A multi-sensor fusion method for measuring blind spots of reversing radar

Through the multi-sensor fusion method, Kalman filtering and recursive least squares method are used to construct the fitting function, which solves the problem of low distance measurement accuracy in the ultrasonic sensor blind spot, and achieves accurate obstacles within the range of 0 to 5 meters. The distance measurement of objects overcomes the problem of inaccurate positioning of inertial components at low speeds.

CN114002685BActive Publication Date: 2024-11-12NANJING AUTOMOBILE GROUP CORP +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111251601.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-26
Publication Date
2024-11-12
Estimated Expiration
2041-10-26

AI Technical Summary

Technical Problem

The existing reverse radar system has low distance measurement accuracy in the close-range blind spot of ultrasonic sensors, and the positioning of inertial elements at low speeds is inaccurate, resulting in the inability to accurately measure the distance of the obstacle, especially within a range of 0 to 5 meters.

Method used

Using multi-sensor fusion method, ultrasonic radar, inertial components and wheel speed sensor data are obtained through the MCU, and fitting functions are constructed using Kalman filtering and recursive least squares method to compensate for inertial components and wheel speed sensor data, and generated in the blind spot of ultrasonic sensor Distance observations.

Benefits of technology

Accurate distance measurement in the blind spot of ultrasonic sensors, overcomes the problem of inaccurate positioning of inertial components at low speeds, and can accurately measure the distance of obstacles within a range of 0 to 5 meters, solving the reversing radar in similar logistics trucks. The problem of poor application performance in the blind spots that dock to the unloading platform.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114002685B_ABST
    Figure CN114002685B_ABST
Patent Text Reader

Abstract

The present invention relates to a method for measuring distance in the blind spot of a reversing radar using multi-sensor fusion. When a vehicle is reversing, data is collected and calculated through a distance measurement system, and finally an observation value of the distance between the logistics vehicle and the end point in the blind spot of an ultrasonic sensor is obtained. The present invention solves the problem of low accuracy in measuring distance in the blind spot of an ultrasonic sensor, and overcomes the problem of inaccurate positioning of inertial elements at low speeds. By using ultrasonic sensors to compensate for the data of the inertial elements and wheel speed sensors, it can be applied to distance observation in the blind spot of a reversing radar. Accurate distance measurement of 0 to 5 meters from obstacles is achieved through the fusion of multiple sensors, thereby solving the problem of poor application effect of reversing radar in scenarios such as logistics trucks docking at unloading platforms, which require accurate distances in blind spots.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a reversing radar blind spot ranging method, in particular to a multi-sensor fusion reversing radar blind spot ranging method, belonging to the technical field of vehicle-mounted radars. Background Art

[0002] In the current reversing assistance system, ultrasonic radar is generally used for distance measurement, combined with rearview radar and voice broadcast system to provide assistance to the driver in reversing. Compared with other non-contact distance measurement technologies, such as millimeter wave radar, laser radar, binocular camera, etc., ultrasonic sensors have the advantages of low cost, good environmental adaptability, low requirements on the computing power of the controller, simple distance measurement method, etc., and can be well applied within the distance measurement range of 0.25 to 5 meters. The detection of ultrasonic sensors includes three stages: wave emission, delay, and reception. The transit time detection method is usually used, that is, the probe starts the timer count while emitting sound waves. In the receiving stage, the sensor stops the timer count after receiving the echo with an energy amplitude greater than the set threshold for the first time, and the distance of the object that bounces the echo is calculated from this time data. The delay after the ultrasonic sensor emits the wave can reduce the interference caused by the reflection of nearby structures and instruments, but it also causes the ultrasonic sensor to detect blind spots at close range.

