Automobile LiDAR device cleaning system and method

Abstract

The invention discloses an automobile LiDAR device cleaning system which is characterized in that a LiDAR assembly is provided with a mirror surface used for detecting a window, an integrated nozzle is fixed above the LiDAR assembly, the integrated nozzle sprays a cleaning medium to the window mirror surface, the integrated nozzle is connected with a pump through a pipeline, the pipeline is provided with a heating module, and the heating module is connected with the pump. The LiDAR assembly outputs a signal to the ADAS domain controller through the CAN, the ADAS domain controller performs information interaction with the cleaning system controller through the CAN, the cleaning system controller is connected with the pump and the heating module and outputs a driving signal to the pump and the heating module, and the cleaning system controller is connected with an environment temperature sensor. The cleaning system is high in integration degree, all the control units are integrated in the cleaning system controller and connected with other sensors, actuators and ECUs through external interfaces and wire harnesses, and the overall complexity of the system is low; the whole cleaning process is automatically completed by the system, manual operation of a driver is not needed, the driver take-over frequency required by the system in the auxiliary driving starting process is greatly reduced, and the driving experience is improved.

Description

Technical Field

[0001] This invention relates to the field of automotive autonomous driving sensor cleaning technology. Background Technology

[0002] Most mainstream ADAS solutions currently employ multi-sensor fusion, relying on specific algorithms to process data collected by cameras, LiDAR, millimeter-wave radar, and infrared spectral sensors to achieve advanced autonomous driving. LiDAR, as a core sensor for L2+ to L4 level autonomous driving, is used for environmental perception, obstacle detection, and high-precision positioning. It generates 3D point cloud data by emitting laser beams, enabling high-precision distance, angle, and velocity measurements. Combined with deep learning algorithms, it is used for target recognition, dynamic tracking, and scene understanding. Rain, snow, dust, and other contaminants cover the LiDAR window mirror, causing the LiDAR point cloud to be lost or expanded, significantly reducing detection accuracy. L3+ autonomous driving requires sensor availability greater than 99.9%, necessitating active cleaning to ensure system redundancy. The LiDAR cleaning system can spray water or blow air to clean the LiDAR in real time based on ADAS feedback signals, ensuring the cleanliness of its window mirror.

[0003] The existing technical solutions for cleaning systems include the following three approaches:

[0004] 1. Timed Triggered Cleaning: The cleaning system starts working immediately after the vehicle is started, without needing to interact with the ADAS domain controller. It cleans the entire vehicle's LiDAR every certain time interval (3~5 minutes).

[0005] 2. Manual cleaning: While the vehicle is in motion, the instrument panel receives the CAN signal from the ADAS domain controller and displays the information that the LiDAR needs to be cleaned on the instrument panel. The driver can manually press the switch at the rear of the vehicle to trigger the cleaning (similar to the operation of the windshield wiper switch).

[0006] 3. Real-time control of cleaning: After the vehicle is powered on, the cleaning system controller receives the CAN signal from the ADAS domain controller. Through the internal timer of the controller MCU, within a certain time range (5~10min according to ADAS command messages), the nozzles are controlled to spray water or blow air onto the LiDAR at regular intervals (10~20s).

[0007] A common cleaning device structure is exemplified by the patent publication document with authorization announcement number CN216424350U, authorization announcement date of 2022-05-03, entitled "A Hidden LiDAR Cleaning Device and Automobile". The disclosed hidden LiDAR cleaning device includes a telescopic liquid guiding mechanism installed above the LiDAR housing. A decorative cover is provided at the end of the telescopic end of the telescopic liquid guiding mechanism. At least one nozzle is provided at the telescopic end of the telescopic liquid guiding mechanism near the decorative cover. The nozzle sprays water downward and is inclined away from the decorative cover. It also includes a liquid guiding hose, one end of which is connected to the telescopic liquid guiding mechanism and the other end is connected to a pump.

[0008] Similar existing systems have the following problems:

[0009] 1. Timed trigger cleaning: This is excessively redundant. The system continues to operate even when cleaning is not required, which reduces the system's lifespan and requires frequent addition of cleaning fluid.

