Lane changing method and system, storage medium, and vehicle

Abstract

The application relates to a lane changing method and system, a storage medium and a vehicle. The lane changing method comprises the following steps: receiving continuous multiple frames of condition information, wherein the condition information comprises speed information of a current vehicle, adjacent vehicle state information and lane line information; taking the condition information as an input of a neural network, and obtaining an initial lane changing strategy through neural network processing; and correcting the initial lane changing strategy based on a predetermined rule and the condition information to generate a corrected lane changing strategy and output. According to the lane changing method, intelligent, safe and efficient lane changing can be realized in the process of automatic driving or assisted driving.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automatic driving / assisted driving of vehicles, in particular to a lane changing method, a lane changing system, a storage medium and a vehicle. BACKGROUND

[0002] Lane changing is a common decision-making behavior in the process of automatic driving (or assisted driving, the same below). How to make intelligent and safe lane changing decisions in a complex and variable driving environment is an important issue of automatic driving, and is also one of the important indicators for automatic driving technology to move to a higher level. In actual driving scenarios, the current vehicle and the surrounding environment are in a state of high interaction, and the surrounding vehicles may also have driving behaviors such as acceleration, deceleration, lane changing, etc., so this puts high requirements on the decision-making system of automatic driving (especially the intelligent lane changing function).

[0003] There are schemes in the prior art to generate intelligent lane changing decisions through the following approaches. One scheme is to achieve lane changing through human-designed rules. However, due to the complexity of the driving scene, the rules cannot exhaust all lane changing conditions, so this scheme is difficult to implement in actual application. Another scheme uses machine learning technology to establish a machine learning model for lane changing decisions, and makes intelligent lane changing decisions in different scenarios through the model. SUMMARY

[0004] Embodiments of the present application provide a lane changing method, a lane changing system, a storage medium and a vehicle, so that intelligent, safe and efficient lane changing can be achieved in the process of automatic driving or assisted driving.

[0005] According to an aspect of the present application, a lane changing method is provided, comprising the following steps: receiving condition information, wherein the condition information comprises speed information of a current vehicle, adjacent vehicle state information and lane line information; taking the condition information as input of a neural network, and obtaining an initial lane changing strategy through processing of the neural network; and correcting the initial lane changing strategy based on a predetermined rule and the condition information to generate a corrected lane changing strategy and output.

[0006] In some embodiments of the present application, optionally, the adjacent vehicles include immediately adjacent vehicles in front, behind, left, right, upper left, upper right, lower left and lower right of the current vehicle in the direction of travel of the current vehicle.

[0007] In some embodiments of the present application, optionally, the state information includes lateral and longitudinal speeds of the adjacent vehicles, lateral and longitudinal distances between the adjacent vehicles and the current vehicle.

[0008] In some embodiments of the present application, optionally, the lane line information includes coefficients of lane line fitting curves of the lane in which the current vehicle is located and its adjacent lanes.

[0009] In some embodiments of the application, optionally, the neural network is a long short-term memory neural network.

[0010] In some embodiments of the application, optionally, the predetermined rules comprise at least one of: a speed difference suppression rule that increases a probability of changing lane to an adjacent lane and decreases a probability of keeping the original lane in the initial lane change strategy if a speed difference between a desired speed of the current vehicle and a speed of a vehicle in front of a lane in which the current vehicle is located is higher than a first predetermined value; a fast lane priority rule that increases a probability of changing lane to the left and decreases a probability of changing lane to the right in the initial lane change strategy if a probability of keeping the original lane is lower than a second predetermined value and a difference between a probability of changing lane to the left and a probability of changing lane to the right is lower than a third threshold value; and a decision cooling rule that suppresses changing lane to an adjacent lane in the initial lane change strategy if a predetermined time has elapsed since a last time the current vehicle changes lane to an adjacent left lane.

[0011] According to another aspect of the application, there is provided a lane changing system, comprising: a receiving unit configured to receive condition information, wherein the condition information comprises speed information of a current vehicle, adjacent vehicle state information, and lane line information; a neural network unit configured to generate an initial lane change strategy output with the condition information as input; and an expert rule unit configured to correct the initial lane change strategy based on predetermined rules and the condition information to generate a corrected lane change strategy and output.

