Emergency auxiliary take-over system and method for pilotless automobile
Through the emergency auxiliary takeover system, combined with collision detection and emergency control modules, the driver's takeover status is predicted and the vehicle is automatically controlled to avoid collision, which solves the safety issues of unmanned driving systems in emergency situations and realizes efficient collision risk management.
Patent Information
- Application Number
- CN202410417857.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-09
- Publication Date
- 2025-10-14
AI Technical Summary
In the existing unmanned driving system, during emergency steering and obstacle avoidance conditions, the collision detection device is not linked to the unmanned driving system, resulting in a high probability of misjudgment, inability to effectively take over control, and a collision risk.
An emergency auxiliary takeover system is adopted, including a controller, a collision detection module and an emergency control module. It obtains vehicle and obstacle information, analyzes the driver's status, uses extreme learning machines and neuromuscular models to model driver behavior, combines steering wheel torque detection, predicts the takeover status, and automatically controls the vehicle to avoid collision when the driver is unable to take over in time.
It improves the accuracy of collision risk detection and the safety of unmanned vehicles, ensuring that vehicles follow a safe trajectory in emergency situations to avoid collisions.
Smart Images

Figure CN120773768A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of unmanned driving, in particular to an emergency auxiliary takeover system and method for an unmanned vehicle. BACKGROUND
[0002] In recent years, the unmanned driving technology has developed rapidly. In most time, the unmanned vehicle does not need the intervention of the driver. When the unmanned vehicle encounters an emergency turning and obstacle avoidance working condition, the driver needs to take over the unmanned vehicle for emergency turning to avoid collision with the front vehicle or pedestrian. Therefore, in order to improve the performance of the human-machine cooperative turning control of the unmanned takeover system in the emergency turning and obstacle avoidance working condition, it is very important to identify and understand the lateral behavior of the driver.
[0003] Unmanned driving is a complex system, including sensor fusion, positioning, perception, planning, control and other modules. If a problem occurs in each module, it may cause a collision risk. In the prior art, a separate collision detection device is used to determine whether there is a collision risk according to the distance between the current vehicle and the front vehicle. The determination is not linked with the unmanned system, and the probability of misjudgment is high. Moreover, the collision detection device only detects and does not take over the control of the vehicle, which may cause a collision risk in the actual road conditions, and makes the unmanned vehicle in a dangerous situation.
[0004] Therefore, how to predict the takeover state of the driver and how to ensure the driving safety of the unmanned vehicle when the driver cannot take over the vehicle in time become urgent problems to be solved. SUMMARY
[0005] The present application provides an emergency auxiliary takeover system and method for an unmanned vehicle, which solves the problem that the collision detection device is not linked with the unmanned system and only detects without taking over the control of the vehicle, which leads to a collision risk.
[0006] To achieve the above object, the present application adopts the following technical scheme:
[0007] An emergency auxiliary takeover system and method for an unmanned vehicle, the emergency auxiliary takeover system comprising:
[0008] A controller arranged on the unmanned vehicle and connected with a vehicle-mounted intelligent terminal, used to acquire the information and state of the current target vehicle and the obstacle on the driving path;
[0009] A collision detection module electrically connected with the controller, used to detect whether the current unmanned target vehicle will collide with the obstacle on the driving path, and also used to detect the state of the current driver;
[0010] An emergency control module is electrically connected with the controller and the collision detection device, and the controller can control the target vehicle in emergency;
[0011] The emergency auxiliary takeover method comprises the following steps:
[0012] S1, obtaining information of a target vehicle currently in unmanned driving and obstacle information on a driving path;
[0013] S2, detecting whether the current driving path will collide with the obstacle on the path;
[0014] S3, obtaining driver information of the current vehicle;
[0015] S4, inputting the obtained driver information into a pre-trained takeover state prediction model to obtain a takeover state of the driver;
[0016] S5, obtaining a danger level of the current vehicle, and predicting whether the driver can take over the current vehicle in time according to the takeover state of the current driver and the danger level;
[0017] S6, when it is judged that the driver cannot take over the vehicle in time, a corresponding vehicle emergency control module is specified according to the danger level and the takeover state to control the vehicle.
[0018] The emergency auxiliary takeover system and method in the application monitor the state of the unmanned vehicle and the information of the obstacle on the driving path, and the state of the internal driver, when a sudden situation occurs on the driving path of the unmanned vehicle, analyze whether the vehicle can be automatically controlled to avoid collision, analyze whether the information of the driver in the vehicle meets the takeover requirement, when it is detected that the information of the driver in the vehicle can be taken over, send a reminder information to the driver for takeover, and when it is detected that the driver in the vehicle cannot take over in time, the vehicle automatically avoids, so as to ensure the driving safety of the vehicle.
