Multi-modal radar detection optimization method based on humidity driving

Abstract

The invention discloses a multi-modal radar detection optimization method based on humidity driving, and the method comprises the steps: defining fuzzy logic, and constructing a membership function of humidity; humidity data of a humidity sensor and millimeter wave radar and laser radar data are collected; obtaining the membership degree of the humidity data through a membership degree function, and determining the type of the humidity; constructing a fuzzy rule base and a fuzzy logic controller, and determining weight distribution of the radar system under different humidity category conditions; obtaining weight distribution of the radar system; performing data fusion on the millimeter wave radar data and the laser radar data according to weight distribution; and constructing a reward function to optimize the performance of the fused data, and obtaining an optimal radar detection scheme. By monitoring the environment humidity in real time and dynamically adjusting the working mode and weight distribution of the radar system, more accurate environment perception is achieved, acquisition and processing of sensor data are optimized, and therefore the reliability and performance of the whole system are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent sensing and control, relates to multi-modal radar detection, and specifically relates to a humidity-driven multi-modal radar detection optimization method. Background Art

[0002] With the rapid development of intelligent transportation and autonomous driving technologies, environmental perception systems have become particularly critical. Millimeter-wave radar, with its excellent penetration and anti-interference capabilities, can maintain its performance even in adverse weather conditions; while lidar, with its high-precision spatial resolution, provides strong support for precise measurement. Multimodal radar systems combine the advantages of both and provide detailed information about the vehicle's surroundings, including key parameters such as distance, speed, and shape.

[0003] However, the multimodal radar systems in the existing technology often lack comprehensive consideration of environmental factors. Especially under conditions with large changes in humidity, the propagation characteristics of the radar signal will be affected, which will lead to a decrease in radar detection accuracy and make the fusion results unable to accurately reflect the actual environmental conditions.

[0004] Therefore, a new technical solution is needed to solve this problem. Summary of the invention

[0005] Purpose of the invention: In order to overcome the deficiencies in the prior art, a humidity-driven multimodal radar detection optimization method is provided, which monitors the ambient humidity in real time and dynamically adjusts the working mode and weight distribution of the radar system. The present invention aims to achieve more accurate environmental perception, optimize the collection and processing of sensor data, and thus improve the reliability and performance of the overall system.

[0006] Technical solution: To achieve the above purpose, the present invention provides a humidity-driven multi-modal radar detection optimization method, comprising the following steps:

[0007] S1: define fuzzy logic and construct the membership function of humidity;

[0008] S2: collects humidity data from humidity sensors, millimeter wave radar and lidar data;

[0009] S3: Obtain the membership of the humidity data through the membership function to determine the category to which the humidity belongs;

[0010] S4: construct fuzzy rule base and fuzzy logic controller to determine the weight distribution of radar system under different humidity categories;

[0011] S5: Obtaining a weight distribution of the radar system according to the humidity category determined in step S3;

[0012] S6: According to the weight distribution in step S5, the millimeter wave radar data and the laser radar data are fused;

[0013] S7: Construct a reward function to optimize the performance of the fused data and obtain the best radar detection solution.

[0014] Furthermore, the step S1 is specifically as follows:

[0015] The fuzzy set of humidity is defined as {low, medium, high}, and a membership function is defined for each fuzzy set element, which maps the actual measured humidity value to the interval [0, 1], indicating the degree to which the value belongs to the corresponding fuzzy set;

[0016] According to the above requirements, the membership function μ of humidity H (h) is defined as follows:

[0017]

[0018] Among them, H low , H mid and H high Represent the low threshold, middle threshold and high threshold of humidity respectively, and e is a parameter that controls the smoothness of the membership function. The humidity on rainy days is usually between 80% and 90%, while the humidity on foggy days is usually greater than 90%. Therefore, the low threshold can be set to 82%, the middle threshold to 85%, and the high threshold to 88%.

[0019] Furthermore, in step S2, the collected humidity data is normalized, and the humidity data is defined as h, and the normalized humidity data is defined as h'.

