Method, device, vehicle and program product for detecting lane change of vehicle

Abstract

The invention relates to a method and device for detecting lane change of a vehicle, the vehicle and a program product. The method includes obtaining a plurality of pieces of state information for a continuous period of time of the vehicle. The method further determines the lane changing probability of the vehicle based on the multiple pieces of state information and a vehicle lane changing prediction model, the vehicle lane changing prediction model is an inference model based on a convolutional neural network, and the step of determining the lane changing probability of the vehicle comprises transverse convolution and longitudinal convolution of the multiple pieces of state information. The method further includes determining that the vehicle will lane in response to the lane change probability being above a predetermined threshold. By means of the method, lane changing of the vehicle can be more accurately predicted, the following target is updated for the vehicle in the automatic cruise driving process in time, the driving safety of the vehicle is improved, and in addition, the driving experience and comfort of a driver are greatly improved.

Description

Technical Field

[0001] The present disclosure relates to the field of vehicle intelligent driving, and more specifically, to a method, device, vehicle, and program product for detecting vehicle lane changes. Background Art

[0002] In modern vehicle driving technology, the automatic cruise function, namely ACC (Active Cruise Control), is an important intelligent driving technology in modern vehicle driving technology. The automatic cruise function is not a simple speed control, but the vehicle automatically maintains driving at the speed set by the driver, reducing the driver's driving burden. The automatic cruise technology can also automatically adjust the vehicle speed according to the road conditions to avoid frequent driver intervention, especially in long-distance driving. This can undoubtedly significantly reduce the driver's fatigue and improve driving safety. In addition, the traffic jam assistance system TJA (Traffic Jam Assistant) can also provide great help to the driver. Summary of the invention

[0003] In a first aspect of an embodiment of the present disclosure, a method for detecting a vehicle lane change is provided. The method includes acquiring a plurality of state information of the vehicle within a continuous time period. The method also determines the lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing horizontal convolution and vertical convolution on the plurality of state information. The method further includes determining that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold.

[0004] In a second aspect of an embodiment of the present disclosure, a method for driving assistance is provided. The method includes acquiring a plurality of state information of the vehicle within a continuous time period. The method also determines the lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing horizontal convolution and vertical convolution on the plurality of state information. The method further includes determining that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold.

[0005] In a third aspect of an embodiment of the present disclosure, a domain controller is provided. The domain controller includes a controller configured to execute a method for detecting a vehicle lane change. The method includes acquiring a plurality of state information within a continuous time period of the vehicle. The method also determines the lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing horizontal convolution and vertical convolution on the plurality of state information. The method further includes determining that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold.

[0006] In a fourth aspect of an embodiment of the present disclosure, a computer program product is provided. The computer program product includes machine executable instructions, which, when executed, cause a machine to implement a method for detecting a vehicle lane change. The method includes acquiring a plurality of state information within a continuous time period of the vehicle. The method also determines the lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing horizontal and vertical convolution on the plurality of state information. The method further includes determining that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold.

[0007] In a fifth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. Computer-executable instructions are stored on the computer-readable storage medium, wherein the computer-executable instructions are executed by a processor to implement a method for detecting a vehicle lane change. The method includes acquiring a plurality of state information within a continuous time period of the vehicle. The method also determines the lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing lateral convolution and longitudinal convolution on the plurality of state information. The method further includes determining that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold.

[0008] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0010] Figure 1 A schematic diagram illustrating an example environment in which various embodiments of the present disclosure may be implemented;

[0011] Figure 2 A flow chart of a method for detecting a vehicle lane change according to some embodiments of the present disclosure is shown;

[0012] Figure 3 A schematic diagram showing an example process of training a vehicle lane change prediction model according to some embodiments of the present disclosure;

[0013] Figure 4 A block diagram showing an example process of training a vehicle lane change prediction model according to some embodiments of the present disclosure;

[0014] Figure 5 A flow chart of a method for detecting a vehicle lane change according to some embodiments of the present disclosure is shown;

[0015] Figure 6 A block diagram of an electronic device in which various embodiments of the present disclosure may be implemented is shown. DETAILED DESCRIPTION

[0016] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure. The embodiments of the present disclosure described below with reference to the accompanying drawings are only for exemplary purposes.

