Apparatus and method for controlling airbag of vehicle

Abstract

An apparatus and method for controlling vehicle airbags ensures robust airbag deployment logic and more effectively protects passengers. This is achieved by determining whether to deploy the airbag based on a posterior probability of human injury calculated using a human injury probability model and Bayesian network learning (feedback learning). The apparatus includes: a human injury probability calculator configured to calculate a conditional probability of human injury and a predicted probability of human injury based on vehicle motion information measured by sensing devices; a learner configured to calculate the posterior probability of human injury by performing probability-based real-time feedback machine learning based on the conditional probability of human injury and the predicted probability of human injury; and an airbag deployment determiner configured to determine whether to deploy the airbag based on the posterior probability of human injury.

Description

Technical Field

[0001] This disclosure relates to a vehicle, and more specifically, to airbag deployment control in that vehicle. Background Technology

[0002] Airbags are devices used to protect passengers from the impact of a vehicle collision, and together with seat belts, they are a typical passenger protection system in a vehicle. When a collision is detected by sensors, the working gas device is triggered, and the exploding gas causes the airbag to inflate (deploy) instantly to protect the passengers. Therefore, the shorter the time between a vehicle collision and airbag deployment, the better.

[0003] However, it is necessary to determine whether airbags should be deployed based on whether the impact of a vehicle collision is strong enough to warrant deployment or weak enough to not require deployment. If airbags fail to deploy when they are essential to protect passengers, the passengers will not be protected. Conversely, deploying airbags when the impact is weak enough not to warrant deployment is undesirable because it is unnecessary and incurs the cost of rearranging the airbags (e.g., replacement, repair).

[0004] In other words, it is necessary to accurately determine whether to deploy airbags based on the severity of passenger injuries in a vehicle collision. Summary of the Invention

[0005] One aspect of this disclosure is to provide an airbag control device and method that can determine whether to deploy an airbag based on a posterior probability of human injury calculated through a human injury probability model and Bayesian network learning (feedback learning) to ensure the robustness of the airbag deployment logic and more effectively protect passengers.

[0006] Additional aspects of this disclosure are set forth in part below, and these aspects should be apparent in part from the description or may be understood by practicing this disclosure.

[0007] According to one aspect of this disclosure, an apparatus for controlling a vehicle airbag is provided. The apparatus includes: a human injury probability calculator configured to calculate a conditional probability of human injury and a predicted probability of human injury based on vehicle motion information measured by a sensing device; a learner configured to calculate a posterior probability of human injury based on the conditional probability of human injury and the predicted probability of human injury by performing probability-based real-time feedback machine learning; and an airbag deployment determiner configured to determine whether to deploy the airbag based on the posterior probability of human injury.

[0008] The learner can be configured to calculate the prior probability of human injury based on the conditional probability of human injury and the predicted probability of human injury, and to calculate the posterior probability of human injury by multiplying the conditional probability of human injury by the prior probability of human injury.

[0009] Probabilistic real-time feedback machine learning can be configured to update the prior probability of human injury by feeding back the current posterior probability of human injury to the previous prior probability of human injury.

[0010] The conditional probability of human injury can be configured to be calculated using the following formula 1.

[0011] [Formula 1]

[0012] P(x t |u t x t-1 ), P(z t |x t )

[0013] Expression P(x) t |u t x t-1 ) represents the predicted probability of human injury at the current time point (t) based on the measurements of collision sensors 102 and 106 and the probability of human injury at the previous time (t-1).

[0014] Expression P(z) t |x t ) represents the predicted probability of bodily injury based on passenger injuries measured through simulation.

[0015] Term x t This represents the actual probability of bodily injury for each of the six regions of the head, neck, and chest at the current time point (t).

[0016] Term u t This represents the measurement value of the sensing device 250 at the current time point (t).

[0017] Term x t-1 This represents the actual probability of human injury at the previous time point (t-1).

[0018] Term z t This represents the predicted probability of bodily injury for each of the six regions of the head, neck, and chest at the current time point (t).

[0019] The prior probability of human injury can be configured to be calculated using the following formula 2.

[0020] [Formula 2]

[0021]

[0022] Expression P(x) t |u t x t-1The value represents the predicted probability of human injury at the current time point (t), based on the measured values ​​of the collision sensors 102 and 106 at the previous time (t-1).

