A cooperative lane-change decision-making method for mixed manned and unmanned traffic flow

Abstract

The present invention discloses a collaborative lane change decision method for mixed unmanned and manned traffic flow, and relates to the field of traffic control technology. The method includes: monitoring traffic flow to generate real-time traffic flow; predicting lane change behavior based on the state sequence of manned vehicles to generate manned lane change prediction results; based on the manned lane change prediction results, using a game decision network to make collaborative lane change decisions for multiple unmanned vehicles according to the state sequence of unmanned vehicles to generate unmanned lane change decisions; determining the first unmanned vehicle that needs to change lanes and the first lane change constraint information; and monitoring and reminding the first unmanned vehicle to change lanes according to the first lane change constraint information. The method solves the technical problem in the prior art that it is difficult for unmanned vehicles to effectively respond to the lane change behavior of manned vehicles in mixed traffic flows, and achieves the technical effect of improving the accuracy of unmanned vehicle lane change decisions and traffic safety.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic control, and in particular to a coordinated lane change decision method for mixed traffic flows of people and vehicles. Background Art

[0002] With the rapid development of driverless technology, driverless vehicles are increasingly used in road traffic. However, in actual road environments, driverless vehicles and manned vehicles often need to operate together, forming a mixed traffic flow of driverless and manned vehicles. In this complex traffic environment, due to the high uncertainty of the behavior of manned vehicles, their lane-changing behavior has a significant impact on the normal driving and safety decisions of driverless vehicles. Driverless vehicles not only need to perceive traffic conditions in real time, but also must accurately predict the driving intentions of manned vehicles to ensure the safety and effectiveness of their lane-changing decisions. In the prior art, there is a lack of a collaborative decision-making method that can effectively respond to the lane-changing behavior of manned vehicles in mixed traffic flows for driverless vehicles, resulting in decision-making delays and increased risks when driverless vehicles change lanes. Summary of the invention

[0003] The present application provides a collaborative lane-changing decision method for mixed unmanned and manned traffic flow, which solves the technical problem in the prior art that unmanned vehicles have difficulty in effectively responding to the lane-changing behavior of manned vehicles in mixed traffic flow.

[0004] In view of the above problems, the present application provides a collaborative lane change decision method for mixed traffic flow with humans and no humans.

[0005] The present application provides a coordinated lane change decision method for mixed traffic flow of people and vehicles, the method comprising:

[0006] A traffic monitoring module in a preset road area is connected to perform traffic flow monitoring to generate real-time traffic flow, wherein the real-time traffic flow includes the state sequence of unmanned vehicles and the state sequence of manned vehicles; lane change behavior is predicted based on the state sequence of the manned vehicle to generate a manned lane change prediction result; based on the manned lane change prediction result, a game decision network is used to perform collaborative lane change decisions for multiple unmanned vehicles according to the state sequence of the unmanned vehicles to generate an unmanned lane change decision; based on the unmanned lane change decision, a first unmanned vehicle that needs to change lanes and first lane change constraint information are determined; and lane change monitoring and reminders are performed on the first unmanned vehicle according to the first lane change constraint information.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] First, the traffic monitoring module in the preset road area is connected to monitor the traffic flow and generate real-time traffic flow, wherein the real-time traffic flow includes the state sequence of the unmanned vehicle and the state sequence of the manned vehicle. Next, the lane change behavior is predicted based on the state sequence of the manned vehicle to generate the manned lane change prediction result. Then, based on the manned lane change prediction result, the game decision network is used to make a coordinated lane change decision for multiple unmanned vehicles according to the state sequence of the unmanned vehicle to generate an unmanned lane change decision. Further, the first unmanned vehicle and the first lane change constraint information that need to change lanes are determined based on the unmanned lane change decision. Finally, the first unmanned vehicle is monitored and reminded to change lanes according to the first lane change constraint information. The technical problem that it is difficult for unmanned vehicles to effectively respond to the lane change behavior of manned vehicles in mixed traffic flows in the prior art is solved, and the technical effect of improving the accuracy of unmanned vehicle lane change decisions and traffic safety is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic flow chart of a coordinated lane change decision method for mixed traffic flow with humans and humans provided in an embodiment of the present application;