[0017] As mentioned above, when using a vehicle system (such as ACC or TJA, etc., ACC is used as an example in the following) to follow a target vehicle, since the movement of vehicles on the road is dynamic, ACC's target selection also needs to be dynamic. ACC can calculate the driving task probability of its own vehicle: for example, whether the vehicle is lane keeping, or changing lanes left or right. The driving task probability calculation method mainly follows the traditional Markov process and Gaussian probability distribution. Figure 1 1 is a schematic diagram of an example environment 100 in which various embodiments of the present disclosure may be implemented. Figure 1As shown, many drivers complain that the timing of ACC's following target selection is too late when the own vehicle 110 changes lanes. For example, when the own vehicle 110 turns left from the own vehicle lane to the left adjacent lane, ACC can only change the original following target vehicle 120 to the following target vehicle 130 in the adjacent lane after the own vehicle completely enters the left adjacent lane. This performance is difficult for drivers to accept, because during the entire left turn process, ACC still uses the original following target vehicle 120 as the following target, which is obviously not appropriate.

[0018] It can be seen that a method is needed to detect the lane change of the vehicle in advance so that the correct following target can be selected earlier. Figure 1 If it is known in advance that the own vehicle 110 will change to the adjacent left lane, the correct following target vehicle 130 is selected before the own vehicle 110 reaches the adjacent left lane, or when the own vehicle 110 is turning left, instead of continuing to select the original following target vehicle 120. In this way, the driver will feel that it is more natural, more reasonable and safe, thereby greatly improving the driver's driving experience.

[0019] In summary, the present disclosure can provide a method for detecting vehicle lane changes on one hand. The method includes obtaining multiple state information of the vehicle within a continuous time period. The method also determines the lane change probability of the vehicle based on the multiple state information and the vehicle lane change prediction model. And, the method further determines that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold. Through the method of the present disclosure, the lane change of the vehicle can be detected in advance, and based on the lane change information, the ACC can better adjust its following target, thereby making driving more natural, safe and reliable, and greatly improving the driver's experience and comfort. The following is combined with Figure 1 and Figure 2 Some embodiments of the method are described in detail.

[0020] Figure 1 1 is a schematic diagram of an example environment 100 in which various embodiments of the present disclosure may be implemented. Figure 1 As shown, the own vehicle 110 has the ability of environmental perception, which can be provided by an environmental perception device, such as a lidar, or a millimeter wave radar, an ultrasonic sensor, an infrared sensor or a camera, etc. It can collect its own status information, and it can also collect the status information of the same-direction vehicles within a certain range around it (such as the following target vehicle 120).