[0023] expression bel(x) t-1 ) represents the posterior probability of the previous (t-1) human injury.

[0024] The posterior probability of human injury can be configured to be calculated using the following formula 3.

[0025] [Formula 3]

[0026]

[0027] The term η represents the normalization factor.

[0028] Expression P(z) t |x t ) represents the predicted probability of bodily injury based on passenger injuries measured through simulation.

[0029] expression This represents the prior probability of human injury.

[0030] The airbag deployment determiner can be configured to determine whether to deploy the airbag based on the posterior probability of human injury exceeding a preset reference value.

[0031] Vehicle motion information can include vehicle acceleration and angular velocity values, collision values, pressure values, roll values, pitch angle values, and yaw angle values.

[0032] According to one aspect of this disclosure, a method for controlling a vehicle airbag is provided. The method includes: calculating a conditional probability of human injury and a predicted probability of human injury based on vehicle motion information measured by a sensing device using a human injury probability calculator; calculating a posterior probability of human injury based on the conditional probability of human injury and the predicted probability of human injury by a learner through performing probability-based real-time feedback machine learning; and determining whether to deploy the airbag based on the posterior probability of human injury using an airbag deployment determiner.

[0033] The method may also include a learner calculating a prior probability of human injury based on a conditional probability of human injury and a predicted probability of human injury, and may also include a learner calculating a posterior probability of human injury by multiplying the conditional probability of human injury by the prior probability of human injury.

[0034] Probabilistic real-time feedback machine learning can be configured to update the prior probability of human injury by feeding back the current posterior probability of human injury to the previous prior probability of human injury.

[0035] The conditional probability of human injury can be configured to be calculated using the following formula 1.

[0036] [Formula 1]

[0037] P(x t |u t x t-1 ), P(z t |x t )

[0038] Expression P(x) t |u t x t-1 ) represents the predicted probability of human injury at the current time point (t) based on the measurements of collision sensors 102 and 106 and the previous (t-1) probability of human injury.

[0039] Expression P(z) t |x t ) represents the predicted probability of bodily injury based on passenger injuries measured through simulation.

[0040] Term x t This represents the actual probability of bodily injury for each of the six regions of the head, neck, and chest at the current time point (t).

[0041] Term u t This represents the measurement value of the sensing device 250 at the current time point (t).

[0042] Term x t-1 This represents the actual probability of human injury at the previous time point (t-1).

[0043] Term z t This represents the predicted probability of bodily injury for each of the six regions of the head, neck, and chest at the current time point (t).

[0044] The prior probability of human injury can be configured to be calculated using the following formula 2.

[0045] [Formula 2]

[0046]

[0047] Expression P(x) t |u t x t-1 The value represents the predicted probability of human injury at the current time point (t), based on the measured values ​​of the collision sensors 102 and 106 at the previous time (t-1).

[0048] expression bel(x) t-1) represents the posterior probability of the previous (t-1) human injury.

[0049] The posterior probability of human injury can be configured to be calculated using the following formula 3.

[0050] [Formula 3]

[0051]

[0052] The term η represents the normalization factor.

[0053] Expression P(z) t |x t ) represents the predicted probability of bodily injury based on passenger injuries measured through simulation.

[0054] expression This represents the prior probability of human injury.

[0055] The airbag deployment determiner can be configured to determine whether to deploy the airbag based on the posterior probability of human injury exceeding a preset reference value.

[0056] Vehicle motion information can include vehicle acceleration and angular velocity values, collision values, pressure values, roll values, pitch angle values, and yaw angle values.

[0057] According to another aspect of this disclosure, an apparatus for controlling a vehicle airbag is provided. The apparatus includes: a human injury probability calculator configured to calculate a conditional probability of human injury and a predicted probability of human injury based on vehicle motion information measured by a sensing device; a learner configured to calculate a prior probability of human injury based on the conditional probability of human injury and the predicted probability of human injury, calculate a posterior probability of human injury by multiplying the conditional probability of human injury by the prior probability of human injury, and update the prior probability of human injury by feeding back the current posterior probability of human injury to the previous prior probability of human injury through probability-based real-time feedback machine learning; and an airbag deployment determiner configured to determine whether to deploy the airbag based on the posterior probability of human injury.

