Unmanned vehicle braking failure power cut-off method and system based on weight distribution

Abstract

The invention discloses an unmanned vehicle braking failure power cut-off method and system based on weight distribution, and the method comprises the steps: employing an analytic hierarchy process to construct a multi-level decision model, defining a sensor importance judgment matrix A, solving a matrix feature vector, and determining the initial weight distribution; constructing a probability density function of the characteristics of each sensor by adopting kernel density estimation, and carrying out weight dynamic optimization; carrying out standardization processing on sensor data, carrying out multi-modal data fusion based on a dynamic weight, and calculating a safety coefficient S; presetting a safety threshold and a time tolerance window, and triggering risk early warning according to the change of the monitoring safety factor S; the system comprises a sensing monitoring module, an unmanned vehicle braking state analysis and control module and a power emergency cut-off device. Through dynamic sensor weight allocation, a data fusion mechanism and power cut-off, in the case of brake failure crisis, response is quickly made, power supply is cut off, and safe parking of the unmanned vehicle is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned vehicle driving safety, and particularly to a method and system for power cut-off in case of brake failure of an autonomous vehicle based on weight allocation. Background Art

[0002] In the process of the current transportation field moving towards intelligence, autonomous vehicles, as a highly disruptive innovation, are gradually changing people's travel modes. However, the prerequisite for their wide application is to ensure extremely high safety, especially in the critical link of braking, where any minor mistake may lead to catastrophic consequences.

[0003] The safe braking of autonomous vehicles highly depends on the coordinated operation of various types of sensors. They are like the "sensory organs" of the vehicle, constantly monitoring various key parameters of the vehicle operation. However, many thorny problems have emerged in the actual application of the existing technology.

[0004] The fixed sensor weights are a major problem. In traditional braking control schemes, sensors such as wheel speed sensors, acceleration sensors, and brake pressure sensors each perform their own functions. However, the weights assigned to them remain fixed once set. This shows obvious drawbacks when facing complex and changeable actual road conditions. For example, when the vehicle is driving on a wet road surface, the data fluctuation of the wheel speed sensor is more intense than on a dry road surface, and the change in the vehicle's motion state it reflects is more critical. At this time, if the fixed weights are still followed, it will lead to misjudgment of the brake failure risk; another example is on a steep slope section, the conversion of the vehicle's gravitational potential energy has a great impact on acceleration, and the importance of the acceleration sensor data suddenly increases. The fixed weight mode cannot accurately capture this change, easily leading to missed judgments and leaving the vehicle unaware of being in a dangerous situation.

[0005] The poor real-time performance of data fusion seriously restricts the emergency response ability of the braking system. Multi-source heterogeneous sensors continuously collect data in different formats and with different characteristics. To comprehensively understand the vehicle braking state, it is necessary to fuse this data. However, the current fusion algorithms are often too complex, introducing a large amount of time delay in the calculation process. In an emergency braking scenario, every second is crucial. This delay makes the vehicle unable to make a braking decision in a timely manner based on accurate comprehensive information, increasing the possibility of accidents.

[0006] The lack of redundancy mechanisms makes existing systems extremely vulnerable. In most cases, existing driverless braking systems overly rely on a single braking module and lack multiple safeguards. Once a key sensor fails, or the data transmitted by different sensors conflicts with each other, the entire braking control system is prone to chaos, which may further lead to misoperations. Especially in the face of an extreme situation such as brake failure, there is no reliable emergency power cut-off mechanism to quickly cut off the power source and prevent the vehicle from accelerating continuously due to loss of control, resulting in a further escalation of the danger.

[0007] In summary, how to ensure the driving safety of driverless vehicles and improve the real-time performance and reliability of emergency braking when driverless vehicles are speeding out of control is a technical problem that urgently needs to be solved. Summary of the Invention

[0008] Object of the Invention: Aiming at the problem of brake failure encountered by existing driverless vehicles during driving and the many defects of traditional emergency handling means for brake failure, such as rigid sensor weights, inefficient data fusion, and slow emergency decision-making, the present invention proposes a method and system for cutting off power in case of brake failure of a driverless vehicle based on weight distribution, aiming to quickly respond by means of an intelligent and dynamic sensor weight allocation strategy, an efficient and accurate data fusion mechanism, and a decisive and timely power cut-off measure when a brake failure crisis emerges, cut off the power supply, ensure that the driverless vehicle stops safely and smoothly, and effectively avoid the occurrence of serious accidents.