[0011] Figure 2 A flowchart of lane change behavior prediction in a collaborative lane change decision method for mixed traffic flow with humans and no humans provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application solves the technical problem in the prior art that unmanned vehicles have difficulty in effectively responding to the lane changing behavior of manned vehicles in mixed traffic flows by providing a collaborative lane changing decision method for unmanned and manned mixed traffic flows.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0015] Examples, such as Figure 1 As shown, the embodiment of the present application provides a coordinated lane change decision method for mixed traffic flow of people and vehicles, wherein the method includes:

[0016] A traffic monitoring module in a preset road area is connected to perform traffic flow monitoring to generate real-time traffic flow, wherein the real-time traffic flow includes the state sequence of unmanned vehicles and the state sequence of manned vehicles.

[0017] In the pre-set road area, a traffic monitoring module is deployed. The traffic monitoring module includes cameras, radars, lidars, etc., which can collect the movement and status data of vehicles in the area in real time. Based on the data collected from the traffic monitoring module, real-time traffic flow information is generated. By connecting the traffic monitoring module in the preset road area, real-time traffic flow can be obtained. Traffic flow refers to the movement and overall trend of all vehicles in the entire traffic system. These data will be arranged in time to form a state sequence, that is, the specific state of each vehicle at different time points. The state sequence of unmanned vehicles refers to the detailed motion state information of each unmanned vehicle at different time points, including the speed, position, acceleration / deceleration, direction, whether the turn signal is on, whether the lane is changing, and the distance from other surrounding vehicles at a certain moment. Similarly, the state sequence of manned vehicles also includes all motion state data of the vehicle when it is driving on the road.

[0018] A lane change behavior prediction is performed based on the state time series of the manned vehicle to generate a manned lane change prediction result.

[0019] Based on the state time series data of the manned vehicle, the lane change behavior is predicted to generate a manned lane change prediction result, which describes the possible lane change time and position of the manned vehicle. Optionally, machine learning, deep learning or statistical methods are used to analyze the driver's driving behavior and predict whether he is likely to change lanes at a certain moment in the future.

[0020] Furthermore, if Figure 2 As shown, the lane change behavior prediction is performed based on the state time sequence of the manned vehicle to generate a manned lane change prediction result, including:

[0021] Collect a historical vehicle lane change data set, wherein the historical vehicle lane change data set includes historical driving behavior, historical mixed traffic flow characteristics and corresponding lane change behavior characteristics; establish a lane change behavior prediction model based on the historical driving behavior and the historical mixed traffic flow characteristics; analyze the state time series of the manned vehicle through the lane change behavior prediction model to generate a first lane change prediction result; perform a lane change impact analysis based on the state time series of the manned vehicle, optimize the first lane change prediction result, and generate the manned lane change prediction result.

[0022] Preferably, a historical vehicle lane change data set is collected, including historical driving behavior information, such as the driver's acceleration, braking, steering and other behavior records, and historical mixed traffic characteristics, such as vehicle density, speed distribution, lane occupancy, etc., and also includes corresponding lane change behavior characteristics, such as lane change time, lane change direction, lane change success or failure, etc.; using historical driving behavior and historical mixed traffic characteristics as training data, a machine learning algorithm (such as decision tree, neural network, etc.) is selected to train a lane change behavior prediction model. By analyzing the relationship between historical driving behavior and actual lane change behavior, the lane change behavior prediction model can learn to identify the driver's lane change intention under different driving behavior characteristics and mixed traffic characteristics; the state time series of the current manned vehicle is input into the lane change behavior prediction model. A real-time analysis is performed on the prediction model to obtain a first lane change prediction result, which is a prediction based on the current vehicle behavior and includes a preliminary conclusion on whether the driver has the intention to change lanes. A lane change impact analysis is performed based on the state sequence of the manned vehicle, that is, based on factors such as the current vehicle state, surrounding environment, and traffic rules, the impact of the vehicle on its own driving after changing lanes is analyzed, such as the driving smoothness after changing lanes, whether it is safe when changing lanes, and whether congestion can be reduced. The results of the lane change impact analysis are combined with the first lane change prediction result, and the first lane change prediction result is optimized to generate a manned lane change prediction result, which more accurately reflects the current manned vehicle's lane change intention.