[0021] Figure 2A flow chart of a method 200 for detecting a vehicle lane change according to some embodiments of the present disclosure is shown. At box 210, multiple state information of the vehicle in a continuous time period is obtained. Preferably, the state information collected by the own vehicle 110 may be the motion state of the vehicle. Preferably, the motion state may include any one of the following or a combination thereof: the position of the vehicle; the speed of the vehicle; the acceleration of the vehicle; the yaw angle of the vehicle; or the yaw rate of the vehicle. The position of the vehicle includes the longitudinal position X of the vehicle (i.e., the driving direction) and the lateral position Y of the vehicle (i.e., the position relative to the adjacent lane). The speed of the vehicle includes the longitudinal speed Vx and the lateral speed Vy (i.e., the lateral speed relative to the adjacent lane). The acceleration of the vehicle includes the longitudinal acceleration ax and the lateral acceleration ay. The yaw angle of the vehicle is the angle between the changed driving direction and the original driving direction, and the yaw angle of the vehicle is generally recorded as θ. The yaw rate of the vehicle refers to the rate of change of the yaw angle of the vehicle over time, and the yaw rate of the vehicle is generally recorded as dθ. Preferably, the selected time period T1, ACC can collect the vehicle status information T1: [X, Y, Vx, Vy, ax, ay, θ, dθ]. It is worth noting that the above state information selection is only exemplary and does not constitute any limitation to the present disclosure. Those skilled in the art can make any motion parameter selection applicable to the present disclosure according to the present disclosure, and these different selections and combinations thereof should also be within the scope of protection of the present disclosure. Preferably, multiple state information within a continuous time period can be selected, and the time period is between 100 milliseconds and 500 milliseconds. For example, a 4-second time period of the own vehicle 110 can be selected for continuous sampling at an interval of 200 milliseconds to obtain multiple state information of the own vehicle 110, thus forming multiple state information within 20 continuous time periods: T1, T2, ..., T20. It is worth noting that the above time period and its sampling time period are only exemplary and do not constitute any limitation to the present disclosure. Those skilled in the art can make any time interval and time period applicable to the present disclosure according to the present disclosure, such as using 100 milliseconds as the time period, using a time interval of 3 seconds, etc., and these different selections and combinations thereof should also be within the scope of protection of the present disclosure. By sampling status information over a longer time period, it is possible to more accurately predict vehicle lane changes, thereby better solving the problem of vehicle lane change prediction and taking subsequent measures based on accurate vehicle lane change predictions, thereby greatly improving vehicle safety and providing the driver with a good driving experience.

[0022] At box 220, the lane change probability of the vehicle is determined based on the multiple state information and the vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing horizontal convolution and vertical convolution on the multiple state information. Among them, method 200 also includes receiving multiple historical state information of the vehicle within a continuous time period, and also includes training the vehicle lane change prediction model using the historical state information of the vehicle. Preferably, wherein training the vehicle lane change prediction model using the historical state information of the vehicle includes: collecting the historical state information of the vehicle; and marking the historical state information as one of the following marks: lane change; or not lane change. Preferably, considering the booming artificial intelligence algorithms nowadays, the vehicle lane change prediction model is any one of the following or a combination thereof: convolutional neural network CNN, recurrent neural network RNN, long short-term memory network LSTM or deep neural network DNN, etc. Preferably, the vehicle lane change prediction model can be a CNN model, and when training the model, an activation function is also used to increase the nonlinearity of the inter-layer output, and it also includes performing lateral convolution and longitudinal convolution on the multiple historical state information, and merging the results of the lateral convolution and longitudinal convolution. The vehicle lane change prediction model uses historical state information, and after training, it can be used to predict the probability of the vehicle changing lanes based on the input current relevant state information of the vehicle. In the subsequent description of the present disclosure, multiple embodiments of the present disclosure will be combined to explain in detail how to train the vehicle lane change prediction model based on the CNN model. Based on the multiple state information obtained, using the vehicle lane change prediction model, method 200 can determine the lane change probability of the vehicle. The use of a vehicle lane change prediction model based on an artificial intelligence algorithm is conducive to improving the accuracy of the vehicle's lane change, so that ACC can take accurate subsequent following actions, improve the safety of the vehicle, and also improve the driving experience and comfort of the driver.