[0009] Technical Solution: The method for cutting off power in case of brake failure of a driverless vehicle based on weight distribution according to the present invention is achieved through the layout of each sensor, the intelligent allocation of weights, the accurate fusion of data, and the effective early warning of risks in an all-round way.

[0010] The method for cutting off power in case of brake failure of a driverless vehicle based on weight distribution according to the present invention includes the following steps:

[0011] (1) Integrate a brake pressure sensor at key nodes of the hydraulic brake pipeline, install an acceleration sensor at the vehicle's center of mass to capture the three-dimensional motion state, embed a wheel speed sensor inside each wheel hub to monitor the rotational speed change in real time, and deploy a lidar at the 360° panoramic viewing point on the roof to ensure all-round environmental perception. This step strictly follows the principles of vehicle physical characteristics and maximizing space utilization.

[0012] (2) Use the analytic hierarchy process to construct a multi-level decision-making model, and determine the sensor importance judgment matrix A = [a ij 4×4 where a ij ​To quantify the relative criticality of the i-th sensor and the j-th sensor based on the nine-scale method (the subscripts 1, 2, 3, and 4 represent the brake pressure sensor, acceleration sensor, lidar, and wheel speed sensor respectively, and the 1-4 digital subscripts appearing later are equivalent to them). Initial weight Among them, is the unnormalized weight; by solving the matrix eigenvector W=(w1, w2, w3, w4) and satisfying Determine the initial weight distribution and ensure the rationality of the weight through the consistency test (CR<0.1).

[0013] (3) Based on the historical database (covering multi-condition data sets of dry, wet, ice and snow) and the expert experience database (including the physical modeling results of brake failure cases), use kernel density estimation to construct the key feature f of each sensor i Probability density function of Among them, K(x) is the Gaussian kernel function and h is the bandwidth; set the initial judgment feature threshold for each sensor Among them, p accept is the risk probability.

[0014] (4) Through the CAN-FD bus and SPI interface of the high-precision control chip, collect the sensor data stream in parallel at a millisecond-level sampling frequency.

[0015] (5) Based on the real-time road surface conditions (friction coefficient μ, slope θ) and environmental perception data, construct a fuzzy rule base, and generate the weight adjustment factor Δw of each sensor under different road surface conditions through fuzzy inference i , and apply the weighted average method to defuzzify; use reasoning and calculation to realize the dynamic optimization of the weight.

[0016] (6) Perform Z-score normalization on the original data x of each sensor i Among them, μ is the mean value within the sliding time window, and σ i is the standard deviation to eliminate the influence of dimension difference and non-self-sufficiency. Subsequently, based on the dynamic weight w' i Perform multi-modal data fusion and calculate the comprehensive safety factor i The Kalman filter is introduced in the fusion process to optimize the instantaneous fluctuation to ensure the smoothness and reliability of S. With the help of standardization and fusion technologies, multi-source data is integrated into key decision-making indicators.

[0017] (7) Preset the safety threshold S th and the time tolerance window t max , and monitor the dynamic change of S in real time. If S th <S continuously exceeds t max ​, triggering the second-level risk warning; sending an excitation signal to the power cut-off module, the power cut-off device cuts off the power supply electrical connection, and the unmanned vehicle stops.

[0018] In step (2), the consistency test index Among them, CI and RI are consistency indicators. In the fourth-order judgment matrix, the RI value is 0.90. λ max is the maximum characteristic root of the judgment matrix. Through strict consistency test, the reliability of weight decision is guaranteed.

[0019] In step (2), the multi-level decision model is divided into two layers: a target layer and a criterion layer, wherein the target layer is the safety factor S, and the criterion layer is the brake pressure sensor data, wheel speed data, acceleration sensor output data, and lidar output data.

[0020] In step (2), the nine-scale method includes: scale 1 indicates that the two elements are equally important; scale 3 indicates that the former is slightly more important than the latter; scale 5 indicates that the former is significantly more important than the latter; scale 7 indicates that the former is extremely important than the latter; scale 9 indicates that the former is strongly important than the latter; scales 2, 4, 6, and 8 indicate the middle values ​​of the above adjacent judgments; the reciprocal of scales 1-9 indicates the importance of the corresponding two factors when the order of comparison is swapped.