[0010] 2. Manual cleaning: It can easily distract the driver and is not suitable for Level 4 and above autonomous vehicles;

[0011] 3. Real-time control cleaning: It is impossible to clean according to the specific contamination level of the LiDAR, and it is impossible to control the flow rate of water in a single cycle, which can easily lead to over-cleaning and requires frequent addition of cleaning fluid. Summary of the Invention

[0012] The technical problem to be solved by the present invention is to realize a PWM-based automotive LiDAR cleaning control system and method, which can determine the degree of LiDAR impact in real time, perform efficient cleaning while ensuring driving safety and system redundancy, and also has the advantage of saving energy.

[0013] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a cleaning system for an automotive LiDAR device, wherein the LiDAR assembly is provided with a detection window mirror, an integrated nozzle is fixed above the LiDAR assembly, the integrated nozzle sprays a cleaning medium onto the window mirror, the integrated nozzle is connected to a pump through a pipeline, a heating module is provided on the pipeline, the LiDAR assembly outputs a signal to an ADAS domain controller via CAN, the ADAS domain controller interacts with a cleaning system controller via CAN, the cleaning system controller is connected to and outputs drive signals to the pump and the heating module, and an ambient temperature sensor is connected to the cleaning system controller.

[0014] The integrated nozzle includes two strip-shaped air nozzles and a liquid nozzle. Both the air nozzles and the liquid nozzles spray towards the upper edge of the entire window mirror. The air nozzles are connected to an air pump through a gas pipeline, and the liquid nozzles are connected to a water pump through a liquid pipeline and a solenoid valve. The water pump is connected to a washing jug through a pipeline. The cleaning system controller connects to and controls the water pump, the air pump, and the solenoid valve. The heating module is a PTC heating module.

[0015] The upper air path nozzle and the lower liquid path nozzle are two parallel cavities. The integrated nozzle is fixed to the upper surface of the LiDAR assembly frame by a snap-fit ​​structure. The cleaning system controller includes an MCU, a PWM square wave speed control module, a MOS drive circuit, a CAN communication circuit, an OTA module, an HSD diagnostic module, and an external interface.

[0016] The window mirror is a sloping surface that extends forward from the bottom. The spray direction of the integrated nozzle is towards the upper edge of the window mirror. The liquid or gas sprayed by the integrated nozzle is sprayed from the upper edge of the window mirror to the lower edge of the window mirror.

[0017] A cleaning method based on the aforementioned automotive LiDAR device cleaning system:

[0018] Step 1: Acquire LiDAR point cloud data and determine if there are any missing points in the point cloud. If so, proceed to the next step.

[0019] Step 2: Determine whether the lost point cloud is in a critical area. If so, classify it as the most severe level; otherwise, classify it as a severe level.

[0020] Step 3: Output the corresponding duty cycle to the output module according to the severity level;

[0021] Step 4: The output module determines the operating conditions of the air pump and water pump based on the duty cycle signal.

[0022] Step 5: Determine if there are any missing points in the point cloud. If so, return to step 1; otherwise, the cleaning process ends.

[0023] In step 1, the cleaning system controller, based on the point cloud signal acquired by LiDAR and forwarded by the ADAS domain controller, substitutes the probability density function of the Gaussian mixture model of the three-dimensional point cloud coordinates:

[0024]

[0025] In the formula, The number of Gaussian distributions in the mixture model; It is the first The weights of a Gaussian distribution represent the proportion of that distribution, and satisfy the following conditions: ; It is the first The mean vector of a Gaussian distribution represents the location of the center of the distribution; It is the first A Gaussian-distributed covariance matrix is ​​used to describe the spatial distribution of the point cloud. Based on the interval of the P-value, the severity level of point cloud loss is obtained, and the duty cycle of the PWM signal is calculated.

[0026] Step 1, determining whether the point cloud is missing, includes the following steps:

[0027] a) Obtain the first in three-dimensional space Gaussian probability density function , represented as

[0028]

[0029] in For three-dimensional point cloud data dimensions, It is a cosine matrix The determinant;

[0030] b) Model parameters The probability of point cloud generation is estimated using the Expectation-Maximization (EM) algorithm.

[0031]

[0032] c) Combining the point cloud loss severity classification strategy in Table 1, the following will be obtained: Low-probability points are marked, with key regions marked separately.

[0033] The severity levels are divided into five levels: S1, S2, S3, S4, and S5, with the severity levels increasing sequentially.