[0023] Furthermore, performing a lane change impact analysis based on the state time series of the manned vehicle, optimizing the first lane change prediction result, and generating the manned lane change prediction result includes:

[0024] A first manned vehicle is extracted according to the state time series of the manned vehicle, and a first candidate lane change behavior is established according to the lane where the first manned vehicle is located; lane change fitting is performed with the first candidate lane change behavior to obtain a first lane change self-impact of the first manned vehicle after the lane change, and traffic flow impact analysis after the lane change is performed in combination with the state time series of the manned vehicle to generate a first lane change traffic flow impact; the first lane change self-impact and the first lane change traffic flow impact are weightedly fused to generate a first lane change comprehensive impact, wherein the weight of the first lane change self-impact is greater than the weight of the first lane change traffic flow impact; the first lane change prediction result is optimized with the first lane change comprehensive impact to generate the manned lane change prediction result.

[0025] Specifically, the first manned vehicle currently being analyzed is identified and extracted from the state time series data of manned vehicles, and a possible first candidate lane change behavior is established according to the lane in which the vehicle is currently located, including the direction of the lane change (left or right), the time of the lane change, the speed of the lane change, etc. The first candidate lane change behavior is subjected to lane change fitting, that is, the possible position and speed changes of the vehicle after executing the lane change behavior are simulated; through lane change fitting, the first lane change impact of the first manned vehicle after executing the lane change is obtained, and the first lane change impact includes whether the vehicle is safer after the lane change, whether the driving efficiency is improved, whether congestion can be avoided, etc. Combined with the state time series data of manned vehicles, the impact of the lane change behavior on other surrounding vehicles and the entire traffic flow is analyzed, including analyzing whether the lane change of the vehicle will affect the driving of other vehicles. For example, it may cause other vehicles to slow down, change lanes, or cause congestion, thereby generating the first lane change traffic flow impact; the first lane change's own impact and the first lane change traffic flow impact are weighted and fused to generate the first lane change comprehensive impact, that is, a comprehensive lane change impact assessment, wherein the weight of the first lane change's own impact is greater than the weight of the first lane change traffic flow impact, that is, more emphasis is placed on the direct impact of the lane change behavior on the vehicle itself, rather than the impact on surrounding vehicles; based on the first lane change comprehensive impact, the initially generated first lane change prediction result is optimized, that is, the first lane change comprehensive impact is used as an influencing factor to adjust the first lane change prediction result; if the first lane change comprehensive impact shows that the lane change has a significant positive effect on the vehicle itself and has a small impact on traffic flow, the possibility of the lane change will be strengthened, and the lane change behavior is considered to be more reasonable.

[0026] Based on the manned driving lane change prediction result, a game decision network is used to make a coordinated lane change decision for multiple unmanned vehicles according to the state time series of the unmanned vehicles to generate an unmanned driving lane change decision.

[0027] In the game decision network, the unmanned vehicle will make its own optimal decision based on the actions and decisions of the manned vehicle. That is to say, when making a lane change decision, each unmanned vehicle will not only consider its own state, but also make the best overall decision based on the state sequence of other unmanned vehicles and the predicted behavior of manned vehicles. The manned lane change prediction results and the state sequence data of the unmanned vehicle are input into the game decision network as input data to make collaborative lane change decisions for multiple unmanned vehicles and generate unmanned lane change decisions.

[0028] Furthermore, the game decision network includes a first-layer decision network and a second-layer decision network; wherein the first-layer decision network is used to perform independent lane change impact analysis for each unmanned vehicle; and the second-layer decision network is used to make a balanced decision on the lane change impact based on the lane change impact index corresponding to each unmanned vehicle output by the first-layer decision network.