[0021] In step (3), historical data is extracted from the vehicle black box, cloud database and simulation test platform, covering brake failure cases under typical road conditions such as dry, slippery, icy, and gravel. The data dimensions include: brake pressure sensor: hydraulic pipeline pressure time series data (unit: MPa); acceleration sensor: x, y axis acceleration and x, y, z axis deflection angle; wheel speed sensor: four-wheel speed data; laser radar: obstacle distance (d), using sliding window outlier detection and mean shift algorithm to eliminate noise data, ensuring that the data set f i High signal-to-noise ratio.

[0022] In step (4), the brake pressure signal is converted into a voltage through the piezoelectric effect, the acceleration data is solved by the MEMS gyroscope to calculate the three-axis acceleration vector, the wheel speed pulse signal is captured by the Hall sensor and converted into a digital square wave, and the lidar point cloud data is extracted through TOF ranging and clustering algorithms to extract key obstacle features. During the acquisition process, timestamp alignment and sliding window filtering technology are used to eliminate timing deviations and noise interference between multiple sensors. Through hardware interfaces and sophisticated signal processing, the high quality and real-time nature of the collected data are ensured.

[0023] In step (5), a fuzzy rule base is established based on different friction coefficients μ and slopes θ, such as dry roads, slippery roads, and ice and snow roads, etc. For typical road conditions, rules are customized to improve the system adaptability.

[0024] In step (7), if 0.8·S th <S < 1·S th , a first-level warning is triggered to limit the vehicle power.

[0025] The power cut-off system of the present invention includes a sensing and monitoring module, an analysis and control module for the braking state of the driverless vehicle, and a power emergency cut-off device. Among them, the sensing and monitoring module covers a braking pressure sensing unit, a wheel speed sensing unit, an acceleration sensing unit, and a lidar sensing unit; the core component of the braking pressure sensing unit is a braking pressure sensor integrated at the key nodes of the hydraulic braking pipeline. It uses the piezoelectric effect principle. When the braking system pressure P changes, the force T acting on the sensor also changes. According to D = dT (d is the piezoelectric strain constant), the change in pressure T also causes a change in the electric displacement D. And according to E = βD, the electric field strength E also changes accordingly, and the output voltage V = E·d, so as to achieve real-time monitoring of the braking pressure; the wheel speed sensing unit mainly relies on wheel speed sensors embedded inside each wheel hub. It captures wheel speed pulse signals with Hall sensors and quickly converts them into digital square waves, so as to track the dynamic changes of the wheel speed in real time, and then accurately reflect the driving speed and driving state of the vehicle; the acceleration sensing unit takes the acceleration sensor installed at the vehicle center of mass as the key, and uses a MEMS gyroscope to calculate the three-axis acceleration vector, so as to extremely accurately capture the three-dimensional motion state of the vehicle. Whether the vehicle is in the acceleration, deceleration or turning operation stage, the acceleration sensor can quickly sense and timely feedback relevant data to help the system comprehensively master the motion trend of the vehicle; the lidar sensing unit carefully arranges lidars at the 360° panoramic viewing points on the roof. It emits laser beams and accurately measures the time of the reflected light, so as to accurately calculate the depth information of the obstacles and the vehicle, and at the same time conduct a full-range scan of the surrounding environment to provide accurate and comprehensive environmental perception information for the vehicle.

[0026] The analysis and control module for the braking state of the driverless vehicle includes a state analysis unit, a braking regulation unit, a power cut-off system control unit, and a visual interaction interface; the state analysis unit is connected to the sensing and monitoring module and the visual interaction interface through a communication interface. It analyzes the current vehicle driving scenario based on the data transmitted back by the braking pressure sensor, wheel speed sensor, acceleration sensor, and lidar, and selects different sensor fusion weights according to different scenarios. Subsequently, the sensor data is standardized, and further the standardized sensor data is weighted and fused to obtain a safety factor S. The safety factor S is judged with the safety threshold S th to conduct a two-level response mechanism: when 0.8·Sth <S<1·S th When S th <S,且持续时间大于时间容忍窗口t max , then a power cut-off signal is output to the power cut-off system control unit, triggering the power emergency cut-off procedure.

[0027] When the power emergency cut-off device receives a power cut-off signal from the brake status analysis and control module of the unmanned vehicle, the device will act quickly to destroy the electrical connection of the vehicle body, cut off the power source of the vehicle, causing the vehicle to lose power and gradually slow down until it stops completely, thereby protecting the safety of the vehicle and personnel in the emergency situation of brake failure.