[0023] At box 230, in response to the lane change probability being higher than a predetermined threshold, it is determined that the vehicle will change lanes. The vehicle lane change prediction model outputs a predicted vehicle lane change probability based on multiple input state information. If the lane change probability is higher than a predetermined threshold, for example, 70%, it can be determined that the own vehicle 110 will change lanes. It is worth noting that the predetermined threshold value set to 70% is only an example and does not limit the protection scope of the present disclosure. Those skilled in the art can adopt any other predetermined threshold value applicable to the present disclosure, such as 65% or 80%, based on experience and actual usage. Therefore, any predetermined threshold value applicable to the present disclosure selected by those skilled in the art should be within the protection scope of the present disclosure. Preferably, if the own vehicle 110 is in an automatic cruising vehicle, the method 200 also includes, in response to determining that the vehicle will change lanes, reselecting a follow target for the vehicle as the vehicle in front of the lane being changed. For example, in Figure 1In the example, ACC can select the front vehicle 130 in the adjacent lane as the following target for the own vehicle 110. The advantage of adopting this method is that the following target vehicle can be selected in advance for the own vehicle 110, so that when it changes to the adjacent lane, it can follow the new following target more smoothly, instead of waiting for the own vehicle 110 to completely change lanes, as in the existing method, to re-determine a new following target for the own vehicle 110. Therefore, the driving safety of the own vehicle 110 can be greatly improved, and the own vehicle 110 will not follow the old following target vehicle during the lane change, which is easy to cause a safety accident. Moreover, for the driver, when changing lanes, his eyes must be focused on the front vehicle in the lane being changed. At this time, updating the following target to the front vehicle in the lane being changed is obviously in full compliance with the driver's natural driving habits, thereby greatly improving the driver's driving experience and comfort, and of course, more importantly, improving the driver's driving safety. This is something that the existing methods cannot do. Preferably, if the vehicle is a vehicle being followed in automatic cruise control, such as vehicle 120, method 200 further includes, in response to determining that the vehicle will change lanes, reselecting the following target as other front vehicles in the lane. If the own vehicle 110 continues to maintain the own lane, and method 200 determines that the vehicle 120 as the following target will change lanes, then using method 200, ACC can adjust the following target for the own vehicle 110 in advance, and use the front vehicle 140 in the own lane as the new following target of the own vehicle 110. Since the following target 120 is going to change lanes, and the own vehicle 110 maintains the own lane, it is no longer practical to continue to follow it. After using method 200, the following target is adjusted to the front vehicle 140 in the own lane in time, which greatly improves the safety of the own vehicle 110. Moreover, for the driver, when changing lanes by following the target 120, the driver's eyes will naturally turn to the vehicle 140 in front of the lane due to driving habits. At this time, it is natural for the driver to locate the new following target vehicle 140, which greatly improves the driver's driving experience and comfort, and more importantly, improves the driver's driving safety. This is also something that existing methods cannot do.

[0024] Figure 3 A schematic diagram of an example process of training a vehicle lane change prediction model according to some embodiments of the present disclosure is shown. As described above, method 200 receives a plurality of historical state information of a vehicle in a continuous time period, and further includes training the vehicle lane change prediction model using the historical state information of the vehicle, and preferably collecting the historical state information of the vehicle, and marking the historical state information as one of the following marks: lane change; or no lane change. In training method 300, Figure 3The collected historical state information 310 is shown. The historical state information 310 schematically selects 15 historical state information within a certain historical continuous time period: T1: [X, Y, Vx, Vy, ax, ay, θ, dθ], T2: [X, Y, Vx, Vy, ax, ay, θ, dθ], ..., T15: [X, Y, Vx, Vy, ax, ay, θ, dθ]. It should be noted that the use of 15 historical state information is only used to facilitate the diagram to illustrate the process of training the vehicle lane change prediction model, and is not used to limit the protection scope of the present disclosure. Those skilled in the art can use 10, 20, 100, ... 1000, ... 10000 and other historical state information to train the vehicle lane change prediction model according to actual needs. Those skilled in the art use an appropriate number of historical state information for application to the present disclosure, which should fall within the protection scope of the present disclosure. In the training method 300, preferably, the basic model based on CNN is trained to finally form a trained vehicle lane change prediction model. An obvious advantage of the CNN-based basic model is that the basic model has a small amount of computation, and the final generated training model is also relatively small. The trained vehicle lane change prediction model does not require too high hardware computing power, so it can be deployed in some devices with weaker computing power, making real-time calculation and prediction possible, thereby truly ensuring the vehicle's driving safety and truly improving the driver's driving experience and comfort.