[0019] Further, in step S1, the information of the target vehicle at least includes one or more of the driving state, the driving path, the driving speed, the brake time and the safety distance of the target vehicle. By detecting various information of the target vehicle, the safe driving of the vehicle is ensured, and when an abnormal situation occurs on the driving road, the vehicle can brake or avoid in time, so as to ensure the safe driving of the vehicle on the road.
[0020] Further, in step S1, the obstacle information on the driving path includes one or more of obstacle tracking, obstacle path, obstacle speed, obstacle path prediction, obstacle confidence, and distance between the target vehicle and the obstacle. By analyzing the obstacles on the driving path and tracking the obstacles while predicting the movement of the obstacles, the driving habits of experienced drivers are simulated to ensure that the obstacles can be avoided in the event of an abnormal situation.
[0021] Further, in step S2, the collision detection module is used to detect whether the current driving path will collide with the obstacles on the path. The collision detection module is used to determine whether the current driving path has a collision risk and to predict a possible second path while determining whether the current autonomous vehicle turning onto the second path has a collision risk. By analyzing the collision of the vehicle through the collision detection module, the safe driving of the vehicle is ensured, and the second path that can be avoided is predicted in advance to ensure that it can be avoided in time in the event of an accident.
[0022] Further, the collision detection module includes three main units: vehicle analysis, obstacle analysis, and path analysis. The functions of the vehicle analysis unit include vehicle state estimation and brake time and distance estimation. The functions of the obstacle analysis unit include obstacle filtering and obstacle tracking. The functions of the path analysis unit include path analysis.
[0023] After vehicle analysis, obstacle analysis, and path analysis, the collision risk of the autonomous vehicle is finally evaluated by calculating the safe distance of the autonomous vehicle, the distance of the obstacle, and the speed of the obstacle, and the corresponding collision detection strategy analysis is performed.
[0024] The collision detection module is divided into three units: the vehicle analysis unit estimates the driving state of the current autonomous vehicle, the obstacle analysis unit predicts the driving state and possible driving path of the obstacle, and the path analysis unit plans the avoidance path.
[0025] Further, the first path is the driving path of the current autonomous vehicle, and the second path is the predicted safe path. There are mainly two paths in the system: one is the driving path of the current autonomous vehicle, and the other is the predicted emergency avoidance path.
[0026] Further, in step S3, the driver information includes at least image information, voice information, and action information. By detecting the facial information of the driver, whether he is talking to someone, and whether his hands and feet meet the conditions for driving the vehicle, it can be determined whether the driver meets the conditions for taking over the vehicle in time.
[0027] Further, the self-driving vehicle also detects the steering wheel torque of the vehicle, judges whether the torque value information meets the corresponding hands-off driving condition according to the torque value information, and sends a corresponding control signal according to the judgment result. The step comprises:
[0028] Receiving the steering wheel torque value information detected by the steering wheel torque sensor;
[0029] According to the first judgment module, whether the steering wheel torque value is less than the threshold N is judged;
[0030] According to the second judgment module, whether the duration is greater than the threshold M is judged;
[0031] If the torque value information meets the preset condition, a corresponding reminder control signal is sent;
[0032] According to the reminder signal, whether the corresponding automatic takeover condition is met is judged, and the subsequent control step is performed according to the judgment result.
[0033] The system detects the steering wheel torque of the vehicle, that is, detects the hand of the driver, detects whether the hand is on the steering wheel, and whether the steering wheel can be timely taken over for steering, avoiding and other operations when an emergency occurs.
[0034] Further, the emergency auxiliary takeover method further comprises the step of pre-training the takeover state prediction model, which specifically comprises:
[0035] Obtain a plurality of preset historical sample data, each of which comprises face image sample data, voice sample data and human body action sample data of a test person;
[0036] Obtain the driver takeover response time corresponding to each historical sample data, and determine the corresponding label information according to the driver takeover response time;
[0037] According to the historical sample data and the corresponding label information, a training data set is constructed;
[0038] The training data set is input to a pre-constructed convolutional neural network for training to obtain a trained takeover state prediction model;
[0039] The label information includes at least one of the fast response state, the normal response state and the abnormal response state.
[0040] The big data training of the takeover state prediction model can judge the label information according to the information of the current driver, and control the vehicle according to the label information to ensure the safe driving of the vehicle.
[0041] Further, in step S6, the vehicle emergency control module comprises one or more of a safety risk analysis, a vehicle trajectory point generation, a vehicle speed control, a vehicle steering control and a vehicle gear control. The plurality of vehicle emergency control modules control the vehicle, ensuring that in an emergency, the vehicle can quickly react to avoid, brake and the like.