[0028] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0029] (1) The sensor layout in the brake failure power cut-off system of an unmanned vehicle based on weight distribution in the present invention is designed according to the principles of vehicle dynamics and space optimization, and realizes all-round and high-precision perception of the key operating parameters of the vehicle and the surrounding environment, laying a solid foundation for subsequent accurate decision-making.

[0030] (2) The present invention accurately adjusts the sensor weights according to different road conditions, and realizes the transformation from static fixed weights to dynamic intelligent weights with the help of hierarchical analysis method and fuzzy rule base. The sensor weights are flexibly adjusted according to the complex factors of real-time road conditions and vehicle driving status, which improves the accuracy and timeliness of brake failure judgment and effectively avoids the risk of misjudgment and missed judgment.

[0031] (3) The data fusion process of the present invention eliminates the dimensional difference and the influence of stationary and non-stationary characteristics through standardized processing, multimodal fusion and Kalman filter optimization of the raw data of each sensor, significantly improves the accuracy and reliability of the calculation of the comprehensive safety factor, and provides data support for risk warning and emergency decision-making.

[0032] (4) When the brakes fail, the risk warning mechanism of the present invention works closely with the power cut-off mechanism. Through a graded warning strategy, corresponding power control measures are taken under different risk levels. Once the risk is upgraded to a level 2 warning, the power supply electrical connection is decisively cut off to ensure that the unmanned vehicle stops safely, thereby effectively ensuring the safety of vehicle driving.

[0033] (5) The present invention realizes the intelligent monitoring of the braking status of unmanned vehicles, the accurate early warning of the risk of brake failure, and the rapid emergency power cut-off function after failure, which overcomes many difficulties in the field of emergency handling of brake failure of existing unmanned vehicles and improves the driving safety of unmanned vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is the flowchart of the method for power cut-off in case of brake failure of an autonomous vehicle based on weight distribution according to the present invention. Specific implementation manner

[0035] As Figure 1 shown, the power cut-off system for brake failure of the autonomous vehicle in this embodiment includes: a sensing and monitoring module, an analysis and control module for the brake state of the autonomous vehicle, and a power emergency cut-off device. Among them, the sensing and monitoring module includes a brake pressure sensing unit, a wheel speed sensing unit, an acceleration sensing unit, and a lidar sensing unit; the core component of the brake pressure sensing unit is a brake pressure sensor integrated at key nodes of the hydraulic brake pipeline. When the pressure P of the brake system changes, the stress T acting on the sensor also changes. According to D = dT (d is the piezoelectric strain constant), the change in pressure T also causes a change in the electric displacement D, and according to E = βD, the electric field strength E also changes accordingly, and the output voltage V = Ed, thereby achieving real-time monitoring of the brake pressure; the wheel speed sensing unit relies on wheel speed sensors embedded inside each wheel hub, captures wheel speed pulse signals with the help of Hall sensors, and quickly converts the wheel speed pulse signals into digital square waves, so as to track the dynamic changes of the wheel speed in real time, and then accurately reflect the driving speed and driving state of the vehicle; the acceleration sensing unit takes the acceleration sensor installed at the centroid position of the vehicle as the key, and uses the three-axis acceleration vector of the MEMS piezoelectric acceleration sensor to extremely accurately capture the three-dimensional motion state of the vehicle. Whether the vehicle is in the acceleration, deceleration, or turning operation stage, the acceleration sensor quickly senses and timely feeds back relevant data to help the system comprehensively master the motion trend of the vehicle; the lidar sensing unit arranges lidars at 360° panoramic viewing points on the roof of the vehicle. The lidar emits laser beams and accurately measures the time of the reflected light, thereby accurately calculating the depth information between the obstacle and the vehicle, and at the same time performing a full-range scan of the surrounding environment to provide accurate and comprehensive environmental perception information for the vehicle.

[0036] The analysis and control module for the brake state of the autonomous vehicle includes a state analysis unit, a brake regulation unit, a power cut-off system control unit, and a visual interaction interface; the state analysis unit is connected to the sensing and monitoring module and the visual interaction interface through a communication interface, analyzes the current vehicle driving scenario according to the data transmitted back by the brake pressure sensor, wheel speed sensor, acceleration sensor, and lidar, and selects different sensor fusion weights according to different scenarios. Subsequently, the sensor data is standardized, and further the standardized sensor data is weighted and fused to obtain a safety factor S. The safety factor S is compared with the safety threshold S th for judgment, and a two-level response mechanism is adopted: when 0.8·S th < S < 1·S th , the brake regulation unit immediately limits 30% of the power output of the vehicle; when Sth <S and the duration is greater than the time tolerance window t max , then output a power cut-off signal to the power cut-off system control unit to trigger the emergency power cut-off procedure.