[0058] According to another aspect of this disclosure, a method for controlling a vehicle airbag is provided. The method includes: calculating a conditional probability of human injury and a predicted probability of human injury based on vehicle motion information measured by a sensing device using a human injury probability calculator; calculating a prior probability of human injury based on the conditional probability of human injury and the predicted probability of human injury using a learner; calculating a posterior probability of human injury by the learner by multiplying the conditional probability of human injury by the prior probability of human injury; updating the prior probability of human injury by the learner by feeding back the current posterior probability of human injury to the previous prior probability of human injury through probability-based real-time feedback machine learning; and determining whether to deploy the airbag based on the posterior probability of human injury using an airbag deployment determineer. Attached Figure Description

[0059] These and / or other aspects of this disclosure should become apparent and more readily understood from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0060] Figure 1 This is a diagram illustrating a vehicle according to an embodiment.

[0061] Figure 2 This is a diagram illustrating an airbag control device according to an embodiment.

[0062] Figure 3 This is a diagram illustrating a method for controlling the deployment of airbags in a vehicle according to an embodiment.

[0063] Figure 4 It is shown Figure 3 The diagram illustrates the probabilistic Bayesian network machine learning approach to the airbag deployment control method.

[0064] Figure 5 This is a diagram comparing the airbag deployment determination results according to the embodiment with the airbag deployment determination results of the prior art. Detailed Implementation

[0065] Figure 1 This is a diagram illustrating a vehicle according to an embodiment. When a component, device, element, etc., of this disclosure is described as having a purpose or performing an operation, function, etc., the component, device, or element shall be considered herein as being "configured" to satisfy that purpose or perform that operation or function.

[0066] refer to Figure 1 Collision sensors 102 and 106, pressure sensor 104, angular velocity sensor 110, and airbags 108 and 112 are installed in the vehicle.

[0067] At least one pair of frontal collision sensors 102 are mounted on the front of the vehicle to detect whether a frontal collision has occurred and the intensity of the collision. At least one pair of side impact sensors 104 are mounted on both sides of the vehicle to detect the pressure applied to the sides of the vehicle. At least one pair of side collision sensors 106 are mounted on both sides of the vehicle to detect whether a side collision has occurred and the intensity of the collision.

[0068] In the vehicle according to the embodiment, in addition to the collision sensors 102 and 106 and the pressure sensor 104, at least one pair of angular velocity sensors 110 for detecting the direction of collision may also be installed.

[0069] Airbags 108 and 112 may include a front airbag 108 and a side airbag 112. At least one pair of front airbags 108 are respectively installed on the driver's seat front side and the passenger seat front side. The side airbags 112 are respectively installed on the driver's seat left side and the passenger seat right side.

[0070] In order to ensure vehicle motion information, in addition Figure 1 The vehicle may also include other types of sensors for measuring acceleration, angular velocity, impact, pressure, roll, pitch, and yaw values, as shown in the sensor diagram.

[0071] Figure 2 This is a diagram illustrating an airbag control device according to an embodiment.

[0072] refer to Figure 2 The sensing device 250 is covered Figure 1 The term refers to all sensors 102, 104, 106, and 110 described herein, as well as the sensors used to obtain vehicle motion information.

[0073] according to Figure 2 The airbag control unit (airbag control unit, ACU) 202 of the vehicle in the illustrated embodiment can determine whether to deploy airbags 108 and 112 based on the measurement results of the sensing device 250 and generate an airbag deployment command. The airbag control unit 202 may include application software (ASW) 204, a human injury probability model 206, Bayesian network logic 208, airbag deployment determination logic 210, and base software (BSW) 212.

[0074] ASW 204 can drive the application software installed in the airbag control unit 202.

[0075] A human injury probability model 206 (human injury probability calculator) can be set up to calculate the predicted probability of human injury based on the measurement results of the sensing device 250 when a vehicle collides with another vehicle or obstacle. The predicted probability of human injury calculated by the human injury probability model 206 can refer to the probability of human injury predicted by inputting the measurement results of the sensing device 250 at the time of the vehicle collision into a human injury model obtained through various experiments, such as simulation. In the human injury probability model 206, a relatively simplified human anatomy model can be applied to keep the computational load at an appropriate level when calculating the probability of human injury.