[0020] Furthermore, the rules of the fuzzy rule base in step S4 are as follows:

[0021] If the humidity is "low", the weight of the lidar is high; if the humidity is "medium", the weights of the millimeter-wave radar and the lidar are equal; if the humidity is "high", the weight of the millimeter-wave radar is high.

[0022] Furthermore, the fuzzy logic controller in step S4 is used to implement the reasoning mechanism of the fuzzy rule base, and the fuzzy logic controller determines the weight distribution of the radar system through the fuzzy rule base according to the humidity data.

[0023] Furthermore, the design formula of the fuzzy logic controller in step S4 is expressed as:

[0024] W R =f(μ H (h'))

[0025] Among them, W Ris the weight vector of the radar system, and f represents the reasoning mechanism of the fuzzy logic controller;

[0026] Apply fuzzy logic controller for reasoning to determine the weight W1 of millimeter wave radar and the weight W2 of lidar:

[0027] W1=β+(1-β)·(1-α·μ H (h') 2

[0028] W2=1-W1

[0029] Where α is the adjustment factor that controls the rate at which the weight decreases as humidity increases, and β is the offset that ensures that the weight of the mmWave radar does not drop to zero even when humidity is high.

[0030] Furthermore, in step S6, in order to reduce the impact of outliers or unreliable measurements in the millimeter wave radar and lidar data, a confidence interval [L i ,U i ], where L i and U i are the lower and upper bounds of the confidence interval respectively; first, the center point μ of the confidence interval of the i-th sensor measurement value is obtained i :

[0031]

[0032] The formula for fusing millimeter-wave radar and lidar data is:

[0033]

[0034] Where N is the number of sensors, W i is the weight assigned to the ith sensor.

[0035] Furthermore, after the data is fused in step S7, there is still a situation where the correct detection ratio is low and the confidence interval width may not fully represent the complex uncertainty. Therefore, a reward function can be constructed to optimize the performance of the fused data. At the same time, the stability and energy consumption of the fused data also need to be considered.

[0036] The indicators of the reward function include accuracy, consistency, energy consumption and uncertainty. Accuracy and consistency are positive indicators of performance and are enhanced through the reward mechanism. Energy consumption and uncertainty are negative indicators and are reduced through the penalty mechanism.

[0037] Formula design of reward function:

[0038]

[0039] Among them, N 正确 is the number of correct detections, N 总 is the total number of detections, y i is the actual measured value, E 总 is the total energy consumption, T is the time, H(Z fusion ) is the uncertainty penalty, w1, w2, w3, w4 are the weights corresponding to accuracy, consistency, energy consumption and uncertainty respectively. These coefficients are used to adjust the importance of each item in the reward function and are adjusted according to system design and requirements. By adjusting each weight coefficient, these goals can be balanced to achieve the optimization of the radar detection accuracy solution.

[0040] Methods for adjusting the reward function include:

[0041] 1) Initial weight setting

[0042] w1=0.4, w2=0.3, w3=0.2, w4=0.1

[0043] These initial values ​​are based on assumptions, and actual values ​​need to be determined based on the specific needs of the system and test results;

[0044] 2) Conduct system testing and analyze test results;

[0045] 3) Based on the analysis, adjust the weights to reflect the results of system testing and the need for performance improvement:

[0046] Increase the energy consumption weight to reduce its negative impact on overall performance: w3 = 0.3

[0047] Since uncertainty has little impact on performance, slightly increase its weight: w4 = 0.2

[0048] Reduce the weights of precision and consistency because they already perform well: w1=0.25, w2=0.25

[0049] The above is the new weight setting;

[0050] 4) Verify the adjustment effect

[0051] Test the system again with the new weight settings to verify performance improvement;

[0052] 5) Continuous Optimization

[0053] Based on the test results and feedback from actual applications, the weights are continuously adjusted until the optimal configuration is found.

[0054] By testing in simulation or actual environment, the performance of the learned strategy in different situations is evaluated, and the reward function is adjusted to improve the strategy. Relying on continuous iterative learning, the radar system eventually converges to an optimal strategy that can obtain the maximum cumulative reward in most cases. This solves the radar detection accuracy optimization solution.