[0029] The game decision network consists of two layers, namely the first-layer decision network and the second-layer decision network. The first-layer decision network is responsible for independent lane change impact analysis of each unmanned vehicle. The second-layer decision network makes an overall equilibrium decision based on the lane change impact indicators of all unmanned vehicles on the basis of the first-layer decision network.

[0030] The game decision network performs independent lane change impact analysis through the first-layer decision network to generate lane change impact indicators for each driverless vehicle; then, the second-layer decision network makes global equilibrium decisions based on these indicators to ensure the coordination and optimal decision of lane change behaviors under the lane change requirements of multiple vehicles. This two-layer structure of the game decision network not only ensures the safety and efficiency of individual vehicles, but also improves the stability and safety of the entire traffic flow through global equilibrium decisions.

[0031] Furthermore, based on the manned driving lane change prediction result, a game decision network is used to make a coordinated lane change decision for multiple unmanned vehicles according to the state time sequence of the unmanned vehicle to generate an unmanned driving lane change decision, including:

[0032] The traffic flow is updated and fitted according to the manned lane change prediction result to generate a fitted mixed traffic flow; the first-layer decision network is used to take the fitted mixed traffic flow as a constraint to perform lane change impact analysis on the multiple unmanned vehicles respectively to generate multiple independent lane change impact indicators; based on the multiple independent lane change impact indicators, multiple first candidate lane change unmanned vehicles with independent lane change impact indicators greater than or equal to preset indicators are extracted; and the second-layer decision network is used to make a balanced decision on the lane change impact based on the multiple first candidate lane change unmanned vehicles to generate the unmanned lane change decision.

[0033] Specifically, the current traffic flow is updated and fitted based on the manned vehicle lane change prediction results. That is, according to the lane change behavior of manned vehicles, the simulation technology is used to infer the possible changes in the entire traffic flow, and then a mixed traffic flow is obtained. The mixed traffic flow includes both the behavior prediction of manned vehicles and the possible responses of unmanned vehicles. The fitted mixed traffic flow is input into the first-layer decision network as a constraint condition. At the same time, the state time series data of multiple unmanned vehicles are also required to be input. In the first-layer decision network, the lane change impact analysis is performed for each unmanned vehicle, including the safety, efficiency and traffic flow adaptability of the vehicle after performing different lane change actions. For each unmanned vehicle The independent lane change impact index is generated for each vehicle, and the independent lane change impact index comprehensively reflects the various impacts that may be caused by the vehicle after executing the lane change action; a preset index is set to measure the acceptability of the lane change impact, and according to the generated independent lane change impact index, multiple unmanned vehicles with impact indicators greater than or equal to the preset index are screened out as the first candidate lane change vehicles; the set of the first candidate lane change unmanned vehicles and their respective independent lane change impact indicators are input into the second-layer decision network, and a balanced decision on the lane change impact is made for the first candidate lane change unmanned vehicles, and an unmanned lane change decision is generated. The unmanned lane change decision includes specific action guidance such as whether each unmanned vehicle needs to change lanes, the timing of lane change, and the direction of lane change.

[0034] Furthermore, the second-layer decision network makes a balanced decision on the lane change impact according to the plurality of first candidate lane-changing unmanned vehicles to generate the unmanned driving lane change decision, including:

[0035] According to the independent lane change impact indicators corresponding to the multiple first candidate lane-changing unmanned driving vehicles, the multiple first candidate lane-changing unmanned driving vehicles are sorted in descending order to generate a candidate lane-changing vehicle sequence; the candidate lane-changing vehicle sequence is traversed to perform a coordinated lane change impact superposition and equilibrium analysis of the multiple first candidate lane-changing unmanned driving vehicles to generate the unmanned driving lane change decision.