[0034] S1 represents a slight point cloud loss level, indicating that a single lidar sensor has a point cloud loss of no more than 1% in the non-overlapping area of ​​the field of view (FOV).

[0035] S2 is the general point cloud loss level, indicating that a single lidar sensor has a point cloud loss of 1% to 5% in the non-overlapping area of ​​the field of view (FOV).

[0036] S3 represents a relatively severe point cloud loss level, indicating that a single lidar sensor experiences a 5% to 10% point cloud loss in the non-overlapping area of ​​the field of view (FOV).

[0037] S4 is the severe point cloud loss level, indicating that a single lidar has a point cloud loss of 10% to 20% in the non-overlapping area of ​​the field of view (FOV), or a single lidar has a point cloud loss of more than 20% in the overlapping area of ​​the FOV.

[0038] S5 is the point cloud complete loss level, which means that a single LiDAR has more than 20% point cloud loss in the non-overlapping FOV area, or more than 1% point cloud loss in the critical area. The critical area refers to the area where the ADAS algorithm needs to focus on collecting point cloud data.

[0039] PWM frequency 16kHz, resolution 8bit;

[0040] Define the frequency signal range (1%, 20%) of the duty cycle of the PWM square wave speed control module in state S1.

[0041] Define the frequency signal range (20%, 40%) of the duty cycle of the S2 state PWM square wave speed control module.

[0042] Define the frequency signal range (40%, 60%) of the duty cycle of the S3 state PWM square wave speed control module.

[0043] Define the frequency signal range (60%, 80%) of the duty cycle of the S4 state PWM square wave speed control module.

[0044] Define the frequency signal range (80%, 99%) of the duty cycle of the S5 state PWM square wave speed control module.

[0045] 10. The method for cleaning an automotive LiDAR device according to claim 9, characterized in that: in step 4, during the operation:

[0046] a) Receive frequency signal range of duty cycle (1%, 20%), MOS drive air pump runs at 60% power for 10s;

[0047] b) Receive frequency signal range of duty cycle (20%, 40%), MOS drive air pump runs at 100% power for 10s;

[0048] c) Receive frequency signal range of duty cycle (40%, 60%), MOS drive solenoid valve to open at 60% opening degree for 5s, air pump to run at 100% power for 10s;

[0049] d) Receive frequency signal range of duty cycle (60%, 80%), MOS drives solenoid valve to open at 80% opening degree for 5s, air pump runs at 100% power for 10s;

[0050] e) Receive frequency signal range of duty cycle (80%, 100%), MOS drive solenoid valve to open at 100% opening degree for 10s, air pump to run at 100% power for 15s;

[0051] When cleaning begins, if the ambient temperature is detected to be lower than the set temperature threshold, the PTC heating module is controlled to work. The temperature threshold is 1℃-6℃.

[0052] The cleaning system of this invention has a high degree of integration. All control units are integrated into the cleaning system controller and connected to other sensors, actuators, and ECUs through external interfaces and wiring harnesses, resulting in low overall system complexity. The entire cleaning process is completed autonomously by the system without the need for manual operation by the driver, which greatly reduces the number of times the system requires the driver to take over during the activation of assisted driving and improves the driving experience. Attached Figure Description

[0053] The following is a brief explanation of the content and markings in each of the accompanying drawings in this specification:

[0054] Figure 1 This is a schematic diagram of the cleaning system.

[0055] Figure 2 This is a schematic diagram of the cleaning system structure;

[0056] Figure 3 This is a cross-sectional view of the schematic diagram of the cleaning system.

[0057] Figure 4 Here is a flowchart of the cleaning method cycle;

[0058] Figure 5 Functional block diagram of the cleaning system controller;

[0059] The markings in the above figures are: 1. LiDAR assembly; 2. Window mirror; 3. Integrated nozzle; 4. Liquid path nozzle; 5. Gas path nozzle. Detailed Implementation

[0060] The following description, with reference to the accompanying drawings, details the specific implementation of the present invention, including the shape and structure of each component, the relative positions and connections between the parts, the function and working principle of each part, the manufacturing process, and the operation and use methods, to help those skilled in the art to have a more complete, accurate, and in-depth understanding of the inventive concept and technical solution of the present invention.