[0003] Wheel speed sensors generally include magnetoelectric wheel speed sensors and Hall wheel speed sensors, which are usually installed at the rear axle wheels. Taking the magnetoelectric wheel speed sensor as an example, the rotation of the wheel drives the ring gear, and the teeth and gaps on the ring gear change the magnetic resistance of the magnetic circuit through the magnetic field of the sensor, causing the induced electromotive force in the coil to change, generating pulses, which are recorded by the pulse accumulator for use by the ESP (Electronic Stability Program) system. The longitudinal speed of the vehicle can be obtained through conversion. However, since the output signal is a pulse value, the simply calculated vehicle speed is not continuous enough.

[0004] Inertial components are gyroscopes and accelerometers, which can obtain the rotational angular velocity around three axes and the acceleration in three axes, and are often used to determine the position and attitude of the vehicle. However, the acceleration obtained by inertial components at low speeds has more noise and is not very accurate. Summary of the invention

[0005] The purpose of the present invention is to propose a multi-sensor fusion reversing radar blind spot ranging method, which can obtain the distance from a stationary obstacle to a vehicle in the case of a close-range sensor blind spot.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows: a multi-sensor fusion reversing radar blind spot ranging method, characterized in that it includes the following specific steps:

[0007] Step 1: When the vehicle starts to reverse, the vehicle's MCU obtains ultrasonic radar ranging data through the CAN bus, and constructs a data set of the vehicle's reverse distance Y and time according to the starting distance length. The recorded Y value is less than or equal to 25 cm.

[0008] Step 2, the MCU obtains inertial element data and wheel speed sensor pulse data from the vehicle ESP system through the CAN bus, and obtains longitudinal acceleration and vehicle speed after processing the two signals;

[0009] Step 3, performing Kalman filtering on the longitudinal acceleration and vehicle speed obtained in step 2 to obtain the vehicle travel distance X, and constructing a data set of the vehicle travel distance X and time;

[0010] Step 4: After collecting m groups of data, the MCU constructs a function fitting of the vehicle reversing distance Y and the vehicle driving distance X, and updates the fitting parameters by recursive least squares method. When the reversing distance Y recorded in step 1 exceeds 25 cm, the two sets of data sets are reset, the fitting function parameters and covariance matrix are retained, and the above steps are repeated to update the parameters of the fitting function.

[0011] Step 5: When the distance measurement obtained by the ultrasonic sensor drops to 25 cm, the fitting function parameters are updated, and the vehicle travel distance X generated by the method in step 3 is substituted into the fitting function to obtain the compensated vehicle travel distance X. p Finally, the observed distance D = 25-X between the logistics vehicle and the unloading platform in the blind area of ​​the ultrasonic sensor is obtained. p .

[0012] Furthermore, in step 2, the output of the inertial element is subjected to limiting filtering to obtain the longitudinal acceleration a of the vehicle. o , the filter threshold is 1.4m / s 2 ; According to the calibration data of the wheel speed sensor, the pulse value is converted into the displacement distance of the vehicle during one pulse process, and the data is filtered by sliding average to calculate the vehicle speed v o .

[0013] Furthermore, in step 3, the Kalman filter state variables are displacement x, vehicle speed v, and longitudinal acceleration a, and the observed variable is vehicle speed v o , longitudinal acceleration a o , the observed noise size is based on the calibration value of the sensor.

[0014] Furthermore, in step 3, the Kalman filter formula used is as follows:

[0015] State quantity Observation quantity Z = [v o a o ] T ;

[0016] Calculate a priori estimates

[0017]

[0018] The state transfer matrix The control matrix B is 0;

[0019] Calculate the prior error covariance

[0020]

[0021] Where Q is the process noise covariance matrix of the model;

[0022] Calculate the Kalman gain K k :

[0023]

[0024] The observation matrix R is the observation noise, according to the calibration value of the sensor;

[0025] Update error covariance P k :

[0026]

[0027] Compute the posterior estimate

[0028]

[0029] The displacement x value in the posterior estimation matrix is ​​the vehicle's travel distance X.

[0030] Furthermore, in step 4, after collecting m=16 groups of reversing distance Y and vehicle driving distance X data, the fitting function parameters are updated once using the recursive least squares method, and the fitting function is a cubic function; after the recorded reversing distance Y exceeds 25 cm, the current Y and X data arrays are cleared, the reversing distance Y and driving distance X are reset to 0, and then data collection continues, and the fitting function parameters and covariance matrix are retained.