[0042] Compared with the prior art, the beneficial effects of the present application are that the above technical scheme of the present application introduces a collision detection module and a collision avoidance module, and makes the collision detection module and the collision avoidance module interact with the driving system of the vehicle, thereby increasing the accuracy of collision risk detection. In the case of judging that there is a collision risk, the collision avoidance module can take over the unmanned driving system, ensuring that the unmanned vehicle travels along a safe trajectory and avoids collision risks.
[0043] At the same time, the present application also sets a torque sensor on the steering wheel, which monitors the torque value applied by the driver on the steering wheel at all times to judge the hand movement of the driver, so as to judge whether the driver can take over the vehicle in time. If the driver still does not respond, the emergency control module will take over the unmanned vehicle, ensuring the safe driving of the unmanned vehicle. BRIEF DESCRIPTION OF DRAWINGS
[0044] Figure 1 A step flow chart of an unmanned vehicle emergency auxiliary takeover system and method provided by the present application;
[0045] Figure 2 A composition logic diagram of a collision detection module in an unmanned vehicle emergency auxiliary takeover system and method provided by the present application;
[0046] Figure 3 A flowchart of steering wheel torque judgment in an unmanned vehicle emergency auxiliary takeover system and method provided by the present application;
[0047] Figure 4 A composition logic diagram of an emergency control module in an unmanned vehicle emergency auxiliary takeover system and method provided by the present application;
[0048] Figure 5 A structure schematic diagram of a terminal device in an unmanned vehicle emergency auxiliary takeover system and method provided by the present application.
[0049] In the figure: processor 100, memory 200. DETAILED DESCRIPTION
[0050] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.
[0051] In the description of the present application, it needs to be understood that the terms "center", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.
[0052] The terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.
[0053] In the description of the present application, it needs to be noted that, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, or the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances. In addition, when describing pipelines or channels, "connected" and "connected" in the present application have the meaning of conducting. The specific meaning needs to be understood in combination with the context.
[0054] In the embodiments of the present application, the words "exemplary" or "for example" are used to mean serving as an example, instance, or illustration. Any embodiment or design presented as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Rather, the use of "exemplary" or "for example" is intended to present concepts in a concrete manner. The term "one embodiment" or "an embodiment" means that at least one embodiment of the present application has been described.
[0055] In recent years, the unmanned driving technology has developed rapidly. Unmanned driving is a complex system, including sensor fusion, positioning, perception, planning, control and other modules. If a problem occurs in each module, it may cause a collision risk. In the prior art, a separate collision detection device is used to determine whether there is a collision risk according to the distance between the current vehicle and the front vehicle. The determination is not linked with the unmanned driving system, and the misjudgment rate is high. Moreover, the collision detection device only detects and does not take over the control of the vehicle, which may cause a collision risk in the actual road conditions, so that the unmanned vehicle is in a dangerous situation.
[0056] Therefore, how to predict the takeover state of the driver and how to ensure the driving safety of the unmanned vehicle when the driver cannot take over the vehicle in time become problems to be solved.
[0057] Therefore, the present application provides an emergency auxiliary takeover system and method for an unmanned vehicle, which ensures that the unmanned system can be taken over by a collision avoidance module in the case of judging a collision risk, and ensures that the unmanned vehicle travels along a safe trajectory to avoid the collision risk.
[0058] In order to further illustrate the beneficial effects of the present application, the specific embodiments, methods and steps of the present application are described below in combination with the drawings and preferred embodiments.
[0059] As shown in Figure 1 An emergency auxiliary takeover system and method for an unmanned vehicle, the emergency auxiliary takeover system comprising:
[0060] A controller is arranged on the unmanned vehicle and connected with the vehicle-mounted intelligent terminal, which is used to acquire the information and state of the current target vehicle and the obstacles on the driving path;
[0061] A collision detection module is electrically connected with the controller, which is used to detect whether the current unmanned target vehicle will collide with the obstacles on the driving path, and also can detect the state of the current driver;
[0062] An emergency control module is electrically connected with the controller and the collision detection module, which can control the target vehicle through the controller.
[0063] The controller is connected with the vehicle-mounted intelligent terminal, and the current vehicle information and obstacle information on the driving path can be transmitted to the vehicle-mounted intelligent terminal, which collects the information and controls the unmanned vehicle according to the actual situation, so that the unmanned vehicle can normally travel, change lanes, etc.