[0061] like Figure 1-3 As shown, the automotive LiDAR device cleaning system includes a cleaning system controller, an air pump, a water pump, a solenoid valve, an ambient temperature sensor, a PTC heating module, a washing pot, air lines, liquid lines, an air nozzle 5, and a liquid nozzle 4.

[0062] An air pump is used to generate high-pressure airflow.

[0063] A water pump is used to generate a high-pressure water flow at a specific flow rate.

[0064] The solenoid valve is connected to the washing kettle and water pump through a liquid pipeline. By adjusting the opening of the solenoid valve, the flow rate of high-pressure water can be precisely controlled.

[0065] An ambient temperature sensor communicates with the cleaning system controller via LIN to collect ambient temperature data. When the ambient temperature is below a set threshold, the PTC heating module is activated to heat the airflow and water flow.

[0066] The PTC heating module is used to heat high-pressure airflow and water flow to meet cleaning needs in specific environments such as those covered by frost and snow.

[0067] Washing jug, used for holding and adding washing liquid;

[0068] The cleaning system controller communicates with the ADAS domain controller and other ECUs via a common CAN or CANFD.

[0069] The cleaning system controller communicates with the air pump, water pump, solenoid valve, and PTC heating module via hardwired communication to send drive signals;

[0070] Gas and liquid pipelines are used to deliver high-pressure gas and high-pressure water to gas nozzle 5 and liquid nozzle 4, respectively.

[0071] The gas nozzle 5 and the liquid nozzle 4 are integrated into a single nozzle 3, which is mounted on the lidar, and its structure is as follows: Figure 2 As shown.

[0072] The LiDAR assembly 1 is typically mounted on the roof of the vehicle. A detection window mirror 2 is located at the front of the LiDAR assembly 1. The window mirror 2 is a sloping surface extending forward from its bottom. To clean the window mirror 2, an integrated nozzle 3 is fixed above the LiDAR assembly 1. The integrated nozzle 3 sprays a cleaning medium onto the window mirror 2. Preferably, the integrated nozzle 3 can simultaneously spray cleaning agent and air—one for water washing and the other for dust removal and drying—eliminating the need for windshield wipers and ensuring reliable and efficient cleaning. The integrated nozzle 3 contains two parallel cavities: an upper air nozzle 5 and a lower liquid nozzle 4. The nozzle tails are connected to the air and liquid pipes respectively. Both the air nozzle 5 and liquid nozzle 4 have strip-shaped spray nozzles, spraying towards the upper edge of the entire window mirror 2, thus forming an oblique jet stream for cleaning. The air nozzle 5 is connected to an air pump via a gas pipe, and the liquid nozzle 4 is connected to a water pump via a liquid pipe and a solenoid valve. The water pump is connected to a washing jug via a pipe. Existing connection, fixing, and sealing mechanisms can be used. The integrated nozzle 3 is fixed to the upper surface of the LiDAR frame via a snap-fit ​​structure; existing connection and fixing mechanisms can also be used.

[0073] The cleaning system does not include Figure 1The system includes an ADAS domain controller and a LiDAR, but the cleaning system requires the ADAS domain controller to send point cloud data collected by the LiDAR for operation. The cleaning system controller, air pump, water pump, solenoid valve, ambient temperature sensor, and PTC heating module are all powered by the vehicle's 12V low-voltage power supply. The PTC heating module can be fixed to the pipeline to heat the air nozzle 5 and the liquid nozzle 4. For example, it can be placed between the two cavities of the air nozzle 5 and the liquid nozzle 4, allowing a single heating device to simultaneously heat both the air and water circuits.

[0074] The ADAS system includes an ADAS domain controller, a LiDAR, and other sensors. The LiDAR, as the object being cleaned, communicates with the ADAS domain controller via a proprietary CAN bus to collect point cloud data and send it to the ADAS domain controller. The ADAS domain controller processes the point cloud data collected by the LiDAR and feeds back the point cloud loss signal to the cleaning system controller.

[0075] The cleaning system controller is the core control module of the cleaning system. It includes a control chip MCU, a PWM square wave speed regulation module, a MOS drive circuit, a CAN communication circuit, an OTA module, an HSD diagnostic module, and an external interface. The cleaning system controller is connected to other modules through the external interface. The cleaning system controller receives signals from the ambient temperature sensor inside the system to control the PTC heating module, and receives signals from the ADAS domain controller outside the system to control the air pump, water pump, and solenoid valve. It also receives corresponding feedback signals.