[0055] Beneficial effects: Compared with the prior art, the present invention monitors the ambient humidity in real time and dynamically adjusts the working mode and weight distribution of the radar system to adapt to different environmental conditions. The importance of humidity as one of the environmental factors in radar signal propagation and detection accuracy is particularly emphasized, and it is incorporated into the decision-making process of radar data fusion and weight distribution. By combining the qualitative analysis of fuzzy logic control and the quantitative optimization of the reward function. In an unmanned driving environment, the radar system can dynamically adjust its working mode according to the real-time ambient humidity data, improve the overall performance and detection accuracy of the multimodal radar system, aiming to improve the performance and reliability of the multimodal radar system under complex environmental conditions, and provide more accurate and adaptive environmental perception solutions for the fields of intelligent transportation and autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0057] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0058] like Figure 1 As shown, the present invention provides a humidity-driven multi-modal radar detection optimization method, comprising the following steps:

[0059] S1: define fuzzy logic and construct the membership function of humidity;

[0060] The fuzzy set of humidity is defined as {low, medium, high}, and a membership function is defined for each fuzzy set element, which maps the actual measured humidity value to the interval [0, 1], indicating the degree to which the value belongs to the corresponding fuzzy set;

[0061] According to the above requirements, the membership function μ of humidity H (h) is defined as follows:

[0062]

[0063] Among them, H low , H mid and Hhigh Represent the low threshold, middle threshold and high threshold of humidity respectively, and e is a parameter that controls the smoothness of the membership function. The humidity on rainy days is usually between 80% and 90%, while the humidity on foggy days is usually greater than 90%. Therefore, the low threshold can be set to 82%, the middle threshold to 85%, and the high threshold to 88%.

[0064] When the low threshold is reached, the lidar begins to be greatly affected while the performance of the millimeter-wave radar is relatively good; when the high threshold is reached, the performance of the lidar drops significantly while the performance of the millimeter-wave radar is still good.

[0065] S2: Collect humidity data from humidity sensors, millimeter wave radar and lidar data, and normalize the collected humidity data, define the humidity data as h, and the normalized humidity data as h';

[0066] S3: Obtain the membership of the humidity data through the membership function to determine the category to which the humidity belongs;

[0067] S4: construct fuzzy rule base and fuzzy logic controller to determine the weight distribution of radar system under different humidity categories;

[0068] The rules of the fuzzy rule base are as follows:

[0069] If the humidity is "low", the weight of the lidar is high; if the humidity is "medium", the weights of the millimeter-wave radar and the lidar are equal; if the humidity is "high", the weight of the millimeter-wave radar is high.

[0070] The fuzzy logic controller is used to implement the reasoning mechanism of the fuzzy rule base. The fuzzy logic controller will determine the weight distribution of the radar system through the fuzzy rule base based on the humidity data.

[0071] The design formula of the fuzzy logic controller is expressed as:

[0072] W R =f(μ H (h'))

[0073] Among them, W R is the weight vector of the radar system, and f represents the reasoning mechanism of the fuzzy logic controller;

[0074] Apply fuzzy logic controller for reasoning to determine the weight W1 of millimeter wave radar and the weight W2 of lidar:

[0075] W1=β+(1-β)·(1-α·μ H (h') 2

[0076] W2=1-W1

[0077] Where α is the adjustment factor that controls the rate at which the weight decreases as humidity increases, and β is the offset that ensures that the weight of the mmWave radar does not drop to zero even when humidity is high.

[0078] S5: Obtaining a weight distribution of the radar system according to the humidity category determined in step S3;

[0079] S6: According to the weight distribution in step S5, the millimeter wave radar data and the laser radar data are fused;

[0080] To reduce the impact of outliers or unreliable measurements in mmWave radar and lidar data, a confidence interval [L i ,U i ], where L i and U i are the lower and upper bounds of the confidence interval respectively; first, the center point μ of the confidence interval of the i-th sensor measurement value is obtained i :

[0081]

[0082] The formula for fusing millimeter-wave radar and lidar data is:

[0083]

[0084] Where N is the number of sensors, W i is the weight assigned to the ith sensor.

