Method, device and equipment for controlling car window to ascend and descend and medium

Abstract

The invention provides a method, device and equipment for controlling a car window to ascend and descend and a medium. The method comprises the following steps: acquiring a reference signal, wherein the reference signal indicates an expected speed of a vehicle window corresponding to an input command; generating a target speed of the vehicle window based on the obtained reference signal by using a deep neural network; and controlling the car window to ascend or descend according to the target speed based on the target speed and a direction signal for indicating the car window to ascend or descend. According to the method provided by the invention, the non-linear problem existing in vehicle window lifting control can be well solved by utilizing the deep neural network, so that the vehicle window lifting can be accurately controlled based on the method provided by the invention.

Description

Technical Field

[0001] The present disclosure relates to the field of vehicle control, and more specifically, to a method, device, equipment and medium for controlling the lifting and lowering of a vehicle window. Background Art

[0002] System modeling is a bridge between the real world and the mathematical world. In theory, any system in the world can be described by a mathematical equation. For example, a simple resistor circuit system can be described by the mathematical equation V=I*R, where V is voltage, I is current, and R is resistance.

[0003] After system modeling, the real system in the real world will be transferred to a similar system in the computer (like the building system in Matlab Simulink) to experiment with its future results or design controllers to stabilize the system, which is also called the simulation process. Therefore, system modeling is a necessary condition for simulation. Ideal system modeling should be as close to reality as possible.

[0004] From the perspective of system modeling and simulation, linear problems (such as the simple resistor circuit system mentioned above) can be well modeled or simulated through modeling or simulation tools such as Matlab; however, nonlinear problems often occur in real systems in the real world (for example, systems that control the lifting of car windows), for example, nonlinear problems regarding controlling the lifting of car windows according to the actual target speed based on the expected speed of the car windows; for example, drone flight systems involve mechanical, aerodynamic and electrical systems, which include many nonlinear problems.

[0005] For the above nonlinear problems, the existing modeling or simulation methods usually linearize the mathematical relationship near a certain point, so as to perform modeling or simulation for the nonlinear problem through the linearized system. The above processing methods for nonlinear problems can usually only guarantee the correctness within the range near the certain point, and cannot truly reflect the real system in the real world, so that the results after modeling or simulation are applied to the real system in the real world. Deviations will be caused, resulting in inaccurate control of the real system in the real world, and even system loss of control due to certain nonlinear problems.

[0006] Specifically in the field of controlling the lifting of car windows, there is also the nonlinear problem as mentioned above. Currently, for this nonlinear problem, modeling or simulation is also performed through the above linearization method. Then the results of modeling or simulation are applied to the actual lifting control of car windows. The resulting problem is that the lifting control of car windows is inaccurate.

[0007] Therefore, a method is urgently needed to solve the problem of inaccurate window lifting control. Summary of the invention

[0008] In view of the above problems, the present disclosure provides a method for controlling the lifting of vehicle windows. The method provided by the present disclosure can utilize a deep neural network to well handle the nonlinear problem existing in the control of the lifting of vehicle windows, so that the lifting of vehicle windows can be accurately controlled based on the method provided by the present disclosure.

[0009] An embodiment of the present disclosure provides a method for controlling the lifting of a vehicle window, comprising: obtaining a reference signal, wherein the reference signal indicates an expected speed of the vehicle window corresponding to an input command; using a deep neural network to generate a target speed of the vehicle window based on the obtained reference signal; and controlling the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating the rising or falling of the vehicle window.

[0010] According to an embodiment of the present disclosure, the deep neural network includes a long short-term memory neural network, a first fully connected layer and a second fully connected layer, wherein the use of the deep neural network to generate a target speed of the vehicle window based on the acquired reference signal includes: using the long short-term memory neural network to generate a first output vector based on the acquired reference signal; using the first fully connected layer to process the first output vector to generate a second output vector after vector transformation; using the second fully connected layer to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

[0011] According to an embodiment of the present disclosure, the deep neural network also includes a third fully connected layer, a fourth fully connected layer and a first activation function, and the method also includes: using the third fully connected layer to generate a third output vector based on the direction signal; using the first activation function to binarize the third output vector to generate a fourth output vector; using the fourth fully connected layer to generate a direction signal with the same dimension as the target speed based on the fourth output vector.

[0012] According to an embodiment of the present disclosure, the method further includes: acquiring a force signal corresponding to the force applied to the top of the vehicle window; and generating an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

[0013] According to an embodiment of the present disclosure, controlling the window to rise or fall according to the target speed based on the target speed and the direction signal indicating the window to rise or fall includes: controlling the window to continue rising or stop according to the target speed based on the target speed, the direction signal and the indication signal.

[0014] According to an embodiment of the present disclosure, the deep neural network also includes a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network, wherein the indication signal is generated by the following operations: using the fifth fully connected layer to generate a fifth output vector based on the acquired force signal; using the bidirectional long short-term memory neural network to generate a sixth output vector based on the first output vector; using the sixth fully connected layer to generate a seventh output vector based on the sum of the fifth output vector and the sixth output vector; using the second activation function to binarize the seventh output vector to generate an eighth output vector; using the seventh fully connected layer to generate the indication signal with the same dimension as the target speed based on the eighth output vector.

[0015] According to an embodiment of the present disclosure, the deep neural network is trained based on the following operations: obtaining a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is the actual speed at which the vehicle window has risen or fallen in history; based on the historical reference signal, using a deep neural network, generating a predicted target speed of the vehicle window; calculating a training loss based on the predicted target speed and the historical target speed, and adjusting parameters of the deep neural network according to the training loss until the training loss meets a predetermined condition.

[0016] An embodiment of the present disclosure provides a device for controlling the lifting and lowering of a vehicle window, comprising: a first acquisition unit, which is configured to acquire a reference signal, wherein the reference signal indicates an expected speed of the vehicle window corresponding to an input command; a first generation unit, which is configured to generate a target speed of the vehicle window based on the acquired reference signal using a deep neural network; and a control unit, which is configured to control the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating that the vehicle window is rising or falling.

[0017] According to an embodiment of the present disclosure, the deep neural network includes a long short-term memory neural network, a first fully connected layer and a second fully connected layer, wherein the first generation unit is configured to: use the long short-term memory neural network to generate a first output vector based on the acquired reference signal; use the first fully connected layer to process the first output vector to generate a second output vector after vector transformation; use the second fully connected layer to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

[0018] According to an embodiment of the present disclosure, the deep neural network also includes a third fully connected layer, a fourth fully connected layer and a first activation function, and the device also includes: a second generation unit, which is configured to use the third fully connected layer to generate a third output vector based on the direction signal; use the first activation function to binarize the third output vector to generate a fourth output vector; use the fourth fully connected layer to generate a direction signal with the same dimension as the target speed based on the fourth output vector.