[0076] The PWM square wave speed control module adjusts the duty cycle of the PWM signal according to the MCU output signal and outputs the PWM signal to the MOS drive circuit; the MOS drive circuit outputs corresponding voltages to the air pump, water pump, and solenoid valve according to the PWM signal, driving the air pump, water pump, and solenoid valve to work; the CAN communication circuit is used for communication with other ECUs; the OTA module is used for remote upgrades via the vehicle T-BOX; the HSD diagnostic module is used for fault diagnosis of open circuit or short circuit in the system high-side output and sends fault codes.

[0077] The cleaning system controller controls the air pump, water pump, solenoid valve, and PTC heating module. It communicates with the ADAS domain controller, receives point cloud loss signals from the ADAS domain controller, processes the point cloud loss information, calculates the point cloud loss probability density distribution, determines the severity level, and controls the air pump, water pump, solenoid valve, and PTC heating module by outputting corresponding PWM signals according to different severity levels. The cleaning system controller also communicates with other ECUs in the vehicle. The control board is powered by a 12V power supply and includes a control chip (MCU), PWM, square wave speed control module, MOS drive circuit, CAN communication circuit, OTA module, and HSD diagnostic module.

[0078] In practice, the control chip MCU, based on the point cloud signal acquired by LiDAR and forwarded by the ADAS domain controller, substitutes the probability density function of the Gaussian mixture model of the three-dimensional point cloud coordinates.

[0079]

[0080] In the formula, The number of Gaussian distributions in the mixture model; It is the first The weights of a Gaussian distribution represent the proportion of that distribution, and satisfy the following conditions: ; It is the first The mean vector of a Gaussian distribution represents the location of the center of the distribution; It is the first A Gaussian-distributed covariance matrix is ​​used to describe the spatial distribution of the point cloud. Based on the interval of the P-value, the severity level of point cloud loss is obtained, and the duty cycle of the PWM signal is calculated.

[0081] The severity level assessment strategy is shown in Table 1. The severity levels are divided into five levels: S1, S2, S3, S4, and S5, with the severity level increasing sequentially.

[0082] S1 represents a slight point cloud loss level, indicating that a single LiDAR sensor has a point cloud loss of no more than 1% in the non-overlapping area of ​​the FOV (Field of View). This has a minimal impact on system performance, and the system can completely compensate for the missing information by relying on other sensors.

[0083] S2 represents the general point cloud loss level, indicating that a single LiDAR has a 1% to 5% point cloud loss in the non-overlapping area of ​​the FOV (Field of View). In static or known environments, the system can rely on historical data or map information to completely compensate for the missing information.

[0084] S3 indicates a severe point cloud loss level, meaning that a single lidar sensor has lost 5% to 10% of the point cloud in the non-overlapping area of ​​the field of view (FOV). Regardless of whether the system is in a static or dynamic environment, it can only rely on other methods to make up for some of the missing information.

[0085] S4 indicates a severe point cloud loss level, meaning that a single LiDAR has 10% to 20% point cloud loss in the non-overlapping area of ​​the FOV (Field of View), or more than 20% point cloud loss in the overlapping area of ​​the FOV (Field of View). The missing information cannot be compensated for, which seriously affects the accuracy of the system's detection. However, it will not cause the autonomous driving to completely fail, and it can be restored after a short period of takeover by the driver.

[0086] S5 represents the complete loss of point cloud data, indicating that a single LiDAR has more than 20% of the point cloud data lost in the non-overlapping area of ​​the FOV (Field of View), or more than 1% of the point cloud data lost in the critical area. The critical area refers to the area where ADAS algorithms need to collect point cloud data. Point cloud loss will directly affect the implementation of ADAS algorithms and cause autonomous driving to fail.

[0087] If the severity level increases or decreases during the cleaning process, the next cleaning cycle will be performed according to the changed severity level, and the duty cycle of the PWM square wave speed control module will be adjusted.

[0088] Based on the probability density function of the Gaussian mixture model of the 3D point cloud coordinates, the number of detected point clouds is used to calculate the point cloud loss situation, and the value of P is substituted into the severity level range.