[0012] In some embodiments of the application, optionally, the adjacent vehicles comprise vehicles immediately in front of, behind, to the left of, to the right of, above left of, above right of, below left of, and below right of the current vehicle in a direction of travel of the current vehicle.

[0013] In some embodiments of the application, optionally, the state information comprises lateral and longitudinal speeds of the adjacent vehicles, and lateral and longitudinal distances between the adjacent vehicles and the current vehicle.

[0014] In some embodiments of the application, optionally, the lane line information comprises coefficients of lane line fitting curves of a lane in which the current vehicle is located and adjacent lanes thereof.

[0015] In some embodiments of the application, optionally, the neural network unit is constituted by a long short-term memory neural network.

[0016] In some embodiments of the present application, optionally, the predetermined rules comprise at least one of the following: a speed difference suppression rule: increasing the probability of changing lane to the adjacent lane and decreasing the probability of keeping the original lane in the initial lane-changing strategy in the case that the difference between the expected speed of the current vehicle and the speed of the vehicle in front of the lane where the current vehicle is located is higher than a first predetermined value; a fast-lane priority rule: increasing the probability of changing lane to the left and decreasing the probability of changing lane to the right in the initial lane-changing strategy in the case that the probability of keeping the original lane is lower than a second predetermined value and the difference between the probability of changing lane to the left and the probability of changing lane to the right is lower than a third threshold value; and a decision cooling rule: suppressing changing lane to the adjacent lane in the initial lane-changing strategy in the case that a predetermined time has elapsed since the last time the current vehicle changes lane to the adjacent left lane.

[0017] According to another aspect of the present application, there is provided a computer-readable storage medium having instructions stored therein, wherein the instructions, when executed by a processor, cause the processor to perform any one of the lane-changing methods as described above.

[0018] According to another aspect of the present application, there is provided a vehicle comprising any one of the lane-changing systems as described above. BRIEF DESCRIPTION OF DRAWINGS

[0019] The above and other objects and advantages of the present application will become more apparent by describing in detail the following embodiments thereof in connection with the accompanying drawings, in which like reference numerals refer to like elements throughout the several views.

[0020] Figure 1 A lane-changing method according to one embodiment of the present application is shown.

[0021] Figure 2 A lane-changing system according to one embodiment of the present application is shown.

[0022] Figure 3 A lane-changing scenario according to one embodiment of the present application is shown. DETAILED DESCRIPTION

[0023] For brevity and illustrative purposes, the principles of the present application are described herein mainly with reference to exemplary embodiments thereof. However, those skilled in the art will readily appreciate that the same principles are equally applicable and can be practiced in all types of lane-changing methods, lane-changing systems, storage media, and vehicles, and any such variations are intended to be encompassed by the true spirit and scope of the present application.

[0024] One aspect of the present application provides a lane-changing method. As Figure 1As shown, the lane changing method 10 comprises the following steps: receiving condition information in step S102; obtaining an initial lane changing strategy by taking the condition information as the input of a neural network in step S104; and correcting the initial lane changing strategy based on a predetermined rule and the condition information in step S106. The lane changing strategies in the present application include keeping the original lane, changing to the left lane, and changing to the right lane, and each strategy is accompanied by a certain probability. In some examples, the strategy with the highest probability can be selected as the output when the final strategy is determined. It should be noted that the "correction" herein includes the special case of maintaining the original result.

[0025] The lane changing method 10 according to some aspects of the present application receives condition information under a plurality of continuous frames obtained by an image sensor or the like during driving in step S102. The condition information includes speed information of the current vehicle, state information of adjacent vehicles, and lane line information. In this context, the current vehicle refers to the vehicle implementing the lane changing method 10, and the adjacent vehicle refers to a vehicle adjacent to the current vehicle and which can affect the lane changing decision. The condition information received in step S102 is the basic data for implementing the lane changing method 10, and therefore reliable condition information is a prerequisite for generating a scientific lane changing strategy. It should be noted that this document mainly considers same-direction lane changing, and therefore the opposite lane and the vehicles thereon are not within the scope of the study.