[0019] According to an embodiment of the present disclosure, the device further includes: a second acquisition unit, which is configured to acquire a force signal corresponding to the force applied to the top of the vehicle window; and a third generation unit, which is configured to generate an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

[0020] According to an embodiment of the present disclosure, the control unit is configured to control the vehicle window to continue rising or stop at the target speed based on the target speed, the direction signal and the indication signal.

[0021] According to an embodiment of the present disclosure, the deep neural network also includes a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network, wherein the third generation unit is configured to: utilize the fifth fully connected layer to generate a fifth output vector based on the acquired force signal; utilize the bidirectional long short-term memory neural network to generate a sixth output vector based on the first output vector; utilize the sixth fully connected layer to generate a seventh output vector based on the sum of the fifth output vector and the sixth output vector; utilize the second activation function to binarize the seventh output vector to generate an eighth output vector; utilize the seventh fully connected layer to generate the indication signal with the same dimension as the target speed based on the eighth output vector.

[0022] According to an embodiment of the present disclosure, the deep neural network is trained based on the following operations: obtaining a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is the actual speed at which the vehicle window has risen or fallen in history; based on the historical reference signal, using a deep neural network, generating a predicted target speed of the vehicle window; calculating a training loss based on the predicted target speed and the historical target speed, and adjusting parameters of the deep neural network according to the training loss until the training loss meets a predetermined condition.

[0023] An embodiment of the present disclosure provides a device for controlling the lifting and lowering of a vehicle window, comprising: a component for acquiring a reference signal, wherein the reference signal indicates an expected speed of the vehicle window corresponding to an input command; a component for generating a target speed of the vehicle window based on the acquired reference signal using a deep neural network; and a component for controlling the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the vehicle window is raised or lowered.

[0024] An embodiment of the present disclosure provides a device for controlling the lifting of a vehicle window, comprising: a processor, and a memory, wherein the memory stores computer executable instructions, and when the computer executable instructions are executed by the processor, the processor is prompted to execute any one of the methods described above.

[0025] An embodiment of the present disclosure provides a computer-readable recording medium storing computer-executable instructions, wherein when the computer-executable instructions are executed by a processor, the processor is prompted to execute any one of the methods described above.

[0026] The present disclosure provides a method, device, equipment and medium for controlling the lifting of vehicle windows. The method, device, equipment and medium provided by the present disclosure can use deep neural networks to well handle the nonlinear problems existing in the control of vehicle window lifting (such as the nonlinear problems between the expected speed and the target speed, and the nonlinear problems between the direction signal and the lifting of the vehicle window), so that the method provided by the present disclosure can accurately control the lifting of the vehicle window, which can be well applied in modeling, simulation or practical application. In addition, due to the existence of the above-mentioned indication signal, the method, device, equipment and medium provided by the present disclosure can provide an anti-pinch function, thereby preventing the vehicle window from pinching a certain part of the human body, which makes the method, device, equipment and medium provided by the present disclosure have a wide range of application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some exemplary embodiments of the present disclosure, and a person of ordinary skill in the art can obtain other drawings based on these drawings without creative work.

[0028] Figure 1 A flow chart of a method 100 for controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown;

[0029] Figure 2 A block diagram showing an example of controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown;

[0030] Figure 3 A simplified block diagram of an example for controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown;

[0031] Figure 4 A block diagram of a device 400 for controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown;

[0032] Figure 5 is a block diagram showing a device 500 for controlling the lifting of a vehicle window according to an embodiment of the present disclosure;

[0033] Figure 6 A schematic diagram of a recording medium 600 according to the present disclosure is shown. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution of the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings of the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the described embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0035] Unless otherwise defined, the technical terms or scientific terms used in the present disclosure should be understood by people with ordinary skills in the field to which the present disclosure belongs. The "first", "second" and similar words used in the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, similar words such as "one", "one" or "the" do not indicate quantity restrictions, but indicate that there is at least one. Similar words such as "include" or "comprise" mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Similar words such as "connect" or "connected" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0036] In the prior art, the nonlinear problem in window lifting control is handled by linearization in modeling, simulation or actual application. This processing method requires a lot of mathematical calculations and is only available within a limited range. When actually applied to vehicles, it will lead to inaccurate window lifting control and even cause problems such as loss of control of the window lifting control system. In addition, in the existing process of controlling window lifting, a large number of analog circuits are used to control the window lifting according to the actual target speed based on the expected speed of the window. This method of using a large number of analog circuits will cause a large control delay and inconsistency between the expected speed and the actual target speed.

[0037] In order to solve the above problems, the present disclosure provides a method for controlling the lifting of vehicle windows. Since the method provided by the present disclosure uses a deep neural network to control the lifting of vehicle windows, it can well and quickly learn linear and nonlinear behaviors from real systems in the real world, and thus can well handle not only nonlinear problems but also linear problems in modeling, simulation or practical applications. Since the present disclosure replaces the above-mentioned large number of analog circuits in the prior art with a deep neural network, the method provided by the present disclosure greatly reduces the control delay and the inconsistency between the expected speed and the actual target speed.

[0038] The method for controlling the lifting of a vehicle window provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0039] Figure 1 A flow chart of a method 100 for controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown. The method 100 may be executed by any suitable device, apparatus, etc., such as a controller or processor in a vehicle for controlling the lifting of a vehicle window.

[0040] Reference Figure 1 In step S110, a reference signal may be acquired, the reference signal indicating an expected speed of the vehicle window corresponding to the input command.

[0041] As an example, the input command may be generated in response to a user pressing or touching a switch in the vehicle for controlling window lift.

[0042] As an example, the reference signal can be obtained by an acquisition unit in a device for controlling the lifting of a window in a vehicle, such as the following Figure 4 The first acquisition unit 410 is introduced.

[0043] As an example, the expected speed of the window may be the speed at which the window is expected to rise or fall. The speed may be a fixed value, such as 2 cm rise (or fall) per second; or a speed sequence, such as v = [1, 2, 3, 2, 1 ...], which may indicate that the window rises (or falls) by 1 cm in the first second, rises (or falls) by 2 cm in the second second, rises (or falls) by 3 cm in the third second, rises (or falls) by 2 cm in the fourth second, rises (or falls) by 1 cm in the fifth second, and so on.

[0044] In step S120, a deep neural network may be used to generate a target speed of the vehicle window based on the acquired reference signal.

[0045] As an example, the deep neural network may be a pre-trained deep neural network, which may be a deep learning neural network (DLNN), such as a long short-term memory neural network (LSTM).