[0037] When the power emergency cut-off device receives the power cut-off signal sent by the braking state analysis and control module of the driverless vehicle, it acts quickly to damage the electrical connection of the vehicle body, cut off the power source of the vehicle, make the vehicle lose power and decelerate until it stops, thereby protecting the safety of the vehicle and personnel in the emergency situation of brake failure.

[0038] The method for power cut-off in case of brake failure of a driverless vehicle based on weight assignment in the embodiments of the present invention includes the following steps:

[0039] (1) Integrate the brake pressure sensor into the hydraulic brake pipeline, install the acceleration sensor at the vehicle's center of mass to capture the three-dimensional motion state, embed the wheel speed sensor inside each wheel hub to monitor the rotational speed change in real time, and deploy the lidar at the 360° panoramic viewing points on the roof to ensure all-round environmental perception. Among them, the brake sensor outputs hydraulic pressure, the acceleration sensor outputs the x and y axial accelerations and the x, y, and z axis deflection angles, the wheel speed sensor outputs the rotational speeds of the four wheels, and the lidar outputs the all-round depth.

[0040] (2) Use the analytic hierarchy process to construct a multi-level decision-making model. This multi-level decision-making model is divided into two layers: the target layer and the criterion layer. Among them, the target layer is the safety factor S, and the criterion layer is the data of the brake pressure sensor, the wheel speed data, the output data of the acceleration sensor, and the output data of the lidar. Define the sensor importance judgment matrix A = [a ij 4×4 , where a ij is the relative criticality of the i-th sensor and the j-th sensor quantified based on the nine-scale method (the subscripts 1, 2, 3, 4 represent the brake pressure sensor, the acceleration sensor, the lidar, and the wheel speed sensor respectively, and the 1-4 digital subscripts appearing later are equivalent to them). Specifically:

[0041]

[0042] Through the formula Get the unnormalized weight vector Through the formula Get the normalized weight W = [0.3908, 0.3908, 0.1509, 0.0675], that is, the fusion weight of the brake pressure sensor and the acceleration sensor is 0.3908, the fusion weight of the lidar is 0.1509, and the fusion weight of the wheel speed sensor is 0.0675. According to the formula Find the maximum eigenvalue, according to​ Calculate the consistency index C.I., and then calculate the consistency test index according to the 4th-order random consistency index R.I. = 0.90 It is calculated that C.R. = 0.016 < 0.1, which proves that the consistency degree of the judgment matrix A is within the allowable range.

[0043] (3) Based on the historical database (covering datasets of multiple working conditions such as dry, slippery, ice and snow) and the expert experience database (including the physical modeling results of brake failure cases), kernel density estimation is used to set the initial judgment threshold T for the brake pressure sensor, acceleration sensor, lidar and wheel speed sensor i . The specific method includes the key feature f of each sensor i , and use kernel density estimation to construct its probability density function where K(x) is the Gaussian kernel function and h is the bandwidth, which is adaptively selected by the Silverman rule. The feature threshold of each sensor where p accept is the risk probability set according to expert experience.

[0044] (4) Through the CAN-FD bus and SPI interface of the high-precision control chip, the sensor data stream is collected in parallel at a millisecond-level sampling frequency. Among them, the braking pressure signal P acts on the sensor cross-section to generate a force F = PS; where S is the contact area between the piezoelectric component of the sensor and the hydraulic cylinder. This force is applied to the piezoelectric material element, and the element generates a piezoelectric effect. For the entire contact surface of the element, the output voltage where g is the piezoelectric constant of the material; the acceleration sensor generates an external applied force F on a piezoelectric element under the action of inertial force a , so as to output a voltage under the piezoelectric effect where g is the 3×3 material piezoelectric constant matrix, and the accelerations of the x and y axes and the deflection angles of the x, y, and z axes are output; the wheel speed pulse signal is captured by the Hall sensor, and the output Hall voltage V H The calculation formula of is where I is the current passing through the Hall sensor, B is the magnetic induction intensity of the externally applied magnetic field, R H is the Hall coefficient, and d is the material thickness; the lidar extracts the depth Depth information of other vehicles, pedestrians, trees, building obstacles, etc. through TOF ranging and clustering algorithms, and the expression is where c is the speed of light and Δt is the time difference between the lidar emitting laser and receiving laser. A unified clock source based on the NTP protocol or a hardware trigger signal is used to assign a time stamp accurate to the millisecond level to all sensor data. In view of the differences in the sampling rates of each sensor, asynchronous sampling data is remapped to the same time reference point through linear interpolation during the data preprocessing stage.