[0026] In some examples, the adjacent vehicles can be vehicles within the detectable range of the vehicle-mounted detectors (e.g., millimeter wave radar, laser radar, visual sensor, etc.). Depending on the type and number of detectors, the number of adjacent vehicles within the detectable range can vary. The advantage of selecting adjacent vehicles in this way is that the data is more abundant, and therefore the reliability of the decision can be higher. However, if too many adjacent vehicles are considered, the computational load can be large, which can affect the efficiency of the decision.

[0027] In some embodiments of the present application, the most relevant vehicles among the vehicles within the detectable range of the detector can be selected as the adjacent vehicles. For example, as shown in Figure 3 The adjacent vehicles can be the immediately adjacent vehicle C2 in front of the current vehicle C0, the immediately adjacent vehicle C7 behind the current vehicle C0, the immediately adjacent vehicle C4 to the left of the current vehicle C0, the immediately adjacent vehicle C5 to the right of the current vehicle C0, the immediately adjacent vehicle C1 above and to the left of the current vehicle C0, the immediately adjacent vehicle C3 above and to the right of the current vehicle C0, the immediately adjacent vehicle C6 below and to the left of the current vehicle C0, and the immediately adjacent vehicle C8 below and to the right of the current vehicle C0. It should be noted that the adjacent vehicles herein refer to vehicles that can be theoretically considered, and if such vehicles do not exist in actual situations, it does not affect the implementation of the present application. For example, if the detector does not detect the immediately adjacent vehicle C8 below and to the right of the position shown in the figure within the predetermined range, then this position will be "empty" when implementing each step of the present application.

[0028] In some embodiments of the present application, continuing to refer toFigure 3 The state information of the adjacent vehicles includes lateral and longitudinal velocities of the adjacent vehicles, and lateral and longitudinal distances between the adjacent vehicles and the current vehicle. Taking the left adjacent vehicle C4 as an example, the current velocity of the left adjacent vehicle C4 is V, which can be decomposed into a horizontal component V x (lateral velocity) and a vertical component V y (longitudinal velocity). The lateral distance between the left adjacent vehicle C4 and the current vehicle C0 is X, and the longitudinal distance is Y. The state information is conducive to depicting the state of each adjacent vehicle, and thus an accurate lane changing strategy can be generated accordingly.

[0029] In some embodiments of the present application, the lane line information includes coefficients of lane line fitting curves of the lane in which the current vehicle is located and its adjacent lanes (if any). Continuing to refer to Figure 3 , the lane M in which the current vehicle C0 is located shares a lane line L2 with the left adjacent lane K, and shares a lane line L3 with the right adjacent lane N. The left adjacent lane K further includes a lane line L1, and the right adjacent lane N further includes a lane line L4. In some examples of the present application, the lane lines L1, L2, L3 and L4 shown in the figure are mainly studied, and thus the lane line information includes coefficients of the fitting curves of each of the lane lines L1, L2, L3 and L4.

[0030] The lane changing method 10 according to some aspects of the present application takes the condition information as the input of the neural network in step S104, and obtains an initial lane changing strategy via neural network processing. Before the neural network is used to process real-time input data and generate an initial lane changing strategy, the neural network can be trained using artificial driving experience data. The related training process can be carried out according to the prior art, and thus will not be described here.

[0031] In some embodiments of the present application, a long short-term memory (LSTM) neural network is used to generate the initial lane changing strategy in step S104. Long short-term memory is a special recurrent neural network (RNN), which can perform better in longer sequences than ordinary recurrent neural networks. The inventors have found in the development process that long short-term memory neural networks have better effects than other types of neural networks in the process of processing vehicle autonomous lane changing strategies, not only higher efficiency, but also more satisfactory lane changing strategies generated.