[0076] Bayesian network logic (learning device) 208 can calculate the posterior probability of human injury from the predicted probability of human injury using probabilistic machine learning, which is the result of the calculation of the human injury probability model 206. Bayesian network logic 208 can also determine whether to deploy airbags 108 and 112 based on the calculated posterior probability of human injury. However, Bayesian network logic 208 can obtain a more reliable basis for determining whether to deploy airbags by correcting errors through feedback from the posterior probability value of human injury to update the prior probability of human injury from the previous cycle. In this document, according to an embodiment, the posterior probability of human injury is the result of machine learning in Bayesian network logic 208 and is the basis for determining whether to deploy airbags 108 and 112.

[0077] The airbag deployment determination logic 210 (airbag deployment determination device) can determine whether to deploy airbags 108 and 112 based on the posterior probability value of human injury calculated by Bayesian network logic 208. When the posterior probability value of human injury exceeds a predetermined reference value, the airbag deployment determination logic 210 can generate an airbag deployment command, thereby enabling the deployment of airbags 108 and 112.

[0078] BSW 212 can drive the underlying software installed in the airbag control unit 202. The airbag deployment command generated by the airbag deployment determination logic 210 can be transmitted to airbags 108 and 112 via BSW 212.

[0079] Figure 3 This is a diagram illustrating a method for controlling the deployment of airbags in a vehicle according to an embodiment.

[0080] refer to Figure 3 The airbag control unit 202 can receive vehicle motion information measured by the sensing device 250, including acceleration, angular velocity, impact, pressure, roll, pitch, and yaw angles (310). Vehicle motion information can be transmitted via... Figure 2 The ASW 204 of the airbag control device 202 described herein is transmitted to the human injury probability model 206.

[0081] The human injury probability model 206, which receives vehicle motion information, can calculate the predicted probability of human injury in the current collision based on the vehicle motion information (330). In other words, the vehicle motion information measured by the sensing device 250 can be reflected in a predetermined human injury model to predict the human injury probability value. The predicted human injury probability may include the conditional probability of human injury. The conditional probability of human injury is the same as P (impact I injury severity) and can refer to the probability of the amplitude value of the vehicle collision pulse occurring according to the injury severity. In other words, the conditional probability of human injury can refer to the probability of the injury severity based on the current time of the vehicle collision pulse.

[0082] In other words, several detection values ​​detected by sensing device 250 are input into human injury probability model 206 and used to calculate the probability of human injury severity (conditional probability of human injury) for the head, neck, and chest. Human injury probability model 206 can have a total of 12 degrees of freedom for vehicle translation (XY) and rotation (pitch, roll, and yaw). The measurements from sensing device 250 input into human injury probability model 206 can be calculated as a total of 6 predicted human injury probabilities for each part of the human body. The 6 predicted human injury probabilities can include the probabilities of frontal head injury, lateral head injury, frontal neck injury, lateral neck injury, frontal chest injury, and lateral chest injury. The conditional probability value of human injury can be output from human injury probability model 206 along with the 6 predicted human injury probabilities and input into Bayesian network logic 208.

[0083] The Bayesian network logic 208 can perform machine learning using a Bayesian network based on the predicted probability and conditional probability of human injury calculated by the human injury probability model 206. The Bayesian network logic 208 can also generate prior and posterior probability values ​​of human injury as the results of machine learning (350).

[0084] The probability-based Bayesian network machine learning 350 in the Bayesian network logic 208 may include obtaining the conditional probability of human injury (352), obtaining the prior probability of human injury (354), and calculating the posterior probability of human injury (356). The probability-based Bayesian network machine learning of the vehicle according to this embodiment may include real-time feedback learning of the posterior probability of human injury.

[0085] In other words, the prior probability of human injury in the next cycle can be updated by feeding back the posterior probability of human injury obtained through the calculation of the posterior probability of human injury (356) in real time to the prior probability of human injury in the next cycle (354). This update can correct the posterior probability of human injury, thereby enabling a more accurate determination of whether or not to deploy the airbag.

[0086] First, the Bayesian network logic 208 can obtain the conditional probability of human injury and the prior probability of human injury (352, 354) based on the predicted probability of human injury from the human injury probability model 206. Among them, the prior conditional probability of human injury can be obtained through the following formula 1.

[0087] [Formula 1]

[0088] P(x t |u t x t-1 ), P(z t |x t )

[0089] Formula 1 is described in detail below.

[0090] Expression P(x) t |u t x t-1 ) represents the predicted probability of human injury at the current time point (t) based on the measurements of collision sensors 102 and 106 and the previous (t-1) probability of human injury.