[0025] In step 3100, each historical state information in the historical state information 310 is subjected to its own horizontal convolution, so that a horizontal convolution result 320 corresponding to each historical state information is obtained. For each historical state information containing a 6-bit vector, the convolution result 320 is a corresponding 8-bit vector. As mentioned above, one of the main features of CNN is to extract multi-layer features of input data by using convolution operations. After step 3100, the convolution result 320 is obtained, that is, the relevant features of the original historical state information are extracted, and the feature vector of the first layer of the original historical state information, that is, the convolution result 320, is formed.

[0026] In step 3200, each of the convolution results 320 is further convolved horizontally again, so that a 16-bit secondary convolution result 330 is obtained. Based on the secondary convolution result 330, in step 3300, each of the secondary convolution results 330 is convolved vertically, so as to obtain a 16-bit vertical convolution 340. In step 3400, the 16-bit vertical convolution 340 is added to each of the secondary convolution results 330, so as to form a new merged convolution result 350. The merged convolution result 350 is a 32-bit vector. After the horizontal and vertical convolution steps 3200, 3300 and 3400, the features of the historical state information 310 are further extracted. By performing horizontal and vertical convolution on the historical state information 310 and merging the results of the horizontal and vertical convolutions, the feature information of the historical state information 310 can be extracted to a large extent, which provides help for training a more accurate vehicle lane change prediction model.

[0027] In step 3500, the 32-bit merged convolution result 350 is convolved again to form a new 32-bit cubic convolution result 360. In step 3600, the cubic convolution result 360 is convolved again to obtain a new quaternary convolution result 370. The quaternary convolution result 370 is also a 32-bit vector. It is worth noting that convolution is a typical and basic operation in CNN. The specific convolution algorithm used during convolution does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art can adopt any specific convolution algorithm applicable to the present disclosure according to actual needs, and this should fall within the protection scope of the present disclosure. In addition, the above-mentioned multiple embodiments show the use of a four-layer convolution model, but those skilled in the art can adopt other applicable multi-layer convolution models, such as 10 layers, according to actual needs. These other applicable multi-layer convolution models also fall within the protection scope of the present disclosure.

[0028] In step 3700, a global maximum pooling GMP (Global Max Pooling) operation is performed on the four-time convolution result 370. GMP refers to a pooling operation method used in the field of deep learning. GMP performs a sliding window operation on the input sequence and selects the maximum value in each window as the output. After the four-time convolution result 370 is subjected to the GMP operation, a 32-bit pooling result 380 is formed. The 32-bit pooling result 380 is used as input to enter the dense product step 3800.

[0029] In the dense step 3800, as an example, one-dimensional convolution is used to perform dense processing on the pooling result 380, and three one-dimensional convolutions are performed exemplarily to obtain a 16-bit convolution result 390, an 8-bit convolution result 392, and a 1-bit convolution result 394 in sequence.

[0030] In step 3900, the convolution result 394 obtained in step 3800 is subjected to the Sigmoid function. Due to the monotonically increasing properties of the Sigmoid function and the monotonically increasing properties of the inverse function, the convolution result 394 is mapped between 0 and 1, thereby obtaining a probability result 396. Thus, a mapping relationship between the historical state information 310 and the probability result 396 is obtained.

[0031] Thus, in the training method 300, through a large amount of historical state information 310, the above steps 3100-3900 are repeatedly applied to repeatedly train the vehicle lane change prediction model, and a CNN-based vehicle lane change prediction model can be obtained. A large amount of historical state information 310 can be accumulated through daily driving activities. Preferably, a large amount of historical state information 310 can also be obtained by continuous sampling of intelligent driving activities. Preferably, each of the above convolution results is transformed by a Relu function to increase the nonlinearity of the convolution result, thereby further activating the training of the vehicle lane change prediction model. Using the trained vehicle lane change prediction model, the probability of the vehicle changing lanes can be more accurately predicted using the collected vehicle state information. For those vehicles whose predicted lane change probability is greater than 70%, it can be considered that they will change lanes. It is worth noting that the predetermined threshold value set to 70% is only an example and does not limit the scope of protection of the present disclosure. Those skilled in the art can adopt any other predetermined thresholds applicable to the present disclosure, such as 65%, 75% or 80%, based on experience and actual usage. Therefore, any predetermined thresholds applicable to the present disclosure selected by those skilled in the art should be within the protection scope of the present disclosure. The vehicle lane change prediction model obtained by using the above-mentioned multiple embodiments requires less computing power and can be applied to electronic devices such as vehicle-mounted radars and vehicle motion controllers, making the implementation and deployment of the present disclosure more flexible, truly ensuring the driving safety of the vehicle and improving the driving experience of the driver.