[0031] Furthermore, in step 4, the recursive least squares formula used is as follows:

[0032] When 16 sets of data are collected, the fitting curve parameters are solved once. The curve is a cubic function in the following form:

[0033] Y p =a0+a1*X+a2*X 2 +a3*X 3

[0034] Among them, a0, a1, a2, a3 are the fitting curve parameters, Y p As the fitting result, let Substitute the sample into the above formula and write it in matrix form:

[0035] Y p =Mθ

[0036] in m = 16;

[0037] Cost function:

[0038]

[0039] Taking partial derivatives of θ we get:

[0040]

[0041] After solving the equation, we get:

[0042] θ=(M T M) -1 M T Y

[0043] Thus, the initial fitting function parameters are obtained. When 16 sets of data are collected again, a recursive least squares operation is performed to update the fitting function parameters θ k :

[0044] θ k =θ k-1 +K k (Y k -M k θ k-1 )

[0045] in is the gain matrix, k is the number of iterations, R k To observe the noise, according to the calibration value of the sensor, and update the covariance matrix

[0046] P k =(IK k M k ) k-1 .

[0047] Furthermore, the system of the distance measurement method comprises an ultrasonic sensor, a wheel speed sensor, an inertial element and a microcontroller MCU, wherein the MCU obtains ultrasonic sensor data from a CAN bus, and obtains inertial element data and wheel speed sensor pulse data from a vehicle ESP system through the CAN bus; when the ultrasonic sensor is outside a short-range blind spot, the ultrasonic sensor detects the distance information between the vehicle and the loading and unloading platform; and when the ultrasonic sensor is within the blind spot, the processed distance observation value as described in step 5 is output.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention solves the problem of low accuracy of ranging in the blind area of ​​ultrasonic sensors and overcomes the problem of inaccurate positioning of inertial elements at low speeds. By using ultrasonic sensors to compensate the data of inertial elements and wheel speed sensors, it can be applied to distance observation in the blind area of ​​reversing radar. Through the fusion of multiple sensors, accurate ranging of 0 to 5 meters from obstacles is achieved, thereby solving the problem of poor application effect of reversing radar in scenarios such as logistics trucks docking at unloading platforms that require accurate distance in blind areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the distance measurement system of the present invention.

[0050] Figure 2 It is a flow chart of the distance measurement method of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with the accompanying drawings.

[0052] This embodiment proposes a multi-sensor fusion reversing radar blind spot ranging method, including the following specific steps:

[0053] 1) Taking the scenario of a logistics truck reversing and approaching a loading platform as an example, after the vehicle starts to reverse, the MCU obtains the ultrasonic radar ranging data through the CAN bus and constructs a data set of the vehicle's reversing distance Y and time based on the starting distance length;

[0054] 2) MCU obtains inertial element data and wheel speed sensor pulse data from the ESP system through the CAN bus, and obtains longitudinal acceleration and vehicle speed after processing the two signals;

[0055] 3) Performing Kalman filtering on the longitudinal acceleration and vehicle speed obtained in step 2) to obtain the vehicle travel distance X, and constructing a data set of the vehicle travel distance X and time;

[0056] 4) After collecting a certain amount of data, the MCU constructs a function fitting of the vehicle's reversing distance Y and the vehicle's driving distance X, and updates the fitting parameters through the recursive least squares method. When the reversing distance Y recorded in step 1) exceeds 25 cm, the two sets of data sets are reset, the fitting function parameters and covariance matrix are retained, and the above steps are repeated to update the fitting function parameters.

[0057] 5) When the distance measurement obtained by the ultrasonic sensor drops to 25 cm, the fitting function parameters are updated, and the vehicle travel distance X generated by the method in step 3) is substituted into the fitting function to obtain the compensated vehicle travel distance X p Finally, the distance observation value D = 25-X between the logistics vehicle and the unloading platform in the blind area of ​​the ultrasonic sensor is calculated. p .