[0036] Specifically, each candidate lane-changing vehicle is sorted in descending order according to its independent lane-changing impact index, which means that vehicles with larger lane-changing impact indexes have greater lane-changing priority. After sorting, these vehicles are arranged according to priority to form a candidate lane-changing vehicle sequence. Each candidate lane-changing vehicle in the sequence is traversed in turn, and the analysis begins with the vehicle with the largest lane-changing impact index. Whenever a candidate vehicle is analyzed, its lane-changing behavior is analyzed through a synergistic impact superposition analysis with the lane-changing behavior of the previously analyzed vehicle to ensure that the lane-changing behaviors of different vehicles can be coordinated with each other, and the lane-changing safety of other vehicles will not be reduced or traffic congestion will not be aggravated due to the lane change of a certain vehicle. After traversing all candidate lane-changing vehicles, an unmanned driving lane-changing decision is finally generated through a synergistic impact superposition analysis and balanced decision-making of all vehicles.

[0037] Furthermore, traversing the candidate lane-changing vehicle sequence to perform a coordinated lane-changing impact superposition and equilibrium analysis of multiple first candidate lane-changing unmanned driving vehicles to generate the unmanned driving lane-changing decision includes:

[0038] A first candidate lane-changing vehicle at the first position in the candidate lane-changing vehicle sequence is extracted, and lane-changing fitting and lane-changing impact analysis are performed based on the fitted mixed traffic flow to generate a first mixed traffic flow fitting update result and a first lane-changing impact index; a second candidate lane-changing vehicle adjacent to the first candidate lane-changing vehicle in the candidate lane-changing vehicle sequence is extracted, and lane-changing fitting and lane-changing impact analysis are performed based on the first mixed traffic flow fitting update result to generate a second mixed traffic flow fitting update result and a second lane-changing impact index; whether the second lane-changing impact index is greater than or equal to the first lane-changing impact index is determined, and if not, a third candidate lane-changing vehicle is continuously obtained, and the first mixed traffic flow fitting update result is returned to perform lane-changing fitting and lane-changing impact analysis to generate a third mixed traffic flow fitting update result and a third lane-changing impact index; the analysis is repeated until the candidate lane-changing vehicle sequence is traversed to generate a final mixed traffic flow fitting update result; and the unmanned driving lane-changing decision is generated based on the final mixed traffic flow fitting update result.

[0039] The first candidate lane-changing vehicle is extracted from the candidate lane-changing vehicle sequence. This is the vehicle with the largest lane-changing impact index and has the highest priority. Based on the fitting of mixed traffic flow, the first candidate lane-changing vehicle is subjected to lane-changing fitting and lane-changing impact analysis to simulate the behavior of the first candidate lane-changing vehicle after the lane change and its impact on the traffic flow. Through the lane-changing fitting analysis, the mixed traffic flow is updated to generate a new mixed traffic flow state, namely, the first mixed traffic flow fitting update result. The first mixed traffic flow fitting update result reflects the overall traffic flow change after the first candidate vehicle changes lanes. At the same time, the first lane-changing impact of the first candidate vehicle is also generated. The impact index is used to quantitatively evaluate the impact of the vehicle on the traffic flow and itself after changing lanes; the second candidate lane-changing vehicle adjacent to the first candidate lane-changing vehicle is extracted from the candidate lane-changing vehicle sequence; based on the first mixed traffic flow fitting update result, the lane change fitting and lane change impact analysis of the vehicle are performed, that is, the behavior and impact of the second candidate vehicle when changing lanes in the traffic flow after the first candidate vehicle has changed lanes are analyzed; through analysis, an updated traffic flow state is generated, that is, the second mixed traffic flow fitting update result, which reflects the traffic flow changes after the two candidate lane-changing vehicles change lanes. At the same time, A second lane-changing impact index for the second candidate vehicle is also generated to indicate the comprehensive impact of the second candidate vehicle after changing lanes. The second lane-changing impact index is compared with the first lane-changing impact index to determine whether the lane-changing behavior of the second candidate vehicle is relatively more reasonable. If the second lane-changing impact index is greater than or equal to the first lane-changing impact index, it indicates that the lane-changing behavior of the second candidate vehicle is reasonable, and the lane-changing behavior will be accepted and subsequent analysis will be performed. If the second lane-changing impact index is less than the first lane-changing impact index, it indicates that the lane-changing behavior may not be as reasonable as the lane-changing behavior of the previous vehicle, and the next candidate vehicle will be extracted from the sequence. The third candidate lane-changing vehicle (i.e., the third candidate lane-changing vehicle) is selected and the first mixed traffic flow fitting update result is returned to perform lane-changing fitting and impact analysis on the third vehicle. The above analysis process is repeated to process each vehicle in the candidate lane-changing vehicle sequence one by one. Each round of analysis is based on the mixed traffic flow fitting update result of the previous vehicle to ensure that the lane-changing behavior of subsequent vehicles is coordinated on the basis of considering the lane change of the previous vehicle. Through this progressive fitting and impact analysis, a complete and multiple-updated final mixed traffic flow fitting update result is finally generated, which reflects the overall traffic flow status of all candidate lane-changing vehicles after the lane change.