[0064] The collision detection module effectively identifies the collision of the vehicle to avoid the collision danger, and the emergency control module controls the vehicle according to the result of the collision detection module,
[0065] The emergency auxiliary takeover method comprises the following steps:
[0066] Step S1, acquiring the information of the current unmanned target vehicle and the information of the obstacles on the driving path;
[0067] Specifically, in order to ensure the takeover control of the driver to the unmanned vehicle in an emergency, the unmanned vehicle does not need the intervention of the driver most of the time, when the unmanned vehicle encounters an emergency turning avoidance situation, the driver needs to take over the unmanned vehicle for emergency turning, avoidance and the like to avoid collision with the front vehicle, pedestrian or obstacle. Therefore, in order to improve the performance of the unmanned takeover system in the process of emergency human-machine cooperative turning control, it is very important to identify and understand the behavior and state of the driver.
[0068] The behavior and state identification usually adopts a two-point preview model, but it is difficult to determine the appropriate near point and far point in the process of emergency turning avoidance. As a machine learning algorithm, the extreme learning machine is gradually applied to the construction of the driver model with high nonlinearity through data training and fitting, and can describe the preview process of the driver in the process of emergency turning avoidance, but the training of the extreme learning machine is similar to the method of a black box, and the internal mechanism of the driver behavior cannot be effectively identified.
[0069] Therefore, the embodiment of the present application provides an unmanned vehicle emergency auxiliary takeover system and method, which specifically combines the extreme learning machine and the neuromuscular model to model the behavior and state of the driver of the unmanned takeover system, wherein the extreme learning machine is used to accurately describe the driver preview model, and the neuromuscular model is used to describe the interaction behavior between the driver neuromuscular system and the vehicle control system.
[0070] Among them, the controller is installed on the unmanned vehicle, and the emergency auxiliary takeover system is installed on the controller. When the unmanned vehicle encounters an emergency, the emergency auxiliary takeover system acquires the information of the target vehicle and the obstacle information on the driving path.
[0071] The information of the target vehicle at least includes one or more of the driving state, the driving path, the driving speed, the brake time and the safety distance of the vehicle.
[0072] Further, the information of the target vehicle further includes one or more of the vehicle mass center side slip angle, the vehicle yaw angular velocity and the vehicle longitudinal speed.
[0073] The obstacle information at least includes one or more of the obstacle tracking, the obstacle path, the obstacle speed, the obstacle path prediction, the obstacle confidence and the distance between the target vehicle and the obstacle.
[0074] Step S2, detecting whether the current driving path will collide with the obstacle on the path.
[0075] Specifically, the collision detection module is used to detect whether the current driving path collides with the obstacles on the path, and the collision detection module is used to determine whether the current driving path has a collision risk, and the collision detection module predicts a possible second path while determining whether the current autonomous vehicle turning onto the second path has a collision risk.
[0076] Further, as Figure 2 shown, the collision detection module includes three units of vehicle analysis, obstacle analysis and path analysis, wherein the functions of the vehicle analysis unit mainly include vehicle state estimation and brake time and distance estimation, detecting the driving state, driving posture and driving speed of the current vehicle, and predicting the time and distance required for braking when encountering an emergency to ensure the safety of vehicle driving in an emergency.
[0077] The functions of the obstacle analysis unit mainly include obstacle filtering and obstacle tracking, filtering the obstacles around the vehicle, identifying the obstacles that may affect the driving of the vehicle, and tracking these obstacles that may affect the driving of the vehicle to detect the driving path and speed of the obstacles at any time, and to avoid obstacles in an emergency to ensure the safety of vehicle driving in an emergency.
[0078] The path analysis unit mainly analyzes the path, including the first path of the current vehicle driving and the second path for avoiding in an emergency, to ensure that the vehicle can enter the second path through steering operation when the vehicle cannot avoid obstacles by braking, so as to avoid the obstacles that may collide with the current vehicle and ensure the safety of the vehicle.
[0079] In the collision detection module, the vehicle analysis module, the obstacle analysis module and the path analysis module work together to calculate the safety distance of the autonomous vehicle, the distance of the obstacle and the speed of the obstacle, finally to evaluate the collision risk of the autonomous vehicle and analyze the corresponding collision detection strategy.
[0080] Step S3, obtaining the driving personnel information of the current vehicle.
[0081] The driving personnel information at least includes image information, voice information and action information. Specifically, the image information is the face information of the driving personnel, which can be monitored by setting a camera inside the vehicle to monitor the face information of the driving personnel.
[0082] The voice information of the driver can be detected by setting a voice receiving device, such as a microphone or a decibel tester, inside the vehicle to detect the intensity of the sound in the vehicle to determine whether the current driving personnel is talking to someone and whether the current driving personnel can take over the vehicle.