[0089] Table 1. Point Cloud Loss Severity Classification Strategy:

[0090]

[0091] The specific steps are as follows:

[0092] 1. The ADAS domain controller acquires LiDAR point cloud data and determines the point cloud loss situation;

[0093] 2. Based on the point cloud loss situation, the cleaning system controller determines the severity level (S1-S5).

[0094] 3. The MCU has a built-in time signal timer for timing synchronization and a PWM square wave speed control module that outputs the corresponding duty cycle according to different severity levels;

[0095] 4. Output module duty cycle identification, adjusting air pump motor speed and water pump solenoid valve opening;

[0096] 5. The liquid nozzle 4 and the gas nozzle 5 spray and gas wash the LiDAR window mirror 2 in sequence;

[0097] 6. The cleaning system controller receives the point cloud status feedback signal from the ADAS domain controller. If not, the severity level needs to be reassessed and the next cleaning cycle needs to be performed.

[0098] 7. Cleaning complete.

[0099] The following is combined Figure 5 The functional block diagram of the cleaning system controller according to an embodiment of the present invention illustrates the functions and control strategies of the cleaning system:

[0100] 1. Power-on wake-up: The Cleaning System Control Module (CSCM) adopts distributed (AutoSar) network management. A single node in the network sends an NM frame to trigger wake-up. After wake-up, the Cleaning System Control Module sends a request for the current ambient temperature information via LIN. The ambient temperature sensor continuously feeds back temperature signals.

[0101] 2. With the power on (ON), the cleaning system controller receives point cloud loss information from the ADAS domain controller and determines the severity level of point cloud loss based on the point cloud loss probability density calculation method described above:

[0102] a) Obtain the first in three-dimensional space Gaussian probability density function , represented as

[0103]

[0104] in For three-dimensional point cloud data dimensions, It is a cosine matrix The determinant;

[0105] b) Model parameters The probability of point cloud generation is estimated using the Expectation-Maximization (EM) algorithm.

[0106]

[0107] c) Combining the point cloud loss severity classification strategy in Table 1, the following will be obtained: Low-probability points are marked, with key regions marked separately;

[0108] 3. Define the duty cycle signal range for the point cloud loss severity level feedback, and set the PWM frequency to 16kHz and the resolution to 8bit;

[0109] Define the frequency signal range (1%, 20%) of the duty cycle of the PWM square wave speed control module in state S1.

[0110] Define the frequency signal range (20%, 40%) of the duty cycle of the S2 state PWM square wave speed control module.

[0111] Define the frequency signal range (40%, 60%) of the duty cycle of the S3 state PWM square wave speed control module.

[0112] Define the frequency signal range (60%, 80%) of the duty cycle of the S4 state PWM square wave speed control module.

[0113] Define the frequency signal range (80%, 99%) of the duty cycle of the S5 state PWM square wave speed control module.

[0114] Table 2 Cleaning strategies for different ambient temperatures and severity levels:

[0115]

[0116] 4. The MCU collects the current ambient temperature information and combines it with the cleaning strategies for different ambient temperatures and severity levels in Table 2. If the detected temperature is <4℃, the MCU outputs a high-side signal to the PTC heating (relay) module. The heating temperature range is 50℃~80℃.

[0117] 5. The MCU determines the severity level of point cloud loss as described above. If it is level S3 to S5, it outputs a high-side signal to the water pump (relay) module to drive the water pump to work.

[0118] 6. The MOS drive circuit, based on the frequency signal range with different duty cycles and the cleaning strategies for different ambient temperatures and severity levels as shown in Table 2, performs the following actions:

[0119] a) Receive frequency signal range of duty cycle (1%, 20%), MOS drive air pump runs at 60% power for 10s;

[0120] b) Receive frequency signal range of duty cycle (20%, 40%), MOS drive air pump runs at 100% power for 10s;

[0121] c) Receive frequency signal range of duty cycle (40%, 60%), MOS drive solenoid valve to open at 60% opening degree for 5s, air pump to run at 100% power for 10s;

[0122] d) Receive frequency signal range of duty cycle (60%, 80%), MOS drives solenoid valve to open at 80% opening degree for 5s, air pump runs at 100% power for 10s;

[0123] e) Receive frequency signal range of duty cycle (80%, 100%), MOS drive solenoid valve to open at 100% opening degree for 10s, air pump to run at 100% power for 15s;

[0124] 7. The cleaning system controller feeds back a single cleaning cycle signal to the ADAS domain controller. This signal is sent periodically, waiting for the ADAS domain controller to request whether to enter the next cleaning cycle.