[0085] S7: Construct a reward function to optimize the performance of the fused data and obtain the best radar detection solution.

[0086] After data fusion, there is still a low correct detection ratio and the confidence interval width may not fully represent the complex uncertainty. Therefore, a reward function can be constructed to optimize the performance of the fused data. At the same time, the stability and energy consumption of the fused data also need to be considered.

[0087] The indicators of the reward function include accuracy, consistency, energy consumption and uncertainty. Accuracy and consistency are positive indicators of performance and are enhanced through the reward mechanism. Energy consumption and uncertainty are negative indicators and are reduced through the penalty mechanism.

[0088] Formula design of reward function:

[0089]

[0090] Among them, N 正确 is the number of correct detections, N 总 is the total number of detections, y iis the actual measured value, E 总 is the total energy consumption, T is the time, H(Z fusion ) is the uncertainty penalty, w1, w2, w3, w4 are the weights corresponding to accuracy, consistency, energy consumption and uncertainty respectively. These coefficients are used to adjust the importance of each item in the reward function and are adjusted according to system design and requirements. By adjusting each weight coefficient, these goals can be balanced to achieve the optimization of the radar detection accuracy solution.

[0091] By testing in simulation or actual environments, evaluating the performance of the learned strategy in different situations, and adjusting the reward function to improve the strategy, the radar system eventually converges to an optimal strategy through continuous iterative learning, which can obtain the maximum cumulative reward in most cases, thereby solving the radar detection accuracy optimization solution.

[0092] In this embodiment, obtaining the above optimal strategy includes the following steps:

[0093] 1) Initial weight setting

[0094] w1=0.4, w2=0.3, w3=0.2, w4=0.1

[0095] These initial values ​​are based on assumptions and actual values ​​need to be determined based on the specific requirements of your system and test results.

[0096] 2) Conduct system testing and analyze test results

[0097] According to the test results, data analysis is performed:

[0098] 1. Precision and consistency have a large impact on performance, but current precision and consistency are already high, so the marginal benefit of increasing these weights may be low.

[0099] 2. The energy consumption is high, and the impact of energy consumption on overall performance needs to be reduced.

[0100] 3) Based on the analysis, adjust the weights to reflect the results of system testing and the need for performance improvement:

[0101] 1. Increase the energy consumption weight to reduce its negative impact on overall performance: w3 = 0.3

[0102] 2. Since uncertainty has little impact on performance, slightly increase its weight: w4 = 0.2

[0103] 3. Reduce the weights of precision and consistency because they already perform well: w1 = 0.25, w2 = 0.25

[0104] The above are the new weight settings.

[0105] 4) Verify the adjustment effect

[0106] Test the system again with the new weight settings to verify that performance has improved. If energy consumption decreases and accuracy and consistency remain at acceptable levels, the weight adjustment was effective.

[0107] 5) Continuous Optimization

[0108] Based on the test results and feedback from actual applications, the weights are continuously adjusted until the optimal configuration is found.

[0109] Through the above specific processes and numerical adjustments, we can more accurately control the performance of the radar system to achieve the best detection effect.

Claims

1. A humidity-driven multi-modal radar detection optimization method, characterized in that: The steps include: S1: define fuzzy logic and construct the membership function of humidity; S2: collects humidity data from humidity sensors, millimeter wave radar and lidar data; S3: Obtain the membership of the humidity data through the membership function to determine the category to which the humidity belongs; S4: construct fuzzy rule base and fuzzy logic controller to determine the weight distribution of radar system under different humidity categories; S5: Obtaining a weight distribution of the radar system according to the humidity category determined in step S3; S6: According to the weight distribution in step S5, the millimeter wave radar data and the laser radar data are fused; S7: Construct a reward function to optimize the performance of the fused data and obtain the best radar detection solution.