[0046] As an example, the target speed of the car window can be the speed at which the car window actually rises or falls. The speed can be a fixed value, such as rising (or falling) 2 cm per second; or it can be a speed sequence, for example, the speed can be v = [1, 2, 3, 2, 1 ...], which can mean rising (or falling) 1 cm in the first second, rising (or falling) 2 cm in the second second, rising (or falling) 3 cm in the third second, rising (or falling) 2 cm in the fourth second, rising (or falling) 1 cm in the fifth second, etc. It should be noted that the target speed of the car window and the expected speed may also be slightly different.

[0047] According to an embodiment of the present disclosure, the deep neural network may include a long short-term memory neural network, a first fully connected layer and a second fully connected layer, wherein the long short-term memory neural network may be a unidirectional long short-term memory neural network, or a bidirectional long short-term memory neural network.

[0048] The method of using a deep neural network to generate a target speed of the vehicle window based on the acquired reference signal may include: using the long short-term memory neural network to generate a first output vector based on the acquired reference signal; using the first fully connected layer to process the first output vector to generate a second output vector after vector transformation; and using the second fully connected layer to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

[0049] As an example, the acquired reference signal can be input into the above LSTM to generate a first output vector. At this time, the first output vector can be a column vector; then, the column vector can be input into the first fully connected layer to process the column vector, thereby generating a second output vector after vector transformation, and the second output vector can be a lateral quantity. Finally, the second output vector is processed by the second fully connected layer to obtain the target speed of the window of a predetermined dimension, wherein the predetermined dimension can be a predefined dimension that is convenient for signal processing, for example, the predetermined dimension can be one-dimensional, two-dimensional, etc.

[0050] Continue to refer to Figure 1 In step S130, the window may be controlled to rise or fall according to the target speed based on the target speed and the direction signal indicating whether the window is to rise or fall.

[0051] As an example, the direction signal may be generated in response to a user pressing or touching a first switch in the vehicle for controlling a window to rise, to indicate that the window is rising. As another example, the direction signal may be generated in response to a user pressing or touching a second switch in the vehicle for controlling a window to fall, to indicate that the window is falling. The first switch and the second switch may be the same switch.

[0052] According to an embodiment of the present disclosure, the deep neural network may further include a third fully connected layer, a fourth fully connected layer and a first activation function.

[0053] In this case, the method 100 also includes the following operations: using the third fully connected layer to generate a third output vector based on the direction signal; using the first activation function to binarize the third output vector to generate a fourth output vector; using the fourth fully connected layer to generate the direction signal with the same dimension as the target speed based on the fourth output vector.

[0054] As an example, the direction signal can be processed by a third fully connected layer to generate a third output vector after vector transformation. Then, the third output vector after vector transformation is binarized using the first activation function to generate a fourth output vector, wherein the first activation function can be a tanh function, and the fourth output vector obtained after binarization can be a sequence of -1 (e.g., [-1, -1, -1...]) or a sequence of 1 (e.g., [1, 1, 1...]), wherein -1 can represent a decrease and 1 can represent an increase. Finally, the fourth output vector is processed using a fourth fully connected layer to generate a direction signal with the same dimension as the above-mentioned target speed, wherein the value of the direction signal can be a sequence of -1 or a sequence of 1.

[0055] According to an embodiment of the present disclosure, the method 100 may further include: acquiring a force signal corresponding to the force applied to the top of the vehicle window; and generating an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

[0056] As an example, the force applied to the top of the window may be a force applied to the top of the window in response to a body part of a person (e.g., a hand, a head, etc.) pressing the top of the window. In response to the force, a device for lifting the window may generate a corresponding force signal. As another example, the force signal corresponding to the force applied to the top of the window may be a force signal corresponding to the force borne by the window lifter.

[0057] As an example, in response to the force applied to the top of the vehicle window, the acquisition unit in the device for controlling the lifting of the vehicle window (for example, the following combined Figure 4 The second acquisition unit described herein can acquire a force signal corresponding to the force.

[0058] As an example, the indication signal indicating whether the vehicle window continues to rise can be a fixed value, such as 0 or 1, where 0 indicates stopping rising and 1 indicates continuing rising; it can also be a sequence of fixed values, such as a sequence of 0s (e.g., [0,0,0…]) or a sequence of 1s (e.g., [1,1,1…]), where a sequence of 0s indicates stopping rising and a sequence of 1s indicates continuing rising.

[0059] After generating an indication signal indicating whether the vehicle window continues to rise, controlling the vehicle window to rise or fall according to the target speed based on the target speed and the direction signal indicating whether the vehicle window continues to rise or fall may include: controlling the vehicle window to continue rising or stop according to the target speed based on the target speed, the direction signal and the indication signal.

[0060] As an example, based on the target speed, the direction signal and the indication signal, the vehicle window is controlled to continue to rise or stop at the target speed, which can be based on the product result of the target speed, the direction signal and the indication signal. For example, in the case where the direction signal is a sequence of 1 and the indication signal is a sequence of 0, the above-mentioned product result is a sequence of 0, and the method 100 can control the vehicle window to stop. For another example, in the case where the direction signal is a sequence of 1 and the indication signal is a sequence of 1, the value of the above-mentioned product result is the value of the target speed, and the direction is the direction indicated by 1 (i.e., rising), and the method 100 can control the vehicle window to continue to rise.

[0061] According to an embodiment of the present disclosure, the deep neural network may further include a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network.

[0062] In this case, the indication signal is generated by the following operations: using the fifth fully connected layer to generate a fifth output vector based on the acquired force signal; using the bidirectional long short-term memory neural network to generate a sixth output vector based on the first output vector; using the sixth fully connected layer to generate a seventh output vector based on the sum of the fifth output vector and the sixth output vector; using the second activation function to binarize the seventh output vector to generate an eighth output vector; using the seventh fully connected layer to generate the indication signal with the same dimension as the target speed based on the eighth output vector.

[0063] As an example, the acquired force signal is processed by the fifth fully connected layer to generate a fifth output vector. At the same time, the above-mentioned first output vector is processed by a bidirectional long short-term memory neural network to generate a sixth output vector. Then, the fifth output vector and the sixth output vector are added using, for example, an addition function (such as the addition function in MATLAB), and the result of the operation is input into the sixth fully connected layer for processing to generate a seventh output vector. Next, the seventh output vector is binarized using the second activation function to generate an eighth output vector, wherein the second activation function can be a ReLU function, and the eighth output vector obtained after the binarization process can be a sequence of 0s (such as [0,0,0...]) or a sequence of 1s (such as [1,1,1...]), wherein 0 can indicate stopping and 1 can indicate continuing to rise. Finally, the eighth output vector is processed using the seventh fully connected layer to generate an indication signal with the same dimension as the above-mentioned target speed.