[0040] A first unmanned vehicle that needs to change lanes and first lane change constraint information are determined based on the unmanned driving lane change decision.

[0041] A complete unmanned driving lane change decision is generated through the game decision network. These decisions include lane change plans for each unmanned vehicle, including the timing of lane change, the direction of lane change, and the potential impact on other vehicles. From these generated lane change decisions, the specific vehicle that needs to perform the lane change operation is determined. At this time, according to the traffic flow conditions and the lane change requirements of each vehicle, the first unmanned vehicle, that is, the first unmanned vehicle that currently needs to change lanes, will be selected first, and a set of first lane change constraint information will be determined for the first unmanned vehicle. The first lane change constraint information refers to the conditions or restrictions that the first unmanned vehicle needs to follow when performing a lane change to ensure the safety and effectiveness of the lane change; the first lane change constraint information may include time constraints, speed constraints, lane constraints, distance constraints, etc.

[0042] The first unmanned vehicle is monitored and reminded to change lanes according to the first lane change constraint information.

[0043] After the first unmanned vehicle and its lane change constraint information are determined, the lane change process of the vehicle will be monitored and relevant reminders will be issued to ensure the smooth progress of the lane change.

[0044] Furthermore, performing lane change monitoring and reminder on the first unmanned vehicle according to the first lane change constraint information includes:

[0045] The driving state of the first unmanned vehicle is monitored to obtain first monitoring data; whether the first unmanned vehicle satisfies the first lane change constraint information is determined based on the first monitoring data; if so, whether the current road section is suitable for lane change is determined based on the distance between the front and rear vehicles of the first unmanned vehicle; and a lane change command is issued to the first unmanned vehicle based on the determination result.

[0046] Specifically, the driving state of the first unmanned vehicle is monitored in real time, and first monitoring data about the current driving condition of the vehicle is collected, including speed, position, driving direction, distance from surrounding vehicles, etc.; the first monitoring data is analyzed to determine whether the current state of the vehicle meets the first lane change constraint information defined previously; if the first unmanned vehicle meets the first lane change constraint information, it is necessary to further determine whether the current section is suitable for lane change operation based on the distance between the front and rear vehicles; if the distance of the front vehicle is too close, it may not be suitable for lane change to avoid collision or traffic jam; if the distance of the rear vehicle is far, it is suitable for lane change; but if the distance of the rear vehicle is close, lane change may cause the rear vehicle to slow down or cause dangerous situation; based on the analysis results, it is decided whether to issue a lane change command. If the current section is suitable for lane change, a clear lane change command is issued to the first unmanned vehicle, instructing the vehicle to start the lane change operation; if the current section is not suitable for lane change, the lane change will be delayed, or a reminder command will be issued, suggesting that the vehicle wait for a more appropriate time to change lanes.

[0047] In summary, the embodiments of the present application have at least the following technical effects:

[0048] First, the traffic monitoring module in the preset road area is connected to monitor the traffic flow and generate real-time traffic flow, wherein the real-time traffic flow includes the state sequence of the unmanned vehicle and the state sequence of the manned vehicle. Next, the lane change behavior is predicted based on the state sequence of the manned vehicle to generate the manned lane change prediction result. Then, based on the manned lane change prediction result, the game decision network is used to make a coordinated lane change decision for multiple unmanned vehicles according to the state sequence of the unmanned vehicle to generate an unmanned lane change decision. Further, the first unmanned vehicle and the first lane change constraint information that need to change lanes are determined based on the unmanned lane change decision. Finally, the first unmanned vehicle is monitored and reminded to change lanes according to the first lane change constraint information. The technical problem that it is difficult for unmanned vehicles to effectively respond to the lane change behavior of manned vehicles in mixed traffic flows in the prior art is solved, and the technical effect of improving the accuracy of unmanned vehicle lane change decisions and traffic safety is achieved.