[0045] (5) Construct a fuzzy rule base based on real-time road conditions (friction coefficient μ, slope θ) and environmental perception data to achieve dynamic weight adjustment:

[0046] (5.1) Establish a fuzzy rule base through different friction coefficients μ and slopes θ, such as dry roads, slippery roads, and ice and snow roads. When μ > 0.7, it is judged as a dry road; when 0.3 > μ > 0.7, it is judged as a slippery road; when 0.1 > μ > 0.3, it is judged as an ice and snow road. In addition, there are also uphill (θ > 0°), downhill (θ < 0°), and steep slope (|θ| > 30°) situations.

[0047] (5.2) Generate the weight adjustment factor Δw of each sensor under different road conditions through a fuzzy inference engine i , and the weight adjustment for different road conditions is as follows: Under dry road conditions, Δw 制动压力 = 0.13, Δw 加速度 = -0.05, Δw 轮速 = -0.08, Δw 雷达 = 0; Under slippery road conditions, Δw 制动压力 = -0.10, Δw 加速度 = -0.05, Δw 轮速 = 0.08, Δw 雷达 = 0.07; Under ice and snow road conditions, Δw 制动压力 = 0.07, Δw 加速度 = -0.05, Δw 轮速 = -0.08, Δw 雷达 = 0.06; Under uphill conditions, Δw 制动压力 = 0.05, Δw 加速度 = -0.03, Δw 轮速 = -0.02, Δw 雷达 = 0; Under downhill conditions, Δw 制动压力 = 0.03, Δw 加速度 = -0.02, Δw 轮速 = -0.05, Δw 雷达 = 0.04; Under steep slope conditions, Δw 制动压力 = 0.05, Δw 加速度 = -0.05, Δw 轮速 = -0.08, Δw 雷达 = 0.08.

[0048] (6) Perform Z-score standardization on the original sensor data x i , where μ is the mean value within the sliding time window, and σ i is... iis the standard deviation to eliminate the influence of dimension difference and non - independence. Subsequently, based on the dynamic weight w' i perform multi - modal data fusion and calculate the comprehensive safety factor The Kalman filter is introduced in the fusion process to optimize the instantaneous fluctuation, ensuring the smoothness and reliability of S. With the help of standardization and fusion technologies, multi - source data is integrated into key decision - making indicators.

[0049] (7) Preset the safety threshold S th and the time tolerance window t max , and monitor the dynamic changes of S in real - time. If S th <S continuously exceeds t max , trigger a secondary risk warning:

[0050] (7.1) Primary warning (0.8·S th <S<1·S th ): Limit the vehicle's power by 30%; relieve the emergency situation by regulating the power.

[0051] (7.2) Secondary warning S th <S and the duration exceeds t max : Send an excitation signal to the power cut - off module, the power cut - off device cuts off the electrical connection of the power supply, and the unmanned vehicle stops; take ultimate measures resolutely to ensure the safety of the vehicle.

[0052] The visualization interface of the present invention realizes the display of measurement results and provides control functions. The interface adopts a graphical interface or a command - line interface, and the specific design is selected according to user requirements and system complexity. In addition, the visualization interface is also responsible for displaying the status information of the vehicle. For example, after the power emergency cut - off device performs the power cut - off operation, the interface displays "Vehicle failure, power supply has been cut off" to notify the user that the vehicle has entered the emergency braking state. Such a design not only ensures the transparency of the emergency braking process but also provides instant feedback to the user, clearly indicating the current state of the vehicle and the safety measures taken.

[0053] Finally, it should be noted that the purpose of publishing the embodiments is to help further understand the present invention. However, those skilled in the art can understand that: within the spirit and scope of the present invention and the appended claims, various substitutions and modifications are possible. Therefore, the present invention should not be limited to the content disclosed in the embodiments, and the scope of protection claimed by the present invention is defined by the scope of the claims.