[0064] According to an embodiment of the present disclosure, the deep neural network may be trained based on the following operations:

[0065] Step 1, obtaining a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is an actual speed at which the vehicle window was raised or lowered in history;

[0066] As an example, the historical reference signal can be obtained by an acquisition unit in a device for controlling the lifting of a window in a vehicle, such as the following Figure 4 The first acquisition unit 410 is introduced.

[0067] The historical expected speed of the window may be the historical expected speed of the window rising or falling. The speed may be a fixed value, such as rising (or falling) 2 cm per second; or a speed sequence, such as the speed may be v=[1,2,3,2,1...], which may indicate rising (or falling) 1 cm in the first second, rising (or falling) 2 cm in the second second, rising (or falling) 3 cm in the third second, rising (or falling) 2 cm in the fourth second, rising (or falling) 1 cm in the fifth second, and so on.

[0068] The historical target speed of the window may be the speed at which the window has risen or fallen in history. The speed may be a fixed value, such as rising (or falling) 2 cm per second; or a speed sequence, such as the speed may be v = [1, 2, 3, 2, 1 ...], which may indicate rising (or falling) 1 cm in the first second, rising (or falling) 2 cm in the second second, rising (or falling) 3 cm in the third second, rising (or falling) 2 cm in the fourth second, rising (or falling) 1 cm in the fifth second, and so on.

[0069] Step 2, based on the historical reference signal, using a deep neural network to generate a predicted target speed of the vehicle window;

[0070] As an example, the predicted target speed is the predicted speed obtained by the untrained deep neural network. When the predicted speed is consistent with the historical target speed of the car window, it means that the deep neural network is now trained.

[0071] Step 3: Calculate the training loss based on the predicted target speed and the historical target speed, and adjust the parameters of the deep neural network according to the training loss until the training loss meets a predetermined condition.

[0072] As an example, calculating the training loss may be calculating a loss function between the predicted target speed and the historical target speed, wherein the loss function may be a loss function based on mean square error (MSE), a loss function based on mean absolute error (MAE), and the like.

[0073] As an example, the training loss satisfying the predetermined condition may be minimizing the loss function.

[0074] The above has been combined Figure 1 The method for controlling the lifting of vehicle windows provided by the present disclosure is described. According to the method provided by the present disclosure, a deep neural network can be used to well handle the nonlinear problems existing in the control of vehicle window lifting (such as the nonlinear problems between the expected speed and the target speed, and the nonlinear problems between the direction signal and the lifting of the vehicle window), so that the vehicle window lifting can be accurately controlled based on the method provided by the present disclosure, which can be well applied in modeling, simulation or practical application. In addition, due to the presence of the above-mentioned indication signal, the method provided by the present disclosure can provide an anti-pinch function, thereby preventing the vehicle window from pinching a certain part of the human body, which makes the method provided by the present disclosure have a wide range of application prospects.

[0075] In order to facilitate the combination Figure 1 The understanding of the method described below will be combined with Figure 2 , describes the method for controlling the lifting of vehicle windows provided by the present disclosure in the form of an example.

[0076] Figure 2 A block diagram showing an example of controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown.

[0077] Reference Figure 2The deep neural network includes LSTM 220, bidirectional LSTM 230, a first fully connected layer 221, a second fully connected layer 223, a third fully connected layer 251, a fourth fully connected layer 253, a fifth fully connected layer 241, a sixth fully connected layer 261, a seventh fully connected layer 263, an optional fully connected layer 231, a first activation function 252, a second activation function 262, an optional overfitting layer 222, an addition module 260, a multiplication module 270, and an optional regression output module 280. It should be noted that Figure 2 The deep neural network shown is only a preferred example. Optionally, the deep neural network may only include LSTM 220. In this case, the signal processed by LSTM 220 is used as the target speed of the vehicle window, and the vehicle window is directly controlled to rise or fall according to the target speed based on the target speed and direction signal 250.

[0078] Continue to refer to Figure 2 , the signals input to the deep neural network include: a reference signal 210 , a direction signal 250 , and an optional force signal 240 .

[0079] Figure 2 The processing flow of each module or layer shown in is as follows.

[0080] The acquired reference signal 210 is processed by using the LSTM 220, and then the first output vector generated by the processing is processed by using the first fully connected layer 221 to generate a second output vector after vector transformation.

[0081] Then, in the training phase, the second output vector is overfitted using the overfitting layer 222. In the actual use phase of the deep neural network after the training is completed, the overfitting layer 222 does not exist, and the second output vector is processed using the second fully connected layer 223 to generate a target speed of the vehicle window of a predetermined dimension.

[0082] After or at the same time as the reference signal is processed as above, the obtained direction signal 250 is processed by the third fully connected layer 251 to generate a third output vector; then the third output vector is binarized by the first activation function 252 (for example, the above-mentioned tanh function) to generate a fourth output vector. Finally, the fourth output vector is processed by the fourth fully connected layer 253 to generate a direction signal with the same dimension as the above-mentioned target speed.

[0083] In the above method 100 with anti-pinch function, the obtained force signal 240 can be processed by the fifth fully connected layer 241 to generate a fifth output vector. Then, the above first output vector is processed by the bidirectional LSTM 230 to generate a sixth output vector. Optionally, the sixth output vector can be further processed by the optional fully connected layer 231 to obtain a further processed sixth output vector consistent with the dimension of the fifth output vector. Then, the fifth output vector and the sixth output vector are added by the addition module 260, and the result of the addition operation is processed by the sixth fully connected layer 262 to generate a seventh output vector. Then, the seventh output vector is binarized by the second activation function 262 (such as the above ReLU function) to generate an eighth output vector. Next, the eighth output vector is processed by the seventh fully connected layer 263 to generate an indication signal with the same dimension as the above target speed.

[0084] Finally, the indication signal obtained by the fourth fully connected layer 253, the target speed obtained by the second fully connected layer 223, and the direction signal obtained by the seventh fully connected layer 263 are multiplied by the multiplication module 270 to obtain the final upward or downward window movement.

[0085] Optionally, the output obtained by the multiplication module 270 can be processed by the regression output module 280 to obtain a signal that needs to be finally output for display or directly acted on the vehicle window.

[0086] Figure 3 A simplified block diagram of an example for controlling window lifting according to an embodiment of the present disclosure is shown.

[0087] Figure 3 The simplified block diagram shown may exist in a simulation or modeling tool such as Matlab, for example, Simulink in Matlab.

[0088] Reference Figure 3 , Figure 2 The deep neural network described in Figure 3 It is simplified to module 310.

[0089] Figure 3 The module indicated by reference numeral 330 is an analog signal to digital signal conversion module. This is because the input reference signal, force signal and direction signal are often analog signals, while the deep neural network can only process digital signals.

[0090] Figure 3 The module indicated by reference numeral 320 is a conversion module for converting the horizontal quantity into a column vector for the convenience of processing by the deep neural network 310 .