2. The humidity-driven multi-modal radar detection optimization method according to claim 1 is characterized in that: The step S1 is specifically as follows: The fuzzy set of humidity is defined as {low, medium, high}, and a membership function is defined for each fuzzy set element, which maps the actual measured humidity value to the interval [0, 1], indicating the degree to which the value belongs to the corresponding fuzzy set; According to the above requirements, the membership function μ of humidity H (h) is defined as follows: Among them, H low , H mid and H high They represent the low threshold, middle threshold and high threshold of humidity respectively, and e is a parameter that controls the smoothness of the membership function.

3. The humidity-driven multi-modal radar detection optimization method according to claim 1 is characterized in that: In the step S2, the collected humidity data is normalized, and the humidity data is defined as h, and the normalized humidity data is defined as h'.

4. The humidity-driven multi-modal radar detection optimization method according to claim 2 is characterized in that: The rules of the fuzzy rule base in step S4 are as follows: If the humidity is "low", the weight of the lidar is high; if the humidity is "medium", the weights of the millimeter-wave radar and the lidar are equal; if the humidity is "high", the weight of the millimeter-wave radar is high.

5. The humidity-driven multi-modal radar detection optimization method according to claim 2 is characterized in that: The fuzzy logic controller in step S4 is used to implement the reasoning mechanism of the fuzzy rule base. The fuzzy logic controller determines the weight distribution of the radar system through the fuzzy rule base according to the humidity data.

6. The humidity-driven multi-modal radar detection optimization method according to claim 5, characterized in that: The design formula of the fuzzy logic controller in step S4 is expressed as: W R =f(μ H (h′)) Among them, W R is the weight vector of the radar system, and f represents the reasoning mechanism of the fuzzy logic controller; Apply fuzzy logic controller for reasoning to determine the weight W1 of millimeter wave radar and the weight W2 of lidar: W1=β+(1-β)·(1-α·μ H (h') 2 ) W2=1-W1 Among them, α is the adjustment coefficient used to control the rate at which the weight decreases as the humidity increases; β is the offset.

7. The humidity-driven multi-modal radar detection optimization method according to claim 1, characterized in that: In step S6, a confidence interval [L i ,U i ], where L i and U i are the lower and upper bounds of the confidence interval respectively; first, the center point μ of the confidence interval of the i-th sensor measurement value is obtained i : The formula for fusing millimeter-wave radar and lidar data is: Where N is the number of sensors, W i is the weight assigned to the ith sensor.

8. The humidity-driven multi-modal radar detection optimization method according to claim 7, characterized in that: The indicators of the reward function in step S7 include accuracy, consistency, energy consumption and uncertainty, wherein accuracy and consistency are positive indicators of performance and are enhanced through the reward mechanism, and energy consumption and uncertainty are negative indicators and are reduced through the penalty mechanism.

9. The humidity-driven multi-modal radar detection optimization method according to claim 8, characterized in that: The formula design of the reward function in step S7 is: Among them, N 正确 is the number of correct detections, N 总 is the total number of detections, y i is the actual measured value, E 总 is the total energy consumption, T is the time, H(Z fusion ) is the uncertainty penalty, w1, w2, w3, and w4 are the weights corresponding to accuracy, consistency, energy consumption, and uncertainty, respectively. These coefficients are used to adjust the importance of each item in the reward function and are adjusted according to system design and requirements.

10. The humidity-driven multi-modal radar detection optimization method according to claim 9, characterized in that: The method for adjusting the reward function in step S7 includes: 1) Initial weight setting w1=0.4, w2=0.3, w3=0.2, w4=0.1 These initial values ​​are based on assumptions, and actual values ​​need to be determined based on the specific needs of the system and test results; 2) Conduct system testing and analyze test results; 3) Based on the analysis, adjust the weights to reflect the results of system testing and the need for performance improvement: Increase the energy consumption weight to reduce its negative impact on overall performance: w3 = 0.3 Since uncertainty has little impact on performance, slightly increase its weight: w4 = 0.2 Reduce the weights of precision and consistency because they already perform well: w1=0.25, w2=0.25 and above are the new weight settings; 4) Verify the adjustment effect Test the system again with the new weight settings to verify performance improvement; 5) Continuous Optimization Based on the test results and feedback from actual applications, the weights are continuously adjusted until the optimal configuration is found.

Images