Claims

1. A method for cutting off power of brake failure of unmanned vehicles based on weight distribution, characterized in that: The following steps are involved: (1) The brake pressure sensor is integrated into the hydraulic brake line node, the acceleration sensor is installed at the center of mass of the vehicle, the wheel speed sensor is embedded in the wheel hub, and the lidar is placed on the roof; (2) A multi-level decision model is constructed using the hierarchical analysis method to determine the sensor importance judgment matrix A = [a ij ] 4×4 , where a ij To quantify the relative criticality of the i-th sensor and the j-th sensor based on the nine-scale method; initial weight in, is the unstandardized weight; by solving the matrix eigenvector W = (w1, w2, w3, w4) and satisfying Determine the initial weight distribution. When the consistency test index CR is less than 0.1, the consistency test is passed, where CR is the consistency index; (3) Use kernel density estimation to construct the key features f of each sensor i The probability density function of Among them, K(x) is the Gaussian kernel function, h is the bandwidth; set the initial judgment feature threshold for each sensor Among them, p accept is the risk probability; (4) Collect sensor data streams in parallel through the chip’s CAN-FD bus and SPI interface; (5) Based on the road friction coefficient μ, slope θ and environmental perception data, a fuzzy rule base is constructed to generate the weight adjustment factor Δw for each sensor i , perform dynamic weight optimization: (6) For each sensor’s raw data x i Perform Z-score standardization Among them, μ i is the mean value in the sliding time window, σ i is the standard deviation; based on the dynamic weight w' i Perform multi-modal data fusion and calculate comprehensive safety factor (7) Preset safety threshold S th With time tolerance window t max , monitor the dynamic changes of S in real time; if 1≦S th More than t max , triggering the second-level risk warning, sending an excitation signal to the power cut-off module, the power cut-off device cuts off the power supply electrical connection, and the unmanned vehicle stops.​ 2. The method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution according to claim 1, characterized in that: In step (2), the consistency test index Among them, CI and RI are consistency indicators.

3. The method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution according to claim 1, characterized in that: In step (2), the multi-level decision model is divided into a target layer and a criterion layer, the target layer is the safety factor S, and the criterion layer is the brake pressure sensor data, wheel speed data, acceleration sensor output data and lidar output data.

4. The method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution according to claim 1, characterized in that: In step (3), kernel density estimation is used to construct the key features f of each sensor based on the historical database containing multiple operating conditions such as dry, slippery, and refrigerator and the specialized experience database containing brake failure. i The probability density function of 5. The method for cutting off power of unmanned vehicle brake failure based on weight distribution according to claim 1 is characterized in that: In step (4), the brake pressure signal is converted into a voltage through the piezoelectric effect, the acceleration is solved by the MEMS gyroscope to calculate the three-axis acceleration vector, the wheel speed pulse signal is captured by the Hall sensor and converted into a digital square wave, and the lidar point cloud data is used to extract obstacle features through TOF ranging and clustering algorithm.

6. The method for cutting off power of unmanned vehicle brake failure based on weight distribution according to claim 1 is characterized in that: In step (4), in the process of parallel acquisition of sensor data streams, timestamp alignment and sliding window filtering are used to eliminate timing deviation and noise interference among multiple sensors.

7. The method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution according to claim 1, characterized in that: In step (5), the weight adjustment factor Δw of each sensor under different road conditions is generated by fuzzy reasoning. i , and apply the weighted average method to defuzzify and perform dynamic weight optimization.

8. The method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution according to claim 1, characterized in that: In step (6), based on the dynamic weight w' i Perform multi-modal data fusion and calculate comprehensive safety factor The Kalman filter is introduced into the fusion process to optimize the instantaneous fluctuation.

9. The method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution according to claim 1, characterized in that: In step (7), if 0.8≤S<1, a first-level warning is triggered and the vehicle power is restricted.

10. A system for the method for cutting off power when brake failure occurs in an unmanned vehicle based on weight distribution as claimed in claim 1, characterized in that: It includes a sensor monitoring module, an unmanned vehicle braking state analysis and control module, and a power emergency cut-off device; the sensor monitoring module includes a brake pressure sensing unit, a wheel speed sensing unit, an acceleration sensing unit, and a lidar sensing unit; the unmanned vehicle braking state analysis and control module includes a state analysis unit, a brake control unit, a power cut-off system control unit, and a visual interactive interface.

Images