[0083] The action information of the driver is monitored and recognized by a camera arranged at the pedal to monitor the foot action of the driver, and is matched with image information and voice information, so that the state of the driver is analyzed through multiple information to determine whether the driver can timely take over the vehicle to perform steering, avoidance, braking and other operations in an emergency.
[0084] Further, as shown in Figure 3 The self-driving vehicle further detects the steering wheel torque of the vehicle, determines whether the torque value information meets the corresponding hand-off driving condition, and sends a corresponding control signal according to the determination result.
[0085] Receiving the steering wheel torque value information detected by the steering wheel torque sensor;
[0086] The first determination module is used to determine whether the steering wheel torque value is less than a threshold N;
[0087] The second determination module is used to determine whether the duration is greater than a threshold M;
[0088] If the torque value information meets the preset condition, a corresponding reminder control signal is sent;
[0089] According to the reminder signal, it is determined whether the automatic takeover condition is met, and the subsequent control step is performed according to the determination result.
[0090] The steering wheel torque sensor continuously detects the torque value applied by the driver on the steering wheel, and a preset condition for meeting the hand-off driving state is preset, for example, when the steering wheel torque value is detected to be less than a set threshold N and the duration M, the system determines that the driver's hand is not on the steering wheel, and then sends a corresponding reminder information.
[0091] In this embodiment, the reminder information reminds the driver to hold the steering wheel through instrument sound, text, picture and seat belt pretensioning.
[0092] The system detects the steering wheel torque of the vehicle, that is, detects the hand of the driver, detects whether the hand is on the steering wheel, and whether the hand can timely take over the steering wheel to perform steering, avoidance and other operations when an emergency occurs.
[0093] Step S4, input the obtained driver information into a pre-trained takeover state prediction model to obtain the takeover state of the driver.
[0094] In the judgment process, the collision detection module detects the image information, voice information and action information of the driver, and cooperates with the torque detection on the steering wheel to ensure the accuracy of the state recognition of the driver. When the driver cannot take over the vehicle in time, the system can take over the vehicle to turn and avoid in an emergency to ensure the safety of the current unmanned vehicle.
[0095] After the system sends a reminder signal, if the driver does not respond, the system enters an automatic takeover mode, and the emergency control module takes over the control of the vehicle. At the same time, after entering the automatic takeover mode, the system also controls the vehicle to keep driving in the original lane; the front camera and the rear radar monitor the vehicle driving condition, so that the vehicle slows down and brakes, and the electronic hand brake is started to slow down and safely park; the hazard warning light is lit to remind the rear vehicle.
[0096] In this embodiment, after the instrument reminds the driver by sound and text for 2s, the steering torque is continuously monitored. If the driver still does not actively take over the steering wheel, the system will actively take over the vehicle. Through the front camera detection information, the vehicle actively controls the vehicle to keep in the lane to avoid the vehicle deviating from the lane. At the same time, through the rear radar monitoring whether there is a vehicle approaching behind the vehicle, when confirming that there is no approaching vehicle behind the vehicle, the system actively slows down and brakes through the electronic stability control system, so that the vehicle slows down and safely parks, and the hazard warning light is lit and the electronic hand brake EPB is started.
[0097] Further, the emergency auxiliary takeover method in the application further includes the step of pre-training a takeover state prediction model, specifically including:
[0098] A plurality of preset historical sample data are obtained, each historical sample data including a face image sample data, a voice sample data and a human body action sample data of a test person;
[0099] The driver takeover response time corresponding to each historical sample data is obtained, and the corresponding label information is determined according to the driver takeover response time;
[0100] The training data set is constructed according to the historical sample data and the corresponding label information;
[0101] The training data set is input into a pre-constructed convolutional neural network for training to obtain a trained takeover state prediction model.
[0102] The label information includes at least one of a fast response state, a normal response state and an abnormal response state.
[0103] Specifically, in constructing the training data set, a plurality of historical sample data covering a plurality of different takeover states are obtained, and label information corresponding to each historical sample data is determined based on the actual driver takeover response time, the label information being used to indicate the driver takeover state corresponding to each historical sample data, including a fast response state, a normal response state, and an abnormal response state. The training data set can be generated according to the historical sample data and the corresponding label information.
[0104] In some optional embodiments, when the actual driver takeover response time is within 3 seconds, the corresponding label information can be determined as the fast response state; when the actual driver takeover response time is within 5 seconds but more than 3 seconds, the corresponding label information can be determined as the normal response state; and when the actual driver takeover response time is more than 5 seconds, the corresponding label information can be determined as the abnormal response state.