[0032] The lane changing method 10 according to some aspects of the present application corrects the initial lane changing strategy based on predetermined rules and condition information in step S106 to generate a corrected lane changing strategy and output. After obtaining the probability of the initial lane changing strategy (e.g., keep straight, change to left or right) of the current vehicle in the real-time environment through step S104, the probability value can be further processed by a customized expert system. The expert system can use existing knowledge or experience to optimize the results generated by step S104, thereby achieving better results for complex decision-making problems. Specifically, the expert system can combine intuitive driving experience of humans in the driving process on the basis of the output results of the neural network to optimize the output lane changing decision.

[0033] The advantage of using an expert system is that the intelligent lane changing decision output can be made more in line with the expectations of the driver by customizing the addition / modification / deletion of experience in the system, which is also more conducive to the maintenance and iteration of the decision-making machine.

[0034] In some embodiments of the present application, the predetermined rules mentioned in step S106 can include the following:

[0035] (1) Speed difference suppression rule. The neural network uses driving data of human drivers when learning, and different drivers choose different conditions to change lanes when there is a slow vehicle in front. When processing the lane changing probability given by the neural network, the expert system first limits the decision-making scenario to "there is a vehicle in front, and the difference between the speed of the vehicle and the expected driving speed of the current vehicle is greater than a certain threshold, and remains for a certain time". The expected driving speed of the current vehicle can be represented by the current cruise speed set by the driver. When the speed difference suppression condition is met, the output probabilities of the neural network can be revised. For example, for scenarios with a large speed difference from the front vehicle, the probability values of left / right lane changing are appropriately increased and the probability value of straight driving is decreased. Specifically, in the case where the difference between the expected speed of the current vehicle and the speed of the vehicle in front of its lane is higher than a first predetermined value, the probability of changing lanes to the adjacent lane in the initial lane changing strategy is increased and the probability of keeping the original lane is decreased.

[0036] (2) Fast Lane Priority Rule. When the probability of changing lanes to one side is much greater than the other two outputs, it can be assumed that most drivers will choose to change lanes to that side in this scenario, and the expert system will select this direction as the direction for the lane-changing decision. When the probability of going straight is extremely small, but the probabilities of changing lanes to both sides are equal, it can be assumed that changing lanes to the left or right is in line with the driver's expectations. Considering that most roads use the left lane as the fast lane, and there may be unobserved slow vehicles ahead in the right lane, the expert system will appropriately increase the output probability of changing lanes to the left and decrease the output probability of changing lanes to the right, so that the overall decision is more inclined to drive in the fast lane. Specifically, in the initial lane-changing strategy, if the probability of keeping the original lane is lower than the second predetermined value and the difference between the probability of changing lanes to the adjacent left lane and the probability of changing lanes to the adjacent right lane is lower than the third threshold, the probability value of changing lanes to the left is increased and the probability value of changing lanes to the right is decreased.

[0037] (3) Decision Cooling Rules. During driving, due to the switching between various states of the current vehicle, there are some scenarios where frequent lane-changing decisions are undesirable even if the conditions for triggering intelligent lane changing are met. For example, if the vehicle has just completed a lane-changing action, changing lanes again at this time will increase the driver's sense of insecurity. For these scenarios, the expert system will identify them during operation and set a cooling-off time according to each scenario category. During the cooling-off time, even if the driving scenario meets the requirements, intelligent lane-changing decisions will be suppressed by the expert system. Specifically, the system will suppress lane-changing to adjacent lanes in the initial lane-changing strategy within a predetermined time after the current vehicle has completed its last lane-changing to the adjacent left lane.

[0038] Another aspect of this application provides a lane-changing system. For example... Figure 2 As shown, the lane-changing system 20 includes a receiving unit 202, a neural network unit 204, and an expert rule unit 206. Although shown as separate units, these unit modules can be integrated. For example, the neural network unit 204 and the expert rule unit 206 can be implemented using dedicated or general-purpose processors (with necessary storage devices).