[0125] The above content discloses the technical solution of the present invention, which has the following advantages over the prior art:

[0126] 1. The present invention has a high degree of redundancy in the graded strategy for the loss of LiDAR-collected point clouds, a fast cleaning response speed, and strong protection for the driving safety of vehicles with assisted driving enabled.

[0127] 2. In the cleaning system of the present invention, a graded mechanism is adopted for the degree of contamination of LiDAR, and special cleaning strategies are adopted for different working conditions to ensure cleaning effect while saving resources and reducing power consumption and cleaning fluid consumption.

[0128] 3. The cleaning system of the present invention has a high degree of integration. All control units are integrated into the cleaning system controller and are connected to other sensors, actuators and ECUs through external interfaces and wiring harnesses, resulting in low overall system complexity.

[0129] 4. The entire cleaning process of this invention is completed autonomously by the system without the need for manual operation by the driver, which greatly reduces the number of times the system requires the driver to take over during the assisted driving process and improves the driving experience.

[0130] 5. In the system of the present invention, a PTC heating module is used to heat the airflow and liquid flow in low-temperature environments, which effectively preheats the integrated nozzle 3 to ensure its normal operation, while improving the cleaning efficiency under frost and snow cover conditions.

[0131] The above control strategy is merely one embodiment of the cleaning strategy of this invention. For those skilled in the art, the point cloud loss probability density calculation method, cleaning power output mode, PWM frequency, resolution, and the frequency signal range of the PWM square wave duty cycle can all be changed and varied according to actual conditions. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the control strategy of this invention should be included within the protection scope of this invention.

Claims

1. A cleaning system for an automotive LiDAR device, characterized in that: The LiDAR assembly is equipped with a detection window mirror. An integrated nozzle is fixed above the LiDAR assembly. The integrated nozzle sprays cleaning medium onto the window mirror. The integrated nozzle is connected to a pump through a pipeline. A heating module is installed on the pipeline. The LiDAR assembly outputs signals to the ADAS domain controller via CAN. The ADAS domain controller interacts with the cleaning system controller via CAN. The cleaning system controller is connected to and outputs drive signals to the pump and the heating module. The cleaning system controller is also connected to an ambient temperature sensor.

2. The automotive LiDAR device cleaning system according to claim 1, characterized in that: The integrated nozzle includes two strip-shaped air nozzles and a liquid nozzle. Both the air nozzles and the liquid nozzles spray towards the upper edge of the entire window mirror. The air nozzles are connected to an air pump through a gas pipeline, and the liquid nozzles are connected to a water pump through a liquid pipeline and a solenoid valve. The water pump is connected to a washing jug through a pipeline. The cleaning system controller connects to and controls the water pump, the air pump, and the solenoid valve. The heating module is a PTC heating module.

3. The automotive LiDAR device cleaning system according to claim 2, characterized in that: The upper air path nozzle and the lower liquid path nozzle are two parallel cavities. The integrated nozzle is fixed to the upper surface of the LiDAR assembly frame by a snap-fit ​​structure. The cleaning system controller includes an MCU, a PWM square wave speed control module, a MOS drive circuit, a CAN communication circuit, an OTA module, an HSD diagnostic module, and an external interface.

4. The automotive LiDAR device cleaning system according to claim 3, characterized in that: The window mirror is a sloping surface that extends forward from the bottom. The spray direction of the integrated nozzle is towards the upper edge of the window mirror. The liquid or gas sprayed by the integrated nozzle is sprayed from the upper edge of the window mirror to the lower edge of the window mirror.

5. A cleaning method based on the automotive LiDAR device cleaning system according to any one of claims 1-4, characterized in that: Step 1: Acquire LiDAR point cloud data and determine if there are any missing points in the point cloud. If so, proceed to the next step. Step 2: Determine whether the lost point cloud is in a critical area. If so, classify it as the most severe level; otherwise, classify it as a severe level. Step 3: Output the corresponding duty cycle to the output module according to the severity level; Step 4: The output module determines the operating conditions of the air pump and water pump based on the duty cycle signal. Step 5: Determine if there are any missing points in the point cloud. If so, return to step 1; otherwise, the cleaning process ends.