[0105] Further as an optional implementation, the training data set is input to a pre-constructed convolutional neural network for training to obtain a trained takeover state prediction model, and the step specifically includes:
[0106] The training data set is input to the convolutional neural network, and a takeover state prediction result is output;
[0107] The loss value of the convolutional neural network is determined according to the takeover state prediction result and the label information;
[0108] The model parameters of the convolutional neural network are updated by a back propagation algorithm according to the loss value, and the step of inputting the training data set to the convolutional neural network is returned;
[0109] When the loss value reaches a preset first threshold or the number of iterations reaches a preset second threshold, the training is stopped, and the trained takeover state prediction model is obtained.
[0110] Specifically, after the data in the training data set is input to the initialized convolutional neural network model, a recognition result output by the model, i.e., a takeover state prediction result, can be obtained, and the accuracy of the model prediction can be evaluated according to the takeover state prediction result and the aforementioned label information, so that the parameters of the model are updated.
[0111] For the takeover state prediction model, the accuracy of the model prediction result can be measured by a loss function (Loss Function), which is defined on a single training data and is used to measure the prediction error of a training data. Specifically, the loss value of a single training data is determined by the label of the training data and the prediction result of the model for the training data.
[0112] In actual training, a training data set has many training data, so a cost function is generally used to measure the overall error of the training data set, the cost function is defined on the entire training data set, and is used to calculate the average of the prediction error of all training data, so that the prediction effect of the model can be better measured.
[0113] For a general machine learning model, based on the foregoing cost function, a regularization term that measures the complexity of the model can be used as the target function for training, and based on the target function, the loss value of the entire training data set can be obtained.
[0114] There are many commonly used loss functions, such as 01 loss function, square loss function, absolute loss function, logarithmic loss function, cross-entropy loss function, etc., which can be used as the loss function of the machine learning model, and will not be described one by one here.
[0115] In the embodiment of the application, the loss value of the training can be determined from any one of the loss functions. Based on the loss value of the training, the parameters of the model are updated using the back propagation algorithm, and after several iterations, a trained takeover state prediction model can be obtained. The specific number of iterations can be preset, or the training can be considered complete when the test set reaches the accuracy requirement.
[0116] Step S5, according to the collision detection module, the danger level of the current vehicle is obtained, and according to the takeover state of the current driver and the danger level predicted by the takeover state prediction model, whether the driver can take over the current vehicle in time is predicted.
[0117] Further as an optional implementation, the danger scene level of the target vehicle is obtained, and according to the current takeover state and the danger scene level, whether the driver can take over the target vehicle in time is predicted, which specifically includes:
[0118] The predicted collision time of the target vehicle is determined, and the danger scene level of the target vehicle is determined according to the predicted collision time, the danger scene level being a general danger scene, an emergency danger scene or a non-danger scene;
[0119] When the danger scene level is an emergency danger scene, if the current takeover state is a fast response state, it is determined that the driver can take over the target vehicle in time, and if the current takeover state is a normal response state or an abnormal response state, it is determined that the driver cannot take over the target vehicle in time;
[0120] When the danger scene level is a general danger scene, if the current takeover state is a fast response state or a normal response state, it is determined that the driver can take over the target vehicle in time, and if the current takeover state is an abnormal response state, it is determined that the driver cannot take over the target vehicle in time;
[0121] When the dangerous scene level is a non-dangerous scene, it is determined that the driver can take over the target vehicle in time.
[0122] Further, as an optional embodiment, the step of determining the target vehicle's predicted time to collision (TTC) and determining the target vehicle's dangerous scene level according to the predicted time to collision (TTC) specifically comprises:
[0123] The real-time distance between the target vehicle and the preceding vehicle is obtained, and the relative motion speed between the target vehicle and the preceding vehicle is determined.
[0124] The target vehicle's predicted time to collision (TTC) is calculated according to the real-time distance and the relative motion speed.
[0125] When the predicted time to collision (TTC) is less than or equal to a preset third threshold value, it is determined that the dangerous scene level is an emergency dangerous scene.
[0126] When the predicted time to collision (TTC) is greater than the third threshold value and less than or equal to a preset fourth threshold value, it is determined that the dangerous scene level is a general dangerous scene.
[0127] When the predicted time to collision (TTC) is greater than the fourth threshold value, it is determined that the dangerous scene level is a non-dangerous scene.
[0128] Specifically, according to the target vehicle's predicted time to collision (TTC), the target vehicle's dangerous scene level is determined. When the predicted time to collision (TTC) is less than or equal to a third threshold value (such as 3 seconds), the corresponding dangerous scene level is an emergency dangerous scene. When the predicted time to collision (TTC) is greater than the third threshold value and less than or equal to a fourth threshold value (such as 10 seconds), the corresponding dangerous scene level is a general dangerous scene. When the predicted time to collision (TTC) is greater than the fourth threshold value, the unmanned system can avoid collision by itself, and the driver does not need to take over, and the corresponding dangerous scene level is a non-dangerous scene.