[0039] The receiving unit 202 of the lane-changing system 20 is configured to receive conditional information from multiple consecutive frames acquired during driving, such as those obtained by an image sensor. This conditional information includes the current vehicle's speed, adjacent vehicle status information, and lane line information. In this paper, the current vehicle can refer to the vehicle belonging to the lane-changing system 20, and adjacent vehicles refer to vehicles near the current vehicle that may influence the lane-changing decision. The conditional information received by the receiving unit 202 is the fundamental data for the lane-changing system 20 to continue operating; therefore, reliable conditional information is a prerequisite for generating a scientific lane-changing strategy. It should be noted that this paper mainly considers same-direction lane changing; therefore, opposite lanes and vehicles on them are not within the scope of this study.

[0040] In some examples, neighboring vehicles can be those within the detection range of onboard detectors (e.g., millimeter-wave radar, lidar, vision sensors, etc.). The number of neighboring vehicles within the detection range may vary depending on the type and number of detectors. The advantage of selecting neighboring vehicles in this way is richer data, potentially leading to more reliable decisions. However, considering too many neighboring vehicles can result in significant computational costs, which may affect decision-making efficiency.

[0041] In some embodiments of this application, the lane-changing system 20 can select several vehicles most relevant to the lane change from those within the detection range of the vehicle detector (not shown in the figure) as adjacent vehicles. For example, as Figure 3 As shown, adjacent vehicles can be the following vehicles in the direction of travel of the current vehicle C0 (in the direction of the arrow at the front of the vehicle in the figure): the vehicle immediately in front (C2), the vehicle immediately behind (C7), the vehicle immediately to the left (C4), the vehicle immediately to the right (C5), the vehicle immediately to the upper left (C1), the vehicle immediately to the upper right (C3), the vehicle immediately to the lower left (C6), and the vehicle immediately to the lower right (C8). It should be noted that adjacent vehicles here refer to vehicles that can theoretically be considered. If such vehicles do not exist in reality, it does not affect the implementation of this invention. For example, if the detector does not detect the vehicle immediately to the lower right (C8) at the position shown in the figure within the predetermined range, then the lane-changing system 20 can set this position as "empty".

[0042] In some embodiments of this application, reference continues to be made to Figure 3 The state information of adjacent vehicles includes their lateral and longitudinal velocities, as well as their lateral and longitudinal distances from the current vehicle. Taking the vehicle C4 immediately to the left as an example, its current velocity is V, which can be decomposed into a horizontal component V0. x (lateral velocity) and vertical component V y (Longitudinal speed). The lateral distance between the adjacent vehicle C4 on the left and the current vehicle C0 is X, and the longitudinal distance is Y. The state information will help to describe the state of each adjacent vehicle, and the lane-changing system 20 can then generate an accurate lane-changing strategy based on this information.

[0043] In some embodiments of this application, lane line information includes coefficients of the lane line fitting curves of the lane currently occupied by the vehicle and its adjacent lanes (if any). Continue to refer to... Figure 3 The lane M where vehicle C0 is currently located shares lane line L2 with the adjacent lane K on the left and lane line L3 with the adjacent lane N on the right. The adjacent lane K on the left also includes lane line L1, and the adjacent lane N on the right also includes lane line L4. In some examples of the present invention, the lane lines L1, L2, L3, and L4 illustrated are mainly studied, and therefore the lane line information considered by the lane changing system 20 includes the coefficients of the fitted curves of each of lane lines L1, L2, L3, and L4.

[0044] The neural network unit 204 of the lane changing system 20 is configured to take conditional information as input and generate an initial lane changing strategy output. The neural network unit 204 can be trained with human driving experience data before it is used to process real-time input data and produce an initial lane changing strategy. The relevant training process can be carried out according to the prior art, which will not be described here.

[0045] In some embodiments of the present application, the neural network unit 204 is composed of a long short-term memory neural network. Long short-term memory is a special recurrent neural network that can perform better in longer sequences than ordinary recurrent neural networks. The inventors have found in the development process that long short-term memory neural networks have better results than other types of neural networks in processing vehicle autonomous lane changing strategies, not only in terms of efficiency, but also in terms of the lane changing strategies generated.