[0032] Figure 4A block diagram of an example process of training a vehicle lane change prediction model according to some embodiments of the present disclosure is shown. The trainer 400 is composed of an input layer 410, a convolution layer 420, a fully connected layer 430, and an output layer 440. In the input layer 410, the historical state information 310 is input. In the convolution layer 420, the historical state information 310 is subjected to a horizontal convolution 422 and a vertical convolution 423 as described in detail above, and the results of the horizontal convolution and the vertical convolution are merged, and then two horizontal convolutions 424 and 427 are performed. Preferably, GMP processing 425 and 428 can also be performed. Preferably, each convolution result can also be transformed by a Relu function to increase the nonlinearity of the convolution result, thereby further activating the training of the vehicle lane change prediction model. Finally, the final convolution result is subjected to GMP processing to obtain a 32-bit pooling result 380. Preferably, the secondary convolution result 330 can also be subjected to GMP processing, and the processing result is then merged with the vertical convolution result. Preferably, the obtained GMP processing can also be processed with a flattening function. In the fully connected layer 430, three one-dimensional convolutions 431, 432 and 433 are performed as described above, thereby converting the final convolution result into a 1-bit vector. Finally, in the output layer 440, the Sigmoid function is applied to the convolution result 396 so that the convolution result 396 is mapped between 0 and 1, thereby obtaining a probability result 398. Thus, the mapping relationship between the historical state information 310 and the probability result 398 is obtained. Based on a large amount of historical state information of the vehicle, after repeated training by the trainer 400, a vehicle lane change prediction model can be obtained. In summary, the vehicle lane change prediction model disclosed in the present invention is an inference model based on a convolutional neural network, which has the characteristics of low computing power and easy deployment. The model performs horizontal and vertical convolution on the speed and position information of the vehicle in a continuous time period, merges the convolution results of the horizontal and vertical convolutions, and performs multiple convolution and pooling operations on the merged results, and finally obtains the final vehicle lane change probability through a fully connected layer and a Sigmoid function. Based on the trained vehicle lane change prediction model, the vehicle's lane change probability can be predicted more accurately, thereby improving the vehicle's driving safety and improving the driver's driving experience and comfort.

[0033] Figure 5 A flow chart of a method 500 for detecting a vehicle lane change according to some embodiments of the present disclosure is shown. Figure 5 and Figure 1Some embodiments of the present disclosure for detecting vehicle lane changes are introduced in detail. Method 500 is used to detect vehicle lane changes and timely update the follow target for the vehicle according to the situation of the vehicle lane change. At box 510, the motion state of the vehicle is detected in real time. Among them, the self-driving vehicle 110 detects its own motion state on the one hand, and on the other hand, it can also detect the motion state of its follow target 120 and store its motion state information. The motion state information can be stored in the permanent disk of the vehicle, and can also be backed up in the cloud storage, or it can be temporarily stored in the cache of the vehicle as temporary information. The motion information is collected and sampled in a certain time segment, and the time segment can be 2-4 seconds, or longer, and the time period as the sampling interval can be between 100 milliseconds and 500 milliseconds.

