echo pulse envelope The arrival time of the echo pulse is determined using a leading-edge trigger amplitude discrimination method. Specifically, the trigger amplitude discrimination level is set in the rising edge segment of the envelope. Its value is the current noise floor root mean square value. Multiple of preset threshold The product of, i.e. Root mean square value of noise floor The threshold is estimated in real time from the noise floor section of the current echo acquisition window. The calculation method is consistent with the noise power estimation method in the aforementioned signal-to-noise ratio estimation stage, ensuring that the threshold setting dynamically matches the actual noise level. Preset threshold multiple. The system's tolerance for false alarm probability is pre-configured, with a typical value range of 3 to 5 times. Under strong background lighting conditions, the value can be appropriately increased to reduce the false alarm rate.

[0070] Trigger amplitude level The time of intersection with the rising edge of the envelope is determined as the arrival time of the echo pulse. To achieve sub-sampling interval accuracy for time estimation, a linear interpolation method is used to refine the data points between rising edges near the intersection point; that is, when the envelope value first exceeds... Between two adjacent sampling points, the precise crossover time is interpolated according to a linear proportional relationship. Let the envelope be at time... The value at that location is At any moment The value at that location is And satisfy The arrival time obtained by interpolation is The unit is the sampling interval. This interpolation method has low computational cost and strong real-time performance in practical engineering, making it suitable for high-speed execution on embedded processors.

[0071] Obtain the emission time of the laser pulse Then, the time difference between the arrival time and the transmission time of the echo pulse is calculated. Because the arrival time is detected using a leading-edge triggering amplitude discrimination method, the trigger point is not the true peak or center of gravity of the echo pulse, but rather the intersection of the rising edge of the envelope and the threshold level. The position of this intersection point undergoes a systematic shift with changes in the threshold level, known as "walk error." When the echo pulse amplitude is strong, the rising edge is steeper, the trigger point is closer to the pulse leading edge, and the measured time of flight is smaller; when the echo pulse amplitude is weak, the rising edge slope is lower, the trigger point is relatively shifted backward, and the measured time of flight is larger. To eliminate this systematic error, a timing compensation amount related to the constant false alarm rate (CFAR) detection threshold offset needs to be applied to the time of flight difference. .

[0072] Timed compensation amount The determination is based on the threshold-travel time error compensation table established during the system calibration phase. During the calibration phase, a standard target at a known distance is used as the target, and different threshold multiples are applied. Flight time was measured under different echo amplitude conditions and compared with the actual flight time. Systematic deviations under each condition were recorded, and a compensation lookup table was created. During actual ranging, the current effective threshold multiple was used. And from the echo amplitude level estimated from the enhancement envelope, the corresponding timing compensation amount can be obtained by looking up a table or interpolating. This is then subtracted from the original flight time difference to obtain the compensated flight time difference. This compensation mechanism effectively eliminates arrival time detection drift caused by threshold setting, improving the consistency and stability of ranging accuracy, especially in scenarios with large differences in target reflectivity or rapid changes in background lighting conditions.

[0073] The compensated flight time difference With the speed of light constant Half of the product is taken as the target distance between the laser ranging system and the target. ,Right now The speed of light constant is taken as... In practical engineering applications, the speed of light (m / s) is typically corrected using the standard atmospheric refractive index to further improve the absolute accuracy of long-distance measurements. (Target distance) The calculated results are filtered and smoothed before being output. Kalman filtering or moving average filtering can be used to fuse multiple consecutive ranging results, suppressing random errors in a single measurement, and finally outputting a stable and reliable target distance estimate. The entire distance calculation process, from echo acquisition to distance output, is completed within a real-time processing framework, meeting the comprehensive requirements of laser ranging systems for response speed and accuracy.

[0074] A second aspect of the present invention provides a laser ranging signal-to-noise ratio adaptive control system based on AI illumination information, comprising: The parameter optimization unit is used to acquire ambient lighting information of the laser ranging target area and extract features to generate a signal-to-noise ratio degradation prediction value under the current lighting conditions. Using the signal-to-noise ratio degradation prediction value as input, it searches in the preset joint adjustment strategy space of transmission parameters and receiving parameters to maximize the echo signal-to-noise ratio and minimize the laser transmission power consumption. The parameter adjustment unit is used to adjust the peak power and pulse width of the laser emitting unit according to the emission parameters in the Pareto optimal combination, and to adjust the detection threshold and amplifier gain of the echo signal processing unit according to the receiving parameters in the Pareto optimal combination. The feedback correction unit is used to acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with a preset target SNR, generate an SNR deviation, and when the SNR deviation exceeds a preset tolerance threshold, use the SNR deviation as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold. The ranging calculation unit is used to drive the laser to emit ranging pulses with the corrected emission parameters, and to perform signal processing and arrival time detection on the echo pulses with the corrected reception parameters to calculate the target distance.