[0129] After determining the dangerous scene level, it is judged whether the driver can take over the target vehicle in time according to the dangerous scene level and the current takeover state of the driver. When the dangerous scene level is an emergency dangerous scene, if the current takeover state is an ordinary response state or an abnormal response state, it is determined that the driver cannot take over the target vehicle in time. When the dangerous scene level is a general dangerous scene, if the current takeover state is an abnormal response state, it is determined that the driver cannot take over the target vehicle in time. In other cases, it is determined that the driver can take over the target vehicle in time.
[0130] Step S6, when it is judged that the driver cannot take over the vehicle in time, the corresponding vehicle emergency control module is formulated according to the dangerous level and the takeover state to control the vehicle.
[0131] Specifically, when it is determined that the driver cannot take over the target vehicle in time, an emergency auxiliary takeover decision is determined and executed through the target vehicle to avoid collision accidents of the unmanned vehicle.
[0132] Further, as an optional implementation, the corresponding emergency auxiliary takeover decision is determined according to the dangerous scene level and the current takeover state, and then the target vehicle executes the emergency auxiliary takeover decision, which specifically includes:
[0133] A preset emergency auxiliary takeover decision library is acquired, and the dangerous scene level and the current takeover state are matched in the emergency auxiliary takeover decision library to obtain the corresponding emergency auxiliary takeover decision;
[0134] The emergency auxiliary takeover decision is sent to the unmanned driving system of the target vehicle, so that the unmanned driving system executes the emergency auxiliary takeover decision.
[0135] Specifically, the embodiment of the application is preconfigured with an emergency auxiliary takeover decision library, the emergency auxiliary takeover decision library is provided with a plurality of emergency auxiliary takeover decisions, and each emergency auxiliary takeover decision is provided with a corresponding "dangerous scene level current takeover state" array. The dangerous scene level and the current takeover state determined according to the foregoing steps can be matched in the emergency auxiliary takeover decision library to obtain the corresponding emergency auxiliary takeover decision. The target vehicle is controlled by the unmanned driving system of the target vehicle to execute the emergency auxiliary takeover decision, so that the target vehicle can be controlled by the auxiliary takeover when the driver cannot take over the vehicle in time, and the driving safety of the unmanned vehicle is improved.
[0136] The method steps of the embodiment of the application are described above. It can be recognized that the embodiment of the application predicts the takeover state of the driver based on the three-dimensional data of the face image, voice and body action of the driver, can accurately evaluate the current takeover ability of the driver in response to the sudden situation, can accurately judge whether the driver can take over the unmanned vehicle in time in combination with the current dangerous scene, and can match the corresponding emergency auxiliary takeover decision for the unmanned vehicle to execute when the driver cannot take over the unmanned vehicle, thereby improving the driving safety of the unmanned vehicle.
[0137] Further, as shown in Figure 4 The vehicle emergency control module includes one or more of safety risk analysis, vehicle trajectory point generation, vehicle speed control, vehicle steering control and vehicle gear control.
[0138] The emergency control module controls the vehicle through multiple systems, and when an emergency occurs, the vehicle can be controlled to avoid the emergency by imitating the operation of the driver, so as to ensure the safe driving of the vehicle.
[0139] As shown in Figure 5 The embodiment of the application provides an unmanned vehicle emergency auxiliary takeover device, which comprises at least one processor 100; at least one memory 200 for storing at least one program;
[0140] When the at least one program is executed by the at least one processor 100, the at least one processor 100 implements the unmanned vehicle emergency assistance takeover method.