[0034] At box 520, the detected motion state information is sent to the trained vehicle lane change prediction model. At box 530, the vehicle lane change prediction model receives the motion state information and performs calculations based on the motion state information to obtain the probability of the vehicle changing lanes. At box 540, the probability of the vehicle changing lanes is compared with a predetermined threshold to determine the lane change information. As an example, the predetermined threshold can be pre-set to 70%. In response to the probability of the vehicle changing lanes being greater than 70%, it is determined that the vehicle will change lanes, and conversely, in response to the probability of the vehicle changing lanes being less than or equal to 70%, it is determined that the vehicle will continue to drive in the original lane and will not change lanes.

[0035] At box 550, the following target is updated for the own vehicle according to the determined lane change information. In response to the motion information coming from the own vehicle and determining that the own vehicle will change lanes, and at this time the own vehicle control system determines that the own vehicle will turn left, the following target of the own vehicle 110 in the automatic cruise state is changed from vehicle 120 to the vehicle 130 in front of the left lane; in response to the motion information coming from the following target vehicle 120 and determining that the following target vehicle 120 will change lanes, the target vehicle is reselected for the own vehicle 110 as the vehicle 140 in front of the lane. Through multiple embodiments of the method 500 disclosed in the present invention, a safer environment can be provided for the driving of the vehicle, and at the same time, since this change is more in line with the driving habits of the driver, the driver's experience and comfort can be greatly improved. In addition, since the vehicle lane change prediction model is based on the basic CNN model, the required computing power is small, so it can be flexibly deployed and can be closely coordinated with the vehicle control system without adding too much computing burden to the vehicle system.

[0036] In some embodiments, the vehicle motion state information in method 500 includes any one or a combination of the vehicle's position, the vehicle's speed, or the vehicle's yaw angle. These parameters can be used to basically determine whether the vehicle will change lanes. Preferably, the vehicle motion state information may also include any one or a combination of the vehicle's acceleration or the vehicle's yaw rate. Further use of these parameters can better and more fully determine the real-time motion state of the vehicle, and is more conducive to determining whether the vehicle will change lanes. It should be noted that the selection of the above motion state information is merely exemplary, and those skilled in the art may select any reasonable vehicle motion state information for use in the present disclosure, and these reasonable selections by those skilled in the art should be considered to fall within the scope of protection of the present disclosure.

[0037] In some embodiments, method 500 further includes: training the vehicle lane change prediction model using the historical state information of the vehicle, wherein training the vehicle lane change prediction model using the historical state information of the vehicle includes: receiving multiple historical state information of the vehicle within a continuous time period, and wherein training the vehicle lane change prediction model using the historical state information of the vehicle includes: collecting the historical state information of the vehicle; and marking the historical state information as one of the following marks: lane change; or no lane change. Preferably, the vehicle lane change prediction model is a CNN model, and the CNN model has no more than 10 layers to facilitate deployment. Preferably, the number of continuous time periods is more than 10, and the continuous time period is between 100 milliseconds and 500 milliseconds. It also includes performing horizontal convolution and vertical convolution on the historical state information, and merging the results of the horizontal convolution and vertical convolution. Preferably, method 500 collects the historical state information of the vehicle, including collecting state information of the autonomous driving vehicle.

[0038] In some embodiments, the vehicle is a vehicle in automatic cruising, and the method 500 further includes: in response to determining that the vehicle will change lanes, reselecting a following target for the vehicle as a front vehicle in the lane to be changed. The method 500 further includes: in response to determining that the vehicle will change lanes, reselecting a following target as another front vehicle in the lane. The method further includes: in response to the lane change probability being lower than or equal to a predetermined threshold, determining that the vehicle will not change lanes.

[0039] In some embodiments of the present disclosure, the present disclosure further provides a domain controller. The domain controller includes a controller configured to execute a method for detecting a vehicle lane change. The method includes acquiring a plurality of state information within a continuous time period of the vehicle. The method also determines the lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle includes: performing horizontal convolution and vertical convolution on the plurality of state information. The method further includes determining that the vehicle will change lanes in response to the lane change probability being higher than a predetermined threshold.