[0058] In the above step 2), the output of the inertial element is filtered to obtain the longitudinal acceleration a of the vehicle. o The filtering threshold depends on the accuracy of the component; according to the calibration data of the wheel speed sensor, the pulse value is converted into the displacement distance of the vehicle during one pulse, and the data is filtered by sliding average to calculate the vehicle speed v o .

[0059] The Kalman filter formula used in step 3) above is as follows:

[0060] State quantity X = [xva] T , the observed quantity Z = [v o a o ] T . Compute the a priori estimate:

[0061]

[0062] in B is 0;

[0063] Compute the a priori error covariance:

[0064]

[0065] Where Q is the process noise covariance matrix of the model.

[0066] Calculate the Kalman gain:

[0067]

[0068] in R is the observation noise, which is based on the sensor's calibration value.

[0069] Update the error covariance:

[0070]

[0071] Posterior estimate:

[0072]

[0073] The vehicle travel distance X obtained from the inertial element and wheel speed sensor data is calculated.

[0074] The recursive least squares formula used in step 4) above is as follows:

[0075] When 16 sets of data are collected, the fitting parameters are calculated once.

[0076] Y p =a0+a1*X+a2*X 2 +a3*X 3

[0077] Make The above formula is written in matrix form as:

[0078] Y p =Mθ

[0079] in m=16.

[0080] Cost Function

[0081]

[0082] Taking partial derivatives of θ we get:

[0083]

[0084] After solving the equation, we get:

[0085] θ=(M T M) -1 M T Y

[0086] Thus, the initial fitting function parameters are obtained. When 16 sets of data are collected again, a recursive least squares operation is performed to update the fitting function parameters θ k :

[0087] θ k =θ k-1 +K k (Y k -M k θ k-1 )

[0088] in is the gain matrix, and updates the covariance matrix

[0089] P k =(IKk M k ) k-1

[0090] When the recorded reversing distance Y exceeds 25 cm, clear the current Y and X data arrays, reset the reversing distance Y and driving distance X to 0, and continue to collect data, retaining the previous fitting function parameters θ k and the covariance matrix P k , when 16 groups are collected, continue to use recursive least squares to update the fitting function parameters.

[0091] Furthermore, the system of the above-mentioned distance measurement method outputs the distance information between the vehicle and the loading and unloading platform detected by the ultrasonic sensor when it is outside the close-range blind zone of the ultrasonic sensor; and outputs the processed distance observation value described in step 5) when it is within the blind zone.

[0092] Through the above-mentioned design, the ranging data outside the blind spot of the ultrasonic sensor is used to generate a fitting function for the displacement data obtained by the inertial element and the wheel speed sensor, and its accuracy is compensated so that it can accurately position at low speeds. When outside the blind spot, the system uses an ultrasonic sensor for ranging, and when inside the blind spot, it uses a multi-sensor fusion blind spot ranging method for distance observation. Therefore, the present invention can maintain the ranging accuracy in the blind spot of the ultrasonic sensor, and achieve accurate ranging of 0 to 5 meters from obstacles, solving the problem of poor application effect of reversing radar in scenarios such as logistics trucks docking at unloading platforms that require accurate distances within blind spots.

[0093] In addition to the above embodiments, the present invention may also have other implementation modes. Any technical solution formed by equivalent replacement or equivalent transformation falls within the protection scope required by the present invention.