[0141] Although the present application has been described in connection with certain specific embodiments thereof, it will be understood that it is capable of further modifications and this application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is to be understood that the phraseology or terminology herein is for the purpose of description and not of limitation, such that the terminology or phraseology of the present disclosure is to be interpreted by those skilled in relevant art in light of the teachings and
[0142] Although the present application has been described in connection with certain specific embodiments thereof, it will be understood that it is capable of further modifications and this application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is to be understood that the phraseology or terminology herein is for the purpose of description and not of limitation, such that the terminology or phraseology of the present disclosure is to be interpreted by those skilled in relevant art in light of the teachings and
[0143] The above description is merely that of a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims
1. An unmanned vehicle emergency auxiliary takeover system and method, characterized in that: The emergency auxiliary takeover system includes: The controller is installed on the unmanned vehicle and connected to the onboard intelligent terminal to obtain information and status of the current target vehicle and obstacles on the driving path; A collision detection module, electrically connected to the controller, is used to detect whether the currently unmanned target vehicle will collide with an obstacle on the driving path, and can also detect the current driver's status; an emergency control module, electrically connected to the controller and the collision detection module, capable of performing emergency control on the target vehicle through the controller; The emergency auxiliary takeover method comprises the following steps: S1, obtain the information of the current unmanned target vehicle and the obstacle information on the driving path; S2, detects whether the current driving path will collide with obstacles on the path; S3, obtaining the driver information of the current vehicle; S4, inputting the acquired driver information into a pre-trained takeover state prediction model to obtain the driver's takeover state; S5, obtaining the danger level of the current vehicle, and predicting whether the driver can take over the current vehicle in time based on the current driver's takeover status and the danger level; S6, when it is determined that the driver cannot take over the vehicle in time, designate a corresponding vehicle emergency control module to control the vehicle according to the danger level and the takeover status.
2. The unmanned vehicle emergency auxiliary takeover system and method according to claim 1, characterized in that: In step S1, the target vehicle information includes at least one or more of the target vehicle's driving state, driving path, driving speed, braking time, and safety distance.
3. The unmanned vehicle emergency auxiliary takeover system and method according to claim 2, characterized in that: In step S1 , the obstacle information on the driving path includes at least one or more of obstacle tracking, obstacle path, obstacle speed, obstacle path prediction, obstacle confidence, and the distance between the target vehicle and the obstacle.
4. The unmanned vehicle emergency auxiliary takeover system and method according to claim 1, characterized in that: In step S2, a collision detection module is used to detect whether the current driving path will collide with an obstacle on the path. The collision detection module is used to determine whether there is a collision risk on the current driving path, predict a possible second path, and determine whether there is a collision risk when the current unmanned vehicle turns to the second path.
5. The unmanned vehicle emergency auxiliary takeover system and method according to claim 4, characterized in that: The collision detection module includes three units: vehicle analysis, obstacle analysis, and path analysis. The functions of the vehicle analysis unit include vehicle state estimation and braking time and braking distance estimation. The functions of the obstacle analysis unit include obstacle filtering and obstacle tracking. The function of the path analysis unit is to analyze the path. After the vehicle analysis, the obstacle analysis and the path analysis, the safety distance of the unmanned vehicle, the distance to the obstacle and the speed of the obstacle are calculated, and finally the collision risk of the unmanned vehicle is assessed, and a corresponding collision detection strategy analysis is performed.
6. The unmanned vehicle emergency auxiliary takeover system and method according to claim 4, characterized in that: The first path is the current driving path of the driverless car, and the second path is the predicted safe path.
7. The unmanned vehicle emergency auxiliary takeover system and method according to claim 1, characterized in that: In step S3, the driver information includes at least image information, voice information and action information.
8. The unmanned vehicle emergency auxiliary takeover system and method according to claim 7, characterized in that: The driverless car is further provided with a steering wheel torque detection module to detect the vehicle's steering wheel torque, and the steps of determining whether the corresponding hands-free driving condition is met based on the torque value information and issuing a corresponding control signal based on the determination result include: receiving steering wheel torque value information detected by the steering wheel torque sensor; A first judgment module is used to judge whether the steering wheel torque value is less than a threshold N; The second judgment module is used to judge whether the duration is greater than a threshold M; Determining that the torque value information meets a preset condition and issuing a corresponding reminder control signal; It is determined whether the corresponding automatic takeover conditions are met according to the reminder signal, and subsequent control steps are performed according to the determination result.
9. The unmanned vehicle emergency auxiliary takeover system and method according to claim 1, characterized in that: The emergency auxiliary takeover method further includes the step of pre-training the takeover state prediction model, which specifically includes: Acquire a plurality of preset historical sample data, each of which includes facial image sample data, voice sample data, and human body movement sample data of a test person; Obtaining a driver takeover response time corresponding to each of the historical sample data, and determining corresponding label information according to the driver takeover response time; Constructing a training data set based on the historical sample data and the corresponding label information; Inputting the training data set into a pre-built convolutional neural network for training to obtain the trained takeover status prediction model; The tag information includes at least one of a quick response state, a normal response state, and an abnormal response state.
10. The unmanned vehicle emergency auxiliary takeover system and method according to claim 1, characterized in that: In step S6, the vehicle emergency control module includes: one or more of: safety risk analysis, vehicle trajectory point generation, vehicle speed control, vehicle steering control and vehicle gear control.
Citation Information
Cited By
Automatic driving takeover control method and device, intelligent driving system and vehicle
CN120963771A