[0046] The expert rule unit 206 of the lane changing system 20 is configured to correct the initial lane changing strategy based on predetermined rules and conditional information to generate a corrected lane changing strategy and output. After the neural network unit 204 obtains the probability of the initial lane changing strategy (e.g., keep straight, change left or right) of the current vehicle in the real-time environment, the probability value can be further processed by a customized expert rule unit 206. The expert rule unit 206 can use existing knowledge or experience to optimize the results produced by the neural network unit 204, thereby achieving better results for complex decision-making problems. Specifically, the expert rule unit 206 can combine intuitive driving experience in the driving process on the basis of the output of the neural network unit 204 to optimize the output lane changing decision.

[0047] The advantage of using an expert system is that the intelligent lane changing decision output can be made more in line with the expectations of the driver by customizing the addition / modification / deletion of experience in the system, which is also more conducive to the maintenance and iteration of the decision-making machine.

[0048] In some embodiments of the present application, the predetermined rules employed in the expert rule unit 206 can include the following aspects. (1) The speed difference suppression rule, the basic principle of which can be found in the foregoing description and will not be repeated here. Specifically, the probability of changing lanes to the adjacent lane in the initial lane-changing strategy is increased and the probability of keeping the original lane is decreased when the difference between the expected speed of the current vehicle and the speed of the vehicle in front of the lane in which the current vehicle is located is higher than a first predetermined value. (2) The fast lane priority rule, the basic principle of which can be found in the foregoing description and will not be repeated here. Specifically, the probability of changing lanes to the left is increased and the probability of changing lanes to the right is decreased when the probability of keeping the original lane in the initial lane-changing strategy is lower than a second predetermined value and the difference between the probability of changing lanes to the left and the probability of changing lanes to the right is lower than a third threshold value. (3) The decision cooling rule, the basic principle of which can be found in the foregoing description and will not be repeated here. Specifically, changing lanes to the adjacent lane in the initial lane-changing strategy is inhibited within a predetermined time after the current vehicle has completed changing lanes to the adjacent left lane.

[0049] Another aspect of the present application provides a vehicle comprising any one of the lane-changing systems described above. The vehicle configured with the lane-changing system can achieve intelligent, safe and efficient lane-changing during autonomous driving or assisted driving.

[0050] According to another aspect of the present application, a computer-readable storage medium is provided, in which instructions are stored, which when executed by a processor, cause the processor to perform any one of the lane-changing methods described above. The computer-readable medium referred to in the present application includes various types of computer storage media, which can be any available medium that can be accessed by a general or special purpose computer. For example, the computer-readable medium can include RAM, ROM, EPROM, EEPROM, PROM, registers, hard disks, removable disks, CD-ROMs or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other temporary or non-temporary medium that can be used to carry or store desired program code elements in the form of instructions or data structures and can be accessed by a general or special purpose computer, or a general or special purpose processor. As used herein, a disk typically magnetically replicates data, while a disc optically replicates data with a laser. Combinations of the above should also be included within the scope of computer-readable media. The exemplary storage medium is coupled to the processor so that the processor can read information from and write information to the storage medium. In the alternative, the storage medium can be integrated into the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal. 2 PROM, registers, hard disks, removable disks, CD-ROMs or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other temporary or non-temporary medium that can be used to carry or store desired program code elements in the form of instructions or data structures and can be accessed by a general or special purpose computer, or a general or special purpose processor. As used herein, a disk typically magnetically replicates data, while a disc optically replicates data with a laser. Combinations of the above should also be included within the scope of computer-readable media. The exemplary storage medium is coupled to the processor so that the processor can read information from and write information to the storage medium. In the alternative, the storage medium can be integrated into the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal.

[0051] The above some examples of the present application provide solutions to intelligent lane changing decision of an autonomous vehicle in a complex driving environment. According to the solutions, intelligent and safe lane changing decisions can be automatically generated according to current lane line information, continuous running states of surrounding vehicles, continuous running state of the current vehicle, and the like. The control system of the autonomous driving system can realize lane changing actions according to the lane changing decisions.

[0052] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Other feasible changes or replacements can be thought of by those skilled in the art according to the technical range disclosed in the present application, and such changes or replacements are all covered in the protection scope of the present application. The embodiments and features in the embodiments can also be combined with each other without conflict. The protection scope of the present application is subject to the recitation of the claims.