[0075] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0076] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0077] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A laser ranging signal-to-noise ratio adaptive control method based on AI illumination information, characterized in that, include: The system acquires ambient illumination information of the laser ranging target area and extracts its features to generate a predicted signal-to-noise ratio (SNR) degradation value under the current illumination conditions. Using this predicted SNR value as input, it searches within a preset joint adjustment strategy space for transmission and reception parameters to find the Pareto optimal combination of transmission and reception parameters that maximizes the echo SNR while minimizing laser transmission power consumption. This includes: A two-dimensional orthogonal gridded joint adjustment strategy space is constructed, with the peak power and pulse width of the laser emitting unit constituting the emission parameter dimension, and the detection threshold offset and amplifier gain compensation coefficient of the echo signal processing unit constituting the receiving parameter dimension. Each grid intersection node corresponds to a set of pre-calibrated expected echo signal-to-noise ratio and expected laser emission power consumption. Using the predicted signal-to-noise ratio degradation value and the predicted confidence score as search inputs, all grid intersection nodes are traversed and evaluated in the joint adjustment strategy space. With the goal of maximizing the expected value of echo signal-to-noise ratio and minimizing the expected value of laser emission power consumption, the Pareto front grid node set is selected by non-dominated sorting. In the Pareto front grid node set, a weighted comprehensive score is calculated for each candidate grid node using a preset signal-to-noise ratio priority weighting factor, and the transmit and receive parameters corresponding to the grid node with the highest weighted comprehensive score are selected as the Pareto optimal combination. The peak power and pulse width of the laser emitting unit are adjusted according to the emission parameters in the Pareto optimal combination, and the detection threshold and amplifier gain of the echo signal processing unit are adjusted according to the receiving parameters in the Pareto optimal combination. Acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with the preset target SNR, and generate an SNR deviation. When the SNR deviation exceeds a preset tolerance threshold, use the SNR deviation as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold. The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are processed and their arrival time is detected using the corrected reception parameters to calculate the target distance.

2. The method according to claim 1, characterized in that, The system acquires ambient illumination information of the laser ranging target area and extracts its features to generate a predicted signal-to-noise ratio degradation value under the current illumination conditions, including: Multi-dimensional illumination feature separation is performed on the ambient illumination information to extract the average illumination intensity feature, illumination spatial contrast feature, and illumination temporal flicker frequency feature. The average light intensity feature, the spatial contrast of light, and the temporal flicker frequency of light are concatenated to generate a multi-dimensional joint feature vector of light. The multi-dimensional illumination joint feature vector is input into a pre-trained cascaded nonlinear mapping network. The cascaded nonlinear mapping network contains multiple cascaded feature compression coding layers and nonlinear activation transformation layers. Through layer-by-layer dimensionality reduction compression and nonlinear mapping, the multi-dimensional illumination joint feature vector is transformed into a signal-to-noise ratio degradation prediction value under the current illumination conditions. The prediction confidence score corresponding to the signal-to-noise ratio degradation prediction value is generated synchronously at the output layer of the cascaded nonlinear mapping network. The prediction confidence score is used to characterize the reliability of the prediction of the signal-to-noise ratio degradation prediction value under different illumination distribution patterns.

3. The method according to claim 1, characterized in that, The set of Pareto front grid nodes was selected through non-dominated sorting, including: Using the expected echo signal-to-noise ratio as the first optimization objective dimension and the expected laser emission power consumption as the second optimization objective dimension, the pairwise dominance relationship of all grid intersection nodes in the joint adjustment strategy space is determined: if the first node is not inferior to the second node in both the expected echo signal-to-noise ratio and the expected laser emission power consumption, and is strictly superior to the second node in at least one dimension, then the first node is determined to dominate the second node. Extract all grid nodes in the joint regulation strategy space that no other grid node can dominate, and form the Pareto front grid node set; In the Pareto front grid node set, the nodes are sorted from high to low according to the expected echo signal-to-noise ratio. The ratio of the decrease in the expected echo signal-to-noise ratio between adjacent grid nodes in the sorted sequence to the decrease in the expected laser emission power consumption is calculated as the local marginal substitution rate. Grid nodes with a local marginal substitution rate lower than a preset marginal efficiency threshold are removed to eliminate redundant nodes in the Pareto front that have limited power consumption reduction but significant signal-to-noise ratio loss.