[0091] Expression P(z) t |x t ) represents the predicted probability of bodily injury based on passenger injuries measured through simulation.

[0092] Term x t This represents the actual probability of bodily injury for each of the six regions of the head, neck, and chest at the current time point (t).

[0093] Term u t This represents the measurement value of the sensing device 250 at the current time point (t).

[0094] Term x t-1 This represents the actual probability of human injury at the previous time point (t-1).

[0095] Term z t This represents the predicted probability of bodily injury for each of the six regions of the head, neck, and chest at the current time point (t).

[0096] Then, in Bayesian network logic 208, the prior probability update of human injury can be performed as shown in Equation 2 below (354). The conditional probability of human injury obtained in operation 352 is then multiplied by the prior probability value of human injury obtained in operation 354 to obtain the posterior probability of human injury (356).

[0097] At this point, the prior probability value of human injury can be an updated value obtained by feeding back the previous posterior probability value of human injury. The prior probability and posterior probability of human injury can be obtained using Formula 2 and Formula 3 below, respectively.

[0098] [Formula 2]

[0099]

[0100] Formula 2 is the prior probability of human injury, and a detailed description of Formula 2 is as follows.

[0101] Expression P(x) t |u t x t-1The value represents the predicted probability of human injury at the current time point (t), based on the measured values ​​of the collision sensors 102 and 106 at the previous (t-1) probability of human injury.

[0102] expression bel(x) t-1 ) represents the posterior probability of the previous (t-1) human injury.

[0103] [Formula 3]

[0104]

[0105] Formula 3 is the posterior probability of human injury, and a detailed description of Formula 3 is as follows.

[0106] The term η represents the normalization factor.

[0107] Expression P(z) t |x t ) represents the predicted probability of bodily injury based on passenger injuries measured through simulation.

[0108] expression This represents the prior probability of human injury.

[0109] In this article, references Figure 4 The description describes the calculation of the posterior probability of human injury using a probability-based Bayesian network machine learning method according to an embodiment.

[0110] Figure 4 It is shown Figure 3 The diagram illustrates the probabilistic Bayesian network machine learning approach to the airbag deployment control method.

[0111] See Figure 4 The probabilistic Bayesian network machine learning based on Bayesian network logic 208 can include obtaining the conditional probability of human injury, obtaining the prior probability of human injury, and calculating the posterior probability of human injury. Furthermore, the probabilistic Bayesian network machine learning can include real-time feedback learning of the posterior probability of human injury. In other words, by feeding back the calculated posterior probability of human injury to the prior probability of human injury in real time, the prior probability of human injury can be updated, allowing for a more accurate correction of the decision on whether to deploy airbags. Figure 4 As shown, a loop can be continuously repeated to obtain the posterior probability of human injury. In this case, the output of the previous loop (posterior probability of human injury) becomes the input of the next loop (prior probability of human injury).

[0112] return Figure 3The airbag deployment determination logic 210 can determine airbag deployment (370) based on whether the posterior probability value of human injury calculated by the Bayesian network logic 208 exceeds a preset threshold corresponding to airbag deployment. In other words, when the posterior probability value of human injury exceeds the preset threshold, it is determined that airbags 108 and 112 should be deployed. Conversely, when the posterior probability value of human injury is less than or equal to the preset threshold, it is determined that airbags 108 and 112 should not be deployed.

[0113] In other words, the airbag deployment determination logic 210 can calculate the sum of the frontal damage probabilities based on the values ​​of the frontal head injury probability, the frontal neck injury probability, and the frontal chest injury probability (372).

[0114] Furthermore, the airbag deployment determination logic 210 can identify whether the sum of the probabilities of frontal injury exceeds the preset frontal airbag deployment threshold (374).

[0115] In addition, the airbag deployment determination logic 210 can calculate the sum of the side injury probabilities based on the values ​​of the side head injury probability, the side neck injury probability, and the side chest injury probability (376).

[0116] Furthermore, the airbag deployment determination logic 210 can identify whether the sum of the probabilities of side injury exceeds the preset side airbag deployment threshold (378).