[0091] Figure 3 The module (convert) indicated by the label 340 is a conversion module that prints the output result of the deep neural network 310 so that it can be printed and displayed through the printing module (the module indicated by 350) in Matlab.

[0092] In addition to providing the above-mentioned method for controlling the lifting of a vehicle window, the present disclosure also provides a device for controlling the lifting of a vehicle window. Figure 4 This is explained.

[0093] Figure 4 The block diagram of the device 400 for controlling the lifting of a vehicle window according to an embodiment of the present disclosure is shown. The above description of the method for controlling the lifting of a vehicle window is also applicable to the device 400, unless otherwise explicitly stated.

[0094] Reference Figure 4 The device 400 for controlling the lifting of a vehicle window may include a first acquisition unit 410 , a first generation unit 420 and a control unit 430 .

[0095] According to an embodiment of the present disclosure, the first acquiring unit 410 may be configured to acquire a reference signal, where the reference signal indicates an expected speed of the vehicle window corresponding to the input command.

[0096] As an example, the input command may be generated in response to a user pressing or touching a switch in the vehicle for controlling window lift.

[0097] As an example, the expected speed of the window may be the speed at which the window is expected to rise or fall. The speed may be a fixed value, such as 2 cm rise (or fall) per second; or a speed sequence, such as v = [1, 2, 3, 2, 1 ...], which may indicate that the window rises (or falls) by 1 cm in the first second, rises (or falls) by 2 cm in the second second, rises (or falls) by 3 cm in the third second, rises (or falls) by 2 cm in the fourth second, rises (or falls) by 1 cm in the fifth second, and so on.

[0098] According to an embodiment of the present disclosure, the first generating unit 420 may be configured to generate a target speed of the vehicle window based on the acquired reference signal using a deep neural network.

[0099] As an example, the deep neural network may be a pre-trained deep neural network, which may be a deep learning neural network (DLNN), such as a long short-term memory neural network (LSTM).

[0100] As an example, the target speed of the car window can be the speed at which the car window actually rises or falls. The speed can be a fixed value, such as rising (or falling) 2 cm per second; or it can be a speed sequence, for example, the speed can be v = [1, 2, 3, 2, 1 ...], which can mean rising (or falling) 1 cm in the first second, rising (or falling) 2 cm in the second second, rising (or falling) 3 cm in the third second, rising (or falling) 2 cm in the fourth second, rising (or falling) 1 cm in the fifth second, etc. It should be noted that the target speed of the car window and the expected speed may also be slightly different.

[0101] According to an embodiment of the present disclosure, the deep neural network may include a long short-term memory neural network, a first fully connected layer and a second fully connected layer, wherein the long short-term memory neural network may be a unidirectional long short-term memory neural network, or a bidirectional long short-term memory neural network.

[0102] The first generation unit 420 can be configured to: use the long short-term memory neural network to generate a first output vector based on the acquired reference signal; use the first fully connected layer to process the first output vector to generate a second output vector after vector transformation; use the second fully connected layer to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

[0103] As an example, the acquired reference signal can be input into the above LSTM to generate a first output vector. At this time, the first output vector can be a column vector; then, the column vector can be input into the first fully connected layer to process the column vector, thereby generating a second output vector after vector transformation, and the second output vector can be a lateral quantity. Finally, the second output vector is processed by the second fully connected layer to obtain the target speed of the window of a predetermined dimension, wherein the predetermined dimension can be a predefined dimension that is convenient for signal processing, for example, the predetermined dimension can be one-dimensional, two-dimensional, etc.

[0104] Continue to refer to Figure 4 According to an embodiment of the present disclosure, the control unit 430 may be configured to control the window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the window is to rise or fall.

[0105] As an example, the direction signal may be generated in response to a user pressing or touching a first switch in the vehicle for controlling a window to rise, to indicate that the window is rising. As another example, the direction signal may be generated in response to a user pressing or touching a second switch in the vehicle for controlling a window to fall, to indicate that the window is falling. The first switch and the second switch may be the same switch.

[0106] According to an embodiment of the present disclosure, the deep neural network may further include a third fully connected layer, a fourth fully connected layer and a first activation function.

[0107] In this case, the device 100 may also include: a second generation unit, which is configured to use the third fully connected layer to generate a third output vector based on the direction signal; use the first activation function to binarize the third output vector to generate a fourth output vector; use the fourth fully connected layer to generate a direction signal with the same dimension as the target speed based on the fourth output vector.

[0108] As an example, the direction signal can be processed by a third fully connected layer to generate a third output vector after vector transformation. Then, the third output vector after vector transformation is binarized using the first activation function to generate a fourth output vector, wherein the first activation function can be a tanh function, and the fourth output vector obtained after binarization can be a sequence of -1 (e.g., [-1, -1, -1...]) or a sequence of 1 (e.g., [1, 1, 1...]), wherein -1 can represent a decrease and 1 can represent an increase. Finally, the fourth output vector is processed using a fourth fully connected layer to generate a direction signal with the same dimension as the above-mentioned target speed, wherein the value of the direction signal can be a sequence of -1 or a sequence of 1.

[0109] According to an embodiment of the present disclosure, the device 400 may further include: a second acquisition unit, which is configured to acquire a force signal corresponding to the force applied to the top of the vehicle window; and a third generation unit, which is configured to generate an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

[0110] As an example, the force applied to the top of the window can be the force applied to the top of the window in response to a body part of a person (e.g., hand, head, etc.) pressing the top of the window. In response to the force, the relevant equipment for lifting the window will generate a corresponding force signal.

[0111] As an example, the indication signal indicating whether the vehicle window continues to rise can be a fixed value, such as 0 or 1, where 0 indicates stopping rising and 1 indicates continuing rising; it can also be a sequence of fixed values, such as a sequence of 0s (e.g., [0,0,0…]) or a sequence of 1s (e.g., [1,1,1…]), where a sequence of 0s indicates stopping rising and a sequence of 1s indicates continuing rising.

[0112] After generating the indication signal indicating whether the vehicle window continues to rise, the control unit 430 may be configured to control the vehicle window to continue to rise or stop at the target speed based on the target speed, the direction signal and the indication signal.

[0113] As an example, based on the target speed, the direction signal and the indication signal, the vehicle window is controlled to continue to rise or stop at the target speed, which can be based on the product result of the target speed, the direction signal and the indication signal. For example, in the case where the direction signal is a sequence of 1 and the indication signal is a sequence of 0, the above-mentioned product result is a sequence of 0, and the method 100 can control the vehicle window to stop. For another example, in the case where the direction signal is a sequence of 1 and the indication signal is a sequence of 1, the value of the above-mentioned product result is the value of the target speed, and the direction is the direction indicated by 1 (i.e., rising), and the method 100 can control the vehicle window to continue to rise.