4. The method according to claim 1, characterized in that, Acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with a preset target SNR, and generate an SNR deviation, including: The time-domain waveform of the acquired echo pulse signal is subjected to continuous non-overlapping segmented windowing processing according to a preset time window length to generate multiple time-domain signal windows of equal length; For each time-domain signal window, the noise floor segment before the arrival of the echo pulse and the signal superposition segment after the arrival of the echo pulse are extracted. The mean of the squares of the sampling point amplitudes in the signal superposition segment is calculated as the mean signal power, and the mean of the squares of the sampling point amplitudes in the noise floor segment is calculated as the mean noise power. The ratio of the mean signal power to the mean noise power is used as the signal-to-noise ratio estimate for that window. The signal-to-noise ratio (SNR) estimates of all time-domain signal windows are sorted in ascending order, and the median of the sorted values ​​is taken as the actual echo SNR, so as to eliminate abnormal deviations in the SNR estimates of individual windows caused by instantaneous illumination interference. The difference between the actual echo signal-to-noise ratio and the preset target signal-to-noise ratio is calculated, and the absolute value of the difference is taken as the signal-to-noise ratio deviation.

5. The method according to claim 4, characterized in that, When the signal-to-noise ratio (SNR) deviation exceeds a preset tolerance threshold, the SNR deviation is used as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold, including: Taking the grid node where the currently effective transmission parameters and reception parameters are located in the joint adjustment strategy space as the current node, obtain the expected echo signal-to-noise ratio (SNR) of the current node and all adjacent grid nodes, and add the SNR deviation to the expected echo SNR of the current node to obtain the target SNR expected value. Among the neighboring grid nodes of the current node, the grid node whose expected echo signal-to-noise ratio is closest to the expected target signal-to-noise ratio is selected as the target correction node, and the transmission parameters and reception parameters corresponding to the target correction node are used as the parameters after one correction. The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are reacquired to calculate the signal-to-noise ratio. The deviation between the newly obtained signal-to-noise ratio and the preset target signal-to-noise ratio is then calculated. Repeat the target node screening and parameter correction loop until the signal-to-noise ratio deviation converges to within the preset tolerance threshold.

6. The method according to claim 1, characterized in that, The laser is driven to emit ranging pulses using the corrected emission parameters, and the echo pulses are processed and their arrival times detected using the corrected reception parameters to calculate the target distance, including: The time-domain sampling sequence of the echo pulse is subjected to matched filtering with a preset transmitted pulse reference waveform. The signal-to-noise ratio of the echo pulse response under noise background is enhanced by sliding cross-correlation operation, and the enhanced echo pulse envelope is extracted from the cross-correlation result. A trigger amplitude discrimination level is set on the rising edge of the enhanced echo pulse envelope. The trigger amplitude discrimination level is the product of the current noise floor root mean square value and a preset threshold multiple. The arrival time of the echo pulse is determined by the intersection of the trigger amplitude discrimination level and the rising edge of the envelope. The emission time of the laser pulse is obtained and the time difference between the arrival time of the echo pulse and the emission time is calculated. A timing compensation amount associated with the constant false alarm detection threshold offset is applied to the time difference to compensate for the arrival time detection drift caused by the threshold offset. The product of the compensated time difference and the speed of light constant is taken as the target distance between the laser ranging system and the target.

7. A laser ranging signal-to-noise ratio adaptive control system based on AI illumination information, used to implement the method as described in any one of claims 1-6, characterized in that, include: The parameter optimization unit is used to acquire ambient lighting information of the laser ranging target area and extract features to generate a signal-to-noise ratio degradation prediction value under the current lighting conditions. Using the signal-to-noise ratio degradation prediction value as input, it searches in the preset joint adjustment strategy space of transmission parameters and receiving parameters to maximize the echo signal-to-noise ratio and minimize the laser transmission power consumption. The parameter adjustment unit is used to adjust the peak power and pulse width of the laser emitting unit according to the emission parameters in the Pareto optimal combination, and to adjust the detection threshold and amplifier gain of the echo signal processing unit according to the receiving parameters in the Pareto optimal combination. The feedback correction unit is used to acquire the adjusted echo pulse signal, calculate the actual echo signal-to-noise ratio (SNR), compare the actual echo SNR with a preset target SNR, generate an SNR deviation, and when the SNR deviation exceeds a preset tolerance threshold, use the SNR deviation as a feedback quantity to perform gradient direction correction on the transmit parameters and the receive parameters in the joint adjustment strategy space until the SNR deviation converges to within the preset tolerance threshold. The ranging calculation unit is used to drive the laser to emit ranging pulses with the corrected emission parameters, and to perform signal processing and arrival time detection on the echo pulses with the corrected reception parameters to calculate the target distance.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images