[0117] When the sum of damage probabilities exceeds a preset airbag deployment threshold ("Yes" in 374 or "Yes" in 378), airbag deployment determination logic 210 can generate an airbag deployment command to deploy the airbag (390). When the sum of frontal damage probabilities exceeds a preset frontal airbag deployment threshold ("Yes" in 374), airbag deployment determination logic 210 can generate a frontal airbag deployment command to deploy the frontal airbag. When the sum of side damage probabilities exceeds a preset side airbag deployment threshold ("Yes" in 378), airbag deployment determination logic 210 can generate a side airbag deployment command to deploy the side airbag. Frontal airbag deployment commands and side airbag deployment commands may occur simultaneously.

[0118] Figure 5 This is a diagram comparing the airbag deployment determination results according to the embodiment with the airbag deployment determination results of the prior art.

[0119] like Figure 5 As shown in Figure A, under normal circumstances, the airbag deployment threshold can be determined based on limited collision data. However, in conventional threshold-based control, the airbag deployment threshold is set inaccurately. Therefore, it is impossible to accurately determine the scenarios in which airbags deploy or not.

[0120] Optionally, in embodiments of this disclosure, such as Figure 5As shown in B, by using real-time feedback machine learning based on a Bayesian network of a probabilistic model to predict the probability of human injury, and by determining and correcting whether to deploy airbags based on the predicted probability of human injury, it is possible to more accurately determine whether to deploy airbags.

[0121] According to embodiments of this disclosure, determining whether to deploy an airbag based on the posterior probability of human injury calculated through a probabilistic model and Bayesian network learning (feedback learning) ensures the robustness of the airbag deployment logic and more effectively protects passengers.

[0122] The disclosed embodiments are merely illustrative of technical concepts. Those skilled in the art will understand that various modifications, alterations, and substitutions can be made without departing from the essential characteristics. Therefore, the embodiments and drawings disclosed above are not intended to limit the technical concepts, but rather to describe the technical spirit of this disclosure. The scope of these technical concepts is not limited by the embodiments and drawings. The scope of protection should be interpreted by the appended claims, and all technical concepts within the equivalent scope should be interpreted as being included within that scope of rights.

Claims

1. A device for controlling a vehicle airbag, the device comprising: The human injury probability calculator is configured to calculate the conditional probability of human injury and the predicted probability of human injury based on vehicle motion information measured by a sensing device. The learner is configured to compute the posterior probability of human injury by performing probability-based real-time feedback machine learning based on the conditional probability of human injury and the predicted probability of human injury. as well as An airbag deployment determiner is configured to determine whether to deploy the airbag based on the posterior probability of human injury.

2. The apparatus according to claim 1, wherein, The learner is configured as follows: The prior probability of human injury is calculated based on the previously obtained conditional probability of human injury and the previously obtained predicted probability of human injury. as well as The posterior probability of human injury is calculated by multiplying the currently calculated conditional probability of human injury by the prior probability of human injury.

3. The apparatus according to claim 2, wherein, The probability-based real-time feedback machine learning is configured to update the prior probability of human injury by feeding back the current posterior probability of human injury to the previous prior probability of human injury.

4. The apparatus according to claim 2, wherein, The conditional probability of bodily injury is configured to be calculated using Formula 1: [Formula 1] ,in, It is the predicted probability of human injury at the current time point (t) based on the measurement value of the sensing device and the actual probability of human injury at the previous time point (t-1). It is a predicted probability of bodily injury based on passenger injuries measured through simulation. It refers to the actual human injury event that occurs at the current time point (t). It is the measurement value of the sensing device at the current time point (t). It refers to the actual human injury event that occurred at the previous time point (t-1), and It is a predicted human injury event that occurs at the current time point (t).

5. The apparatus according to claim 2, wherein, The prior probability of bodily injury is configured to be calculated using Formula 2: [Formula 2] ,in, It is based on the measured value of the probability of human injury at the previous time point (t-1) and the predicted probability of human injury at the current time point (t). It is the posterior probability of human injury at the previous time point (t-1). It refers to the actual human injury event that occurs at the current time point (t). It is the measurement value of the sensing device at the current time point (t), and It refers to the actual human injury event that occurred at the previous time point (t-1).

6. The apparatus according to claim 2, wherein, The posterior probability of human injury is configured to be calculated using Formula 3: [Formula 3] ,in, η is the normalization factor. It is a predicted probability of bodily injury based on passenger injuries measured through simulation. It is the prior probability of human injury. It refers to the actual human injury event that occurs at the current time point (t), and It is a predicted human injury event that occurs at the current time point (t).