[0114] According to an embodiment of the present disclosure, the deep neural network may further include a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network.

[0115] In this case, the third generation unit is configured to: use the fifth fully connected layer to generate a fifth output vector based on the acquired force signal; use the bidirectional long short-term memory neural network to generate a sixth output vector based on the first output vector; use the sixth fully connected layer to generate a seventh output vector based on the sum of the fifth output vector and the sixth output vector; use the second activation function to binarize the seventh output vector to generate an eighth output vector; use the seventh fully connected layer to generate the indication signal with the same dimension as the target speed based on the eighth output vector.

[0116] As an example, the acquired force signal is processed by the fifth fully connected layer to generate a fifth output vector. At the same time, the above-mentioned first output vector is processed by a bidirectional long short-term memory neural network to generate a sixth output vector. Then, the fifth output vector and the sixth output vector are added using, for example, an addition function (such as the addition function in MATLAB), and the result of the operation is input into the sixth fully connected layer for processing to generate a seventh output vector. Next, the seventh output vector is binarized using the second activation function to generate an eighth output vector, wherein the second activation function can be a ReLU function, and the eighth output vector obtained after the binarization process can be a sequence of 0s (such as [0,0,0...]) or a sequence of 1s (such as [1,1,1...]), wherein 0 can indicate stopping and 1 can indicate continuing to rise. Finally, the eighth output vector is processed using the seventh fully connected layer to generate an indication signal with the same dimension as the above-mentioned target speed.

[0117] According to an embodiment of the present disclosure, the deep neural network may be trained based on the following operations:

[0118] Step 1, obtaining a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is an actual speed at which the vehicle window was raised or lowered in history;

[0119] As an example, the historical reference signal may be obtained by an acquisition unit in a device for controlling the lifting of a vehicle window in a vehicle, such as the first acquisition unit 410 mentioned below.

[0120] The historical expected speed of the window may be the historical expected speed of the window rising or falling. The speed may be a fixed value, such as rising (or falling) 2 cm per second; or a speed sequence, such as the speed may be v=[1,2,3,2,1...], which may indicate rising (or falling) 1 cm in the first second, rising (or falling) 2 cm in the second second, rising (or falling) 3 cm in the third second, rising (or falling) 2 cm in the fourth second, rising (or falling) 1 cm in the fifth second, and so on.

[0121] The historical target speed of the window may be the speed at which the window has risen or fallen in history. The speed may be a fixed value, such as rising (or falling) 2 cm per second; or a speed sequence, such as the speed may be v = [1, 2, 3, 2, 1 ...], which may indicate rising (or falling) 1 cm in the first second, rising (or falling) 2 cm in the second second, rising (or falling) 3 cm in the third second, rising (or falling) 2 cm in the fourth second, rising (or falling) 1 cm in the fifth second, and so on.

[0122] Step 2, based on the historical reference signal, using a deep neural network to generate a predicted target speed of the vehicle window;

[0123] As an example, the predicted target speed is the predicted speed obtained by the untrained deep neural network. When the predicted speed is consistent with the historical target speed of the car window, it means that the deep neural network is now trained.

[0124] Step 3: Calculate the training loss based on the predicted target speed and the historical target speed, and adjust the parameters of the deep neural network according to the training loss until the training loss meets a predetermined condition.

[0125] As an example, calculating the training loss may be calculating a loss function between the predicted target speed and the historical target speed, wherein the loss function may be a loss function based on mean square error (MSE), a loss function based on mean absolute error (MAE), and the like.

[0126] As an example, the training loss satisfying the predetermined condition may be minimizing the loss function.

[0127] The above has been combined Figure 1-Figure 4 The method and device for controlling the lifting of vehicle windows provided by the present disclosure are described. According to the method and device provided by the present disclosure, a deep neural network can be used to well handle the nonlinear problems existing in the control of vehicle window lifting (such as the nonlinear problems between the expected speed and the target speed, and the nonlinear problems between the direction signal and the lifting of the vehicle window), so that the method and device provided by the present disclosure can accurately control the lifting of the vehicle window, which can be well applied in modeling, simulation or practical application. In addition, due to the presence of the above-mentioned indication signal, the method and device provided by the present disclosure can provide an anti-pinch function, thereby preventing the vehicle window from pinching a certain part of the human body, which makes the method and device provided by the present disclosure have a wide range of application prospects.

[0128] In addition to providing the above-mentioned method and device for controlling the lifting of vehicle windows, the present disclosure also provides a device for controlling the lifting of vehicle windows. Figure 5 The above description of the method and device for controlling the lifting of a vehicle window is also applicable to the device described below, unless otherwise explicitly stated.

[0129] Figure 5 is a block diagram showing an apparatus 500 for controlling the lifting and lowering of a vehicle window according to an embodiment of the present disclosure.

[0130] See also Figure 5 , the device 500 may include a processor 501 and a memory 502. The processor 501 and the memory 502 may be connected via a bus 503.

[0131] The processor 501 can perform various actions and processes according to the program stored in the memory 502. Specifically, the processor 501 can be an integrated circuit chip with signal processing capabilities. The above-mentioned processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The various methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc., which can be an X86 architecture or an ARM architecture.

[0132] The memory 502 stores computer instructions, and the above-mentioned method for data quality assessment is implemented when the computer instructions are executed by the processor 501. The memory 402 can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous connection dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM). It should be noted that the memory of the method described herein is intended to include but is not limited to these and any other suitable types of memory.

[0133] The present disclosure also provides another device for controlling the lifting and lowering of a vehicle window, including: a component for acquiring a reference signal, wherein the reference signal indicates an expected speed of the vehicle window corresponding to an input command; a component for generating a target speed of the vehicle window based on the acquired reference signal using a deep neural network; and a component for controlling the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the vehicle window is raised or lowered.

[0134] According to yet another embodiment of the present disclosure, a computer-readable recording medium is also provided. Figure 6 A schematic diagram of a recording medium 600 according to the present disclosure is shown.

[0135] like Figure 6As shown, the recording medium 600 stores computer executable instructions 610. When the computer executable instructions 610 are executed by the processor, the method according to the embodiment of the present disclosure described with reference to the above figures can be executed. The computer-readable recording medium in the embodiment of the present disclosure can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous connection dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM). It should be noted that memory of the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0136] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0137] In general, various example embodiments of the present disclosure may be implemented in hardware or dedicated circuits, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device. When various aspects of the embodiments of the present disclosure are illustrated or described as block diagrams, flow charts, or using some other graphical representation, it will be understood that the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general purpose hardware or controllers or other computing devices, or some combination thereof as non-limiting examples.

[0138] The exemplary embodiments of the present invention described in detail above are merely illustrative, rather than restrictive. It should be understood by those skilled in the art that various modifications and combinations may be made to these embodiments or their features without departing from the principles and spirit of the present invention, and such modifications should fall within the scope of the present invention.