6. The method for cleaning an automotive LiDAR device according to claim 5, characterized in that: In step 1, the cleaning system controller, based on the point cloud signal acquired by LiDAR and forwarded by the ADAS domain controller, substitutes the probability density function of the Gaussian mixture model of the three-dimensional point cloud coordinates: In the formula, The number of Gaussian distributions in the mixture model; It is the first The weights of a Gaussian distribution represent the proportion of that distribution, and satisfy the following conditions: ; It is the first The mean vector of a Gaussian distribution represents the location of the center of the distribution; It is the first A Gaussian-distributed covariance matrix is ​​used to describe the spatial distribution of the point cloud. Based on the range of P value, the severity level of point cloud loss is obtained, and the duty cycle of the PWM signal is calculated.

7. The method for cleaning an automotive LiDAR device according to claim 6, characterized in that: Step 1, determining whether the point cloud is missing, includes the following steps: a) Obtain the first in three-dimensional space Gaussian probability density function , represented as in For three-dimensional point cloud data dimensions, It is a cosine matrix The determinant; b) Model parameters The probability of point cloud generation is estimated using the Expectation-Maximization (EM) algorithm. c) Combining the point cloud loss severity classification strategy, the following will be obtained Low-probability points are marked, with key regions marked separately.

8. The method for cleaning an automotive LiDAR device according to claim 5, 6, or 7, characterized in that: The severity levels are divided into five levels: S1, S2, S3, S4, and S5, with the severity levels increasing sequentially. S1 represents a slight point cloud loss level, indicating that a single lidar sensor has a point cloud loss of no more than 1% in the non-overlapping area of ​​the field of view (FOV). S2 is the general point cloud loss level, indicating that a single lidar sensor has a point cloud loss of 1% to 5% in the non-overlapping area of ​​the field of view (FOV). S3 represents a relatively severe point cloud loss level, indicating that a single lidar sensor experiences a 5% to 10% point cloud loss in the non-overlapping area of ​​the field of view (FOV). S4 is the severe point cloud loss level, indicating that a single lidar has a point cloud loss of 10% to 20% in the non-overlapping area of ​​the field of view (FOV), or a single lidar has a point cloud loss of more than 20% in the overlapping area of ​​the FOV. S5 is the point cloud complete loss level, which means that a single LiDAR has more than 20% point cloud loss in the non-overlapping FOV area, or more than 1% point cloud loss in the critical area. The critical area refers to the area where the ADAS algorithm needs to focus on collecting point cloud data.

9. The method for cleaning an automotive LiDAR device according to claim 8, characterized in that: PWM frequency 16kHz, resolution 8bit; Define the frequency signal range (1%, 20%) of the duty cycle of the PWM square wave speed control module in state S1. Define the frequency signal range (20%, 40%) of the duty cycle of the S2 state PWM square wave speed control module. Define the frequency signal range (40%, 60%) of the duty cycle of the S3 state PWM square wave speed control module. Define the frequency signal range (60%, 80%) of the duty cycle of the S4 state PWM square wave speed control module. Define the frequency signal range (80%, 99%) of the duty cycle of the S5 state PWM square wave speed control module.

10. The method for cleaning an automotive LiDAR device according to claim 9, characterized in that: In step 4, when performing situational work: a) Receive frequency signal range of duty cycle (1%, 20%), MOS drive air pump runs at 60% power for 10s; b) Receive frequency signal range of duty cycle (20%, 40%), MOS drive air pump runs at 100% power for 10s; c) Receive frequency signal range of duty cycle (40%, 60%), MOS drive solenoid valve to open at 60% opening degree for 5s, air pump to run at 100% power for 10s; d) Receive frequency signal range of duty cycle (60%, 80%), MOS drives solenoid valve to open at 80% opening degree for 5s, air pump runs at 100% power for 10s; e) Receive frequency signal range of duty cycle (80%, 100%), MOS drive solenoid valve to open at 100% opening degree for 10s, air pump to run at 100% power for 15s; When cleaning begins, if the ambient temperature is detected to be lower than the set temperature threshold, the PTC heating module is controlled to work. The temperature threshold is 1℃-6℃.

Images