Claims

1. A multi-sensor fusion reversing radar blind spot ranging method, characterized by: The specific steps include: Step 1: When the vehicle starts to reverse, the vehicle's MCU obtains ultrasonic radar ranging data through the CAN bus, and constructs a data set of the vehicle's reverse distance Y and time according to the starting distance length. The recorded Y value is less than or equal to 25 cm. Step 2: The MCU obtains the inertial element data and wheel speed sensor pulse data from the vehicle ESP system through the CAN bus, and obtains the longitudinal acceleration and vehicle speed after processing the two signals; the output of the inertial element is limited and filtered to obtain the longitudinal acceleration a of the vehicle. o , the filter threshold is 1.4m / s 2 ; According to the calibration data of the wheel speed sensor, the pulse value is converted into the displacement distance of the vehicle during one pulse process, and the data is filtered by sliding average to calculate the vehicle speed v o ; Step 3: Perform Kalman filtering on the longitudinal acceleration and vehicle speed obtained in step 2 to obtain the vehicle travel distance X, and construct a data set of vehicle travel distance X and time; the Kalman filter state quantities are displacement x, vehicle speed v, and longitudinal acceleration a, and the observed quantity is vehicle speed v o , longitudinal acceleration a o , the observed noise size is based on the calibration value of the sensor; the Kalman filter formula used is as follows: State quantity Observation quantity Z = [v o a o ] T ; Calculate a priori estimates The state transfer matrix The control matrix B is 0; Calculate the prior error covariance Where Q is the process noise covariance matrix of the model; Calculate the Kalman gain K k : The observation matrix R is the observation noise, according to the calibration value of the sensor; Update error covariance P k : Compute the posterior estimate The displacement x value in the posterior estimation matrix is ​​the vehicle travel distance X; Step 4, after collecting m groups of data, the MCU constructs a function fitting of the vehicle reversing distance Y and the vehicle driving distance X, and updates the fitting parameters by recursive least squares. When the reversing distance Y recorded in step 1 exceeds 25 cm, reset the two sets of data sets, retain the fitting function parameters and the covariance matrix, and repeat the above steps to update the parameters of the fitting function; after collecting m=16 groups of reversing distance Y and vehicle driving distance X data, use the recursive least squares method to update the fitting function parameters once, and the fitting function is a cubic function; after the recorded reversing distance Y exceeds 25 cm, clear the current Y and X data arrays, reset the reversing distance Y and driving distance X to 0, and continue to collect data, retaining the fitting function parameters and the covariance matrix; the recursive least squares formula used is as follows: When 16 sets of data are collected, the fitting curve parameters are solved once. The curve is a cubic function in the following form: Y p =a0+a1*X+a2*X 2 +a3*X 3 Among them, a0, a1, a2, a3 are the fitting curve parameters, Y p As the fitting result, let Substitute the sample into the above formula and write it in matrix form: AND p =Mθ in m = 16; Cost function: Taking partial derivatives of θ we get: After solving the equation, we get: θ=(M T M) -1 M T Y Thus, the initial fitting function parameters are obtained. When 16 sets of data are collected again, a recursive least squares operation is performed to update the fitting function parameters θ k : i k =θ k-1 +K k (Y k -M k i k-1 ) in is the gain matrix, k is the number of iterations, R k To observe the noise, according to the calibration value of the sensor, and update the covariance matrix P k =(I-K k M k )P k-1 ; Step 5: When the distance measurement obtained by the ultrasonic sensor drops to 25 cm, the fitting function parameters are updated, and the vehicle travel distance X generated by the method in step 3 is substituted into the fitting function to obtain the compensated vehicle travel distance X. p Finally, the observed distance D = 25-X between the logistics vehicle and the unloading platform in the blind area of ​​the ultrasonic sensor is obtained. p .

2. The method for measuring blind spot of reversing radar by multi-sensor fusion according to claim 1 is characterized in that: The system of the distance measurement method is composed of an ultrasonic sensor, a wheel speed sensor, an inertial element and a microcontroller MCU. The MCU obtains ultrasonic sensor data from a CAN bus, and obtains inertial element data and wheel speed sensor pulse data from a vehicle ESP system through the CAN bus; when the ultrasonic sensor is outside a short-range blind spot, the distance information between the vehicle and the loading and unloading platform detected by the ultrasonic sensor is output; when the ultrasonic sensor is within the blind spot, the processed distance observation value described in step 5 is output.