[0139] In the above specific embodiments, it can be seen that different features are combined together in the examples. This disclosure should not be understood as an intention that the example clauses have more features than the features explicitly mentioned in each clause. On the contrary, various aspects of the present disclosure may include fewer features than all the features of the disclosed single example clauses. Therefore, the following clauses should be considered to be included in the specification, where each clause itself can be used as a separate example. Although each subordinate clause in these clauses can refer to a specific combination with one of the other clauses, the aspects of the subordinate clauses are not limited to specific combinations. It should be understood that other example clauses may also include a combination of the subject matter of the subordinate clauses with any other subordinate clauses or independent clauses, or any combination of features with other subordinate clauses and independent clauses. Various aspects disclosed herein explicitly include these combinations, unless it is explicitly expressed or it can be easily inferred that a specific combination is not intended (for example, contradictory aspects, such as defining an element as both an insulator and a conductor). In addition, it is also intended that various aspects of the clauses may also be included in any other independent clauses, even if the clause is not directly subordinate to the independent clause.

[0140] The following numbered clauses describe implementation examples:

[0141] Clause 1. A method for controlling the lifting of a vehicle window, comprising: obtaining a reference signal, the reference signal indicating an expected speed of the vehicle window corresponding to an input command; using a deep neural network to generate a target speed of the vehicle window based on the obtained reference signal; and controlling the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the vehicle window is rising or falling.

[0142] Clause 2. A method according to clause 1, wherein the deep neural network comprises a long short-term memory neural network, a first fully connected layer and a second fully connected layer, wherein the use of the deep neural network to generate a target speed of the vehicle window based on the acquired reference signal comprises: using the long short-term memory neural network to generate a first output vector based on the acquired reference signal; using the first fully connected layer to process the first output vector to generate a second output vector after vector transformation; and using the second fully connected layer to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

[0143] Clause 3. A method according to any one of clauses 1 to 2, wherein the deep neural network further includes a third fully connected layer, a fourth fully connected layer and a first activation function, and the method further includes: using the third fully connected layer to generate a third output vector based on the direction signal; using the first activation function to binarize the third output vector to generate a fourth output vector; using the fourth fully connected layer to generate a direction signal with the same dimension as the target speed based on the fourth output vector.

[0144] Clause 4. A method according to any one of clauses 1 to 3, wherein the method further comprises: acquiring a force signal corresponding to the force applied to the top of the vehicle window; and generating an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

[0145] Clause 5. The method according to any one of clauses 1 to 4, wherein the window is controlled to rise or fall according to the target speed based on the target speed and the direction signal indicating the window to rise or fall, comprises: controlling the window to continue rising or stop according to the target speed based on the target speed, the direction signal and the indication signal.

[0146] Clause 6. A method according to any one of clauses 1 to 5, wherein the deep neural network further includes a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network, wherein the indication signal is generated by the following operations: using the fifth fully connected layer to generate a fifth output vector based on the acquired force signal; using the bidirectional long short-term memory neural network to generate a sixth output vector based on the first output vector; using the sixth fully connected layer to generate a seventh output vector based on the sum of the fifth output vector and the sixth output vector; using the second activation function to binarize the seventh output vector to generate an eighth output vector; using the seventh fully connected layer to generate the indication signal with the same dimension as the target speed based on the eighth output vector.

[0147] Clause 7. A method according to any one of clauses 1 to 6, wherein the deep neural network is trained based on the following operations: obtaining a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is the actual speed at which the vehicle window has risen or fallen in history; based on the historical reference signal, using a deep neural network, generating a predicted target speed of the vehicle window; calculating a training loss based on the predicted target speed and the historical target speed, and adjusting the parameters of the deep neural network according to the training loss until the training loss meets a predetermined condition.

[0148] Item 8. A device for controlling the lifting and lowering of a vehicle window, comprising: a first acquisition unit, which is configured to obtain a reference signal, wherein the reference signal indicates an expected speed of the vehicle window corresponding to an input command; a first generation unit, which is configured to generate a target speed of the vehicle window based on the acquired reference signal using a deep neural network; and a control unit, which is configured to control the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating that the vehicle window is rising or falling.

[0149] Clause 9. An apparatus according to clause 8, wherein the deep neural network comprises a long short-term memory neural network, a first fully connected layer and a second fully connected layer, wherein the first generating unit is configured to: generate a first output vector based on the acquired reference signal using the long short-term memory neural network; process the first output vector using the first fully connected layer to generate a second output vector after vector transformation; and generate a target speed of the vehicle window of a predetermined dimension based on the second output vector using the second fully connected layer.

[0150] Clause 10. An apparatus according to any one of clauses 8 to 9, wherein the deep neural network further includes a third fully connected layer, a fourth fully connected layer and a first activation function, and the apparatus further includes: a second generation unit, which is configured to use the third fully connected layer to generate a third output vector based on the direction signal; use the first activation function to binarize the third output vector to generate a fourth output vector; use the fourth fully connected layer to generate a direction signal with the same dimension as the target speed based on the fourth output vector.

[0151] Clause 11. An apparatus according to any one of clauses 8 to 10, wherein the apparatus further comprises: a second acquisition unit configured to acquire a force signal corresponding to the force applied to the top of the vehicle window; and a third generation unit configured to generate an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

[0152] Clause 12. According to the device of any one of clauses 8 to 11, the control unit is configured to: control the vehicle window to continue to rise or stop according to the target speed based on the target speed, the direction signal and the indication signal.

[0153] Clause 13. An apparatus according to any one of clauses 8 to 12, wherein the deep neural network further includes a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network, wherein the third generation unit is configured to: utilize the fifth fully connected layer to generate a fifth output vector based on the acquired force signal; utilize the bidirectional long short-term memory neural network to generate a sixth output vector based on the first output vector; utilize the sixth fully connected layer to generate a seventh output vector based on the sum of the fifth output vector and the sixth output vector; utilize the second activation function to binarize the seventh output vector to generate an eighth output vector; utilize the seventh fully connected layer to generate the indication signal having the same dimension as the target speed based on the eighth output vector.

[0154] Clause 14. An apparatus according to any one of clauses 8 to 13, wherein the deep neural network is trained based on the following operations: obtaining a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is the actual speed at which the vehicle window has risen or fallen in history; based on the historical reference signal, using a deep neural network, generating a predicted target speed of the vehicle window; calculating a training loss based on the predicted target speed and the historical target speed, and adjusting the parameters of the deep neural network according to the training loss until the training loss meets a predetermined condition.