7. The apparatus according to claim 1, wherein, The airbag deployment determiner is configured to determine whether to deploy the airbag based on the posterior probability of human injury exceeding a preset reference value.

8. The apparatus according to claim 1, wherein, The vehicle motion information includes the vehicle's acceleration and angular velocity values, collision values, pressure values, roll values, pitch angle values, and yaw angle values.

9. A method for controlling a vehicle airbag, the method comprising: The human injury probability calculator calculates the conditional probability and predicted probability of human injury based on vehicle motion information measured by sensing devices. The learner calculates the posterior probability of human injury by performing probability-based real-time feedback machine learning, based on the conditional probability of human injury and the predicted probability of human injury; and The airbag deployment determiner determines whether to deploy the airbag based on the posterior probability of human injury.

10. The method of claim 9, further comprising: The learner calculates the prior probability of human injury based on the previously obtained conditional probability of human injury and the previously obtained predicted probability of human injury. as well as The learner calculates the posterior probability of human injury by multiplying the currently calculated conditional probability of human injury by the prior probability of human injury.

11. The method according to claim 10, wherein, The probability-based real-time feedback machine learning is configured to update the prior probability of human injury by feeding back the current posterior probability of human injury to the previous prior probability of human injury.

12. The method according to claim 10, wherein, The conditional probability of bodily injury is configured to be calculated using Formula 1: [Formula 1] ,in, It is the predicted probability of human injury at the current time point (t) based on the measurement value of the sensing device and the actual probability of human injury at the previous time point (t-1). It is a predicted probability of bodily injury based on passenger injuries measured through simulation. It refers to the actual human injury event that occurs at the current time point (t). It is the measurement value of the sensing device at the current time point (t). It refers to the actual human injury event that occurred at the previous time point (t-1), and It is a predicted human injury event that occurs at the current time point (t).

13. The method according to claim 10, wherein, The prior probability of bodily injury is configured to be calculated using Formula 2: [Formula 2] ,in, It is based on the measured value of the human injury probability of the sensing device at the previous time point (t-1) and the predicted probability of human injury at the current time point (t). It is the posterior probability of human injury at the previous time point (t-1). It refers to the actual human injury event that occurs at the current time point (t). It is the measurement value of the sensing device at the current time point (t), and It refers to the actual human injury event that occurred at the previous time point (t-1).

14. The method of claim 10, wherein, The posterior probability of human injury is configured to be calculated using Formula 3: [Formula 3] ,in, η is the normalization factor. It is a predicted probability of bodily injury based on passenger injuries measured through simulation, and It is the prior probability of human injury. It refers to the actual human injury event that occurs at the current time point (t), and It is a predicted human injury event that occurs at the current time point (t).

15. The method according to claim 9, wherein, The airbag deployment determiner is configured to determine whether to deploy the airbag based on the posterior probability of human injury exceeding a preset reference value.

16. The method according to claim 9, wherein, The vehicle motion information includes the vehicle's acceleration and angular velocity values, collision values, pressure values, roll values, pitch angle values, and yaw angle values.

17. An apparatus for controlling a vehicle airbag, the apparatus comprising: The human injury probability calculator is configured to calculate the conditional probability of human injury and the predicted probability of human injury based on vehicle motion information measured by a sensing device. The learner is configured to The prior probability of human injury is calculated based on the conditional probability of human injury and the predicted probability of human injury. The posterior probability of human injury is calculated by multiplying the conditional probability of human injury by the prior probability of human injury. The prior probability of human injury is updated by feeding back the current posterior probability of human injury to the previous prior probability of human injury through probability-based real-time feedback machine learning; and An airbag deployment determiner is configured to determine whether to deploy the airbag based on the posterior probability of human injury.

18. A method for controlling a vehicle airbag, the method comprising: The human injury probability calculator calculates the conditional probability and predicted probability of human injury based on vehicle motion information measured by sensing devices. The learner calculates the prior probability of human injury based on the conditional probability of human injury and the predicted probability of human injury. The learner calculates the posterior probability of human injury by multiplying the conditional probability of human injury by the prior probability of human injury. The learner updates the prior probability of human injury by feeding back the current posterior probability of human injury to the previous prior probability of human injury through probability-based real-time feedback machine learning; and The airbag deployment determiner determines whether to deploy the airbag based on the posterior probability of human injury.

Images