[0040] Figure 6 A schematic block diagram of an example device 600 that may be used to implement embodiments of the present disclosure is shown. Figure 1 The electronic device 104 in the embodiment can be implemented by the device 600. As shown in the figure, the device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 602 or computer program instructions loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device 600 can also be stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0041] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a memory 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0042] The various processes and processing described above, such as methods 200 and 300, may be performed by the processing unit 601. For example, in some embodiments, methods 200 and 300 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the CPU 601, one or more actions of the methods 100 and 300 described above may be performed.

[0043] The present disclosure may be a method, a radar device, an electronic device, a system, a vehicle and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present disclosure.

[0044] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (non-exhaustive list) of computer readable storage medium include: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination of the above. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (e.g., a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0045] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0046] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0047] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0048] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0049] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0050] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0051] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method (200) for detecting a vehicle lane change, comprising: Acquiring (210) a plurality of status information of the vehicle within a continuous time period; Determining (220) a lane change probability of the vehicle based on the plurality of state information and a vehicle lane change prediction model, wherein the vehicle lane change prediction model is an inference model based on a convolutional neural network, and determining the lane change probability of the vehicle comprises: performing horizontal convolution and vertical convolution on the plurality of state information; as well as In response to the lane change probability being above a predetermined threshold, it is determined ( 230 ) that the vehicle will change lanes.

2. The method (200) according to claim 1, wherein the state information is the motion state of the vehicle, and the motion state includes any one of the following items or a combination thereof: The location of the vehicle; The speed of the vehicle; The acceleration of the vehicle; the vehicle's yaw angle; or The vehicle's yaw rate.

3. The method (200) of claim 1, further comprising: The vehicle lane change prediction model is trained using the historical state information of the vehicle.

4. The method (200) according to claim 3, wherein using the historical state information of the vehicle to train the vehicle lane change prediction model comprises: A plurality of historical status information of the vehicle within consecutive time periods is received.

5. The method (200) according to claim 3, wherein using the historical state information of the vehicle to train the vehicle lane change prediction model comprises: Collecting historical status information of the vehicle; as well as Mark the historical status information as one of the following: Change lanes; or Don't change your path.

6. The method (200) according to claim 1, wherein the vehicle is a vehicle in automatic cruising, the method further comprising: In response to determining (230) that the vehicle will change lanes, a following target is reselected for the vehicle as a leading vehicle in the lane being changed.

7. The method (200) according to claim 1, wherein the vehicle is a followed vehicle in automatic cruising, the method further comprising: In response to determining (230) that the vehicle will change lanes, reselecting other leading vehicles whose following targets are in the own lane.

8. The method (200) according to claim 1, wherein the vehicle lane change prediction model is a convolutional neural network (CNN) model, and the CNN model has no more than 10 layers to facilitate deployment.

9. The method (200) according to claim 1, wherein the number of the continuous time periods is greater than 10, and the continuous time periods are between 100 milliseconds and 500 milliseconds.

10. The method (200) according to claim 4, wherein the method further comprises performing horizontal convolution and vertical convolution on the historical state information, merging the convolution results of the horizontal convolution and the vertical convolution, performing multiple convolution and pooling operations on the merged results, and obtaining the vehicle lane change probability through a fully connected layer and a Sigmoid function.

11. The method (200) according to claim 5, wherein collecting historical status information of the vehicle includes collecting status information of an autonomous driving vehicle.

12. The method (200) of claim 1, wherein the method further comprises: In response to the lane change probability being less than or equal to a predetermined threshold, it is determined that the vehicle will not change lanes.

13. A method for driving assistance, the method comprising the steps of any one of claims 1 to 12 for the driving assistance.

14. The method according to claim 13, wherein the driving assistance method includes automatic cruise control or traffic jam assistance.

15. A domain controller, comprising: A controller configured to execute the method according to any one of claims 1 to 12.

16. A computer program product, comprising machine executable instructions, which when executed cause a machine to implement the method according to any one of claims 1 to 12.

Images