[0155] Item 15. A device for controlling the lifting and lowering of a vehicle window, comprising: a component for acquiring a reference signal, the reference signal indicating an expected speed of the vehicle window corresponding to an input command; a component for generating a target speed of the vehicle window based on the acquired reference signal using a deep neural network; and a component for controlling the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the vehicle window is raised or lowered.

[0156] Clause 16. A device for controlling vehicle window lifting and lowering, comprising: a processor, and a memory, wherein the memory stores computer executable instructions, and when the computer executable instructions are executed by the processor, the processor is prompted to perform any one of the methods described above.

[0157] Clause 17. A computer-readable recording medium storing computer-executable instructions, wherein when the computer-executable instructions are executed by a processor, the processor is prompted to perform any of the methods described above.

Claims

1. A method for controlling the lifting of a vehicle window, comprising: obtaining a reference signal indicating an expected speed of the vehicle window corresponding to the input command; Using a deep neural network, a target speed of the vehicle window is generated based on the acquired reference signal; Based on the target speed and a direction signal indicating whether the window is to be raised or lowered, the window is controlled to be raised or lowered according to the target speed.

2. The method of claim 1, wherein: The deep neural network includes a long short-term memory neural network, a first fully connected layer and a second fully connected layer, The method of using a deep neural network to generate a target speed of the vehicle window based on the acquired reference signal includes: Using the long short-term memory neural network, based on the acquired reference signal, a first output vector is generated; Processing the first output vector using the first fully connected layer to generate a second output vector after vector transformation; The second fully connected layer is used to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

3. The method of claim 2, wherein: The deep neural network also includes a third fully connected layer, a fourth fully connected layer and a first activation function, The method further comprises: generating, using the third fully connected layer, a third output vector based on the direction signal; Binarizing the third output vector using the first activation function to generate a fourth output vector; The fourth fully connected layer is used to generate a direction signal having the same dimension as the target speed based on the fourth output vector.

4. The method of claim 3, wherein: The method further comprises: obtaining a force signal corresponding to a force applied to a top portion of the vehicle window; Based on the acquired force signal and the target speed, an indication signal is generated to indicate whether the vehicle window continues to rise.

5. The method according to claim 4, wherein the step of controlling the window to rise or fall according to the target speed based on the target speed and the direction signal indicating the window to rise or fall comprises: Based on the target speed, the direction signal and the indication signal, the vehicle window is controlled to continue to rise or stop according to the target speed.

6. The method of claim 5, wherein: The deep neural network also includes a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network, The indication signal is generated by the following operations: generating a fifth output vector based on the acquired force signal using the fifth fully connected layer; Using the bidirectional long short-term memory neural network, based on the first output vector, a sixth output vector is generated; generating, using the sixth fully connected layer, a seventh output vector based on a sum of the fifth output vector and the sixth output vector; Binarizing the seventh output vector using the second activation function to generate an eighth output vector; The indication signal having the same dimension as the target speed is generated based on the eighth output vector using the seventh fully connected layer.

7. The method of claim 1, wherein: The deep neural network is trained based on the following operations: Acquire a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is an actual speed at which the vehicle window was raised or lowered in history; Based on the historical reference signal, using a deep neural network, generating a predicted target speed of the vehicle window; A training loss is calculated based on the predicted target speed and the historical target speed, and a parameter of a deep neural network is adjusted according to the training loss until the training loss meets a predetermined condition.

8. A device for controlling the lifting and lowering of a vehicle window, comprising: a first acquisition unit configured to acquire a reference signal indicating an expected speed of the vehicle window corresponding to the input command; A first generating unit, configured to generate a target speed of the vehicle window based on the acquired reference signal using a deep neural network; The control unit is configured to control the vehicle window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the vehicle window rises or falls.

9. The device of claim 8, wherein: The deep neural network includes a long short-term memory neural network, a first fully connected layer and a second fully connected layer, Wherein, the first generating unit is configured as follows: Using the long short-term memory neural network, based on the acquired reference signal, a first output vector is generated; Processing the first output vector using the first fully connected layer to generate a second output vector after vector transformation; The second fully connected layer is used to generate a target speed of the vehicle window of a predetermined dimension based on the second output vector.

10. The device of claim 9, wherein: The deep neural network also includes a third fully connected layer, a fourth fully connected layer and a first activation function, The device also includes: A second generating unit is configured to generate a third output vector based on the direction signal using the third fully connected layer; binarize the third output vector using the first activation function to generate a fourth output vector; and generate a direction signal with the same dimension as the target speed based on the fourth output vector using the fourth fully connected layer.

11. The device of claim 10, wherein: The device also includes: a second acquisition unit configured to acquire a force signal corresponding to a force applied to a top portion of the vehicle window; The third generating unit is configured to generate an indication signal indicating whether the vehicle window continues to rise based on the acquired force signal and the target speed.

12. The device according to claim 11, wherein the control unit is configured to: Based on the target speed, the direction signal and the indication signal, the vehicle window is controlled to continue to rise or stop according to the target speed.

13. The device of claim 12, wherein: The deep neural network also includes a fifth fully connected layer, a sixth fully connected layer, a seventh fully connected layer, a second activation function and a bidirectional long short-term memory neural network, The third generating unit is configured as follows: generating a fifth output vector based on the acquired force signal using the fifth fully connected layer; Using the bidirectional long short-term memory neural network, based on the first output vector, a sixth output vector is generated; generating, using the sixth fully connected layer, a seventh output vector based on a sum of the fifth output vector and the sixth output vector; Binarizing the seventh output vector using the second activation function to generate an eighth output vector; The indication signal having the same dimension as the target speed is generated based on the eighth output vector using the seventh fully connected layer.

14. The device of claim 8, wherein: The deep neural network is trained based on the following operations: Acquire a historical reference signal and a historical target speed of the vehicle window, wherein the historical reference signal indicates a historical expected speed of the vehicle window corresponding to a historical input command, and the historical target speed is an actual speed at which the vehicle window was raised or lowered in history; Based on the historical reference signal, using a deep neural network, generating a predicted target speed of the vehicle window; A training loss is calculated based on the predicted target speed and the historical target speed, and a parameter of a deep neural network is adjusted according to the training loss until the training loss meets a predetermined condition.

15. A device for controlling the lifting and lowering of a vehicle window, comprising: means for acquiring a reference signal indicative of an expected velocity of the vehicle window corresponding to the input command; A means for generating a target speed of the vehicle window based on the acquired reference signal using a deep neural network; A component for controlling the window to rise or fall according to the target speed based on the target speed and a direction signal indicating whether the window is rising or falling.

16. A device for controlling the lifting and lowering of a vehicle window, comprising: processor, and A memory storing computer executable instructions, which, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 7.

17. A computer-readable recording medium storing computer-executable instructions, wherein: The computer executable instructions, when executed by a processor, cause the processor to perform the method of any one of claims 1-7.

Images