Claims

1. A lane change method, characterized by, The method comprises the following steps: receiving condition information, wherein the condition information comprises speed information of a current vehicle, state information of adjacent vehicles, and lane line information; inputting the condition information into a neural network, and obtaining an initial lane changing strategy via processing of the neural network; and correcting the initial lane changing strategy based on predetermined rules and the condition information to generate a corrected lane changing strategy and output; the lane line information comprises coefficients of lane line fitting curves of a lane where the current vehicle is located and adjacent lanes thereof; the predetermined rules comprise: a speed difference suppression rule: in a case where a speed difference between a desired speed of the current vehicle and a speed of a vehicle in front of the lane where the current vehicle is located is higher than a first predetermined value, increasing a probability of lane changing to an adjacent lane in the initial lane changing strategy and decreasing a probability of keeping the original lane; a fast lane priority rule: in a case where the probability of keeping the original lane in the initial lane changing strategy is lower than a second predetermined value, and a probability difference between a probability of lane changing to an adjacent left lane and a probability of lane changing to an adjacent right lane is lower than a third threshold value, increasing a probability value of lane changing to the left and decreasing a probability value of lane changing to the right.

2. The method of claim 1, wherein, the adjacent vehicles comprise vehicles immediately in front of, behind, to the left of, to the right of, above left of, above right of, below left of, and below right of the current vehicle in a direction of travel of the current vehicle.

3. The method of claim 1, wherein, the state information comprises lateral and longitudinal speeds of the adjacent vehicles, and lateral and longitudinal distances between the adjacent vehicles and the current vehicle.

4. The method of claim 1, wherein, the neural network is a long short-term memory neural network.

5. The method of claim 1, wherein, the predetermined rules further comprise: a decision cooling rule: inhibiting lane changing to an adjacent lane in the initial lane changing strategy within a predetermined time after completion of last lane changing to an adjacent left lane by the current vehicle.

6. A lane change system, characterized by The system comprises: a receiving unit configured to receive condition information, wherein the condition information comprises speed information of a current vehicle, state information of adjacent vehicles, and lane line information; a neural network unit configured to input the condition information and generate an initial lane changing strategy output; and an expert rule unit configured to correct the initial lane changing strategy based on predetermined rules and the condition information to generate a corrected lane changing strategy and output; the lane line information comprises coefficients of lane line fitting curves of a lane where the current vehicle is located and adjacent lanes thereof; the predetermined rules comprise: a speed difference suppression rule: in a case where a speed difference between a desired speed of the current vehicle and a speed of a vehicle in front of the lane where the current vehicle is located is higher than a first predetermined value, increasing a probability of lane changing to an adjacent lane in the initial lane changing strategy and decreasing a probability of keeping the original lane; a fast lane priority rule: in a case where the probability of keeping the original lane in the initial lane changing strategy is lower than a second predetermined value, and a probability difference between a probability of lane changing to an adjacent left lane and a probability of lane changing to an adjacent right lane is lower than a third threshold value, increasing a probability value of lane changing to the left and decreasing a probability value of lane changing to the right.

7. The system of claim 6, wherein, the adjacent vehicles comprise vehicles immediately in front of, behind, to the left of, to the right of, above left of, above right of, below left of, and below right of the current vehicle in a direction of travel of the current vehicle.

8. The system of claim 6, wherein, The state information comprises lateral and longitudinal velocities of the adjacent vehicle, and lateral and longitudinal distances between the adjacent vehicle and the current vehicle.

9. The system of claim 6, wherein, The neural network unit is composed of a long short-term memory neural network.

10. The system of claim 6, wherein, The predetermined rules further comprise: A decision cooling rule: inhibiting lane changing to an adjacent lane in the initial lane changing strategy within a predetermined time after the current vehicle completes lane changing to an adjacent left lane last time.

11. A computer-readable storage medium having stored therein instructions, the computer-readable storage medium comprising: When the instructions are executed by a processor, the processor executes the method of any one of claims 1-5.

12. A vehicle characterized by comprising: The vehicle comprises the lane changing system of any one of claims 6-10.

Images