[0049] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0050] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0051] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A coordinated lane change decision method for mixed traffic flow with no people or vehicles, characterized in that: The method comprises: Connecting to a traffic monitoring module in a preset road area to monitor traffic flow and generate real-time traffic flow, wherein the real-time traffic flow includes a state sequence of unmanned vehicles and a state sequence of manned vehicles; Performing lane change behavior prediction based on the state time series of the manned vehicle to generate a manned lane change prediction result; Based on the manned driving lane change prediction result, a game decision network is used to make a coordinated lane change decision for multiple unmanned vehicles according to the state time sequence of the unmanned vehicle to generate an unmanned driving lane change decision; Determining a first unmanned vehicle that needs to change lanes and first lane change constraint information based on the unmanned driving lane change decision; The first unmanned vehicle is monitored and reminded to change lanes according to the first lane change constraint information.

2. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 1, characterized in that: The lane change behavior is predicted based on the state time sequence of the manned vehicle to generate a manned lane change prediction result, including: Collecting a historical vehicle lane change data set, wherein the historical vehicle lane change data set includes historical driving behavior, historical mixed traffic flow characteristics and corresponding lane change behavior characteristics; Establishing a lane change behavior prediction model based on the historical driving behavior and the historical mixed traffic characteristics; Analyzing the state time series of the manned vehicle by using the lane change behavior prediction model to generate a first lane change prediction result; A lane change impact analysis is performed based on the state time series of the manned vehicle, the first lane change prediction result is optimized, and the manned lane change prediction result is generated.

3. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 2, characterized in that: Performing a lane change impact analysis based on the state time sequence of the manned vehicle, optimizing the first lane change prediction result, and generating the manned vehicle lane change prediction result, includes: Extracting a first manned vehicle according to the state time sequence of the manned vehicle, and establishing a first candidate lane change behavior according to the lane where the first manned vehicle is located; Performing lane change fitting with the first candidate lane change behavior to obtain a first lane change impact of the first manned vehicle after the lane change, and performing traffic flow impact analysis after the lane change in combination with the state time sequence of the manned vehicle to generate a first lane change traffic flow impact; weighted fusion of the first lane change self-impact and the first lane change traffic flow impact to generate a first lane change comprehensive impact, wherein the weight of the first lane change self-impact is greater than the weight of the first lane change traffic flow impact; The first lane change prediction result is optimized based on the first lane change comprehensive impact to generate the manned lane change prediction result.

4. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 1, characterized in that: The game decision network includes a first-layer decision network and a second-layer decision network; Among them, the first-layer decision network is used to perform independent lane change impact analysis for each unmanned vehicle; the second-layer decision network is used to make a balanced decision on the lane change impact according to the lane change impact index corresponding to each unmanned vehicle output by the first-layer decision network.

5. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 4, characterized in that: Based on the manned driving lane change prediction result, a game decision network is used to make a coordinated lane change decision for multiple unmanned vehicles according to the state time sequence of the unmanned vehicle to generate an unmanned driving lane change decision, including: Performing an update fit of the traffic flow according to the manned driving lane change prediction result to generate a fitted mixed traffic flow; Using the first-layer decision network and taking the fitted mixed traffic flow as a constraint, respectively performing lane change impact analysis on the multiple unmanned vehicles to generate multiple independent lane change impact indicators; Extracting a plurality of first candidate lane-changing unmanned vehicles having independent lane-changing influence indicators greater than or equal to a preset indicator according to the plurality of independent lane-changing influence indicators; The second-layer decision network makes a balanced decision on the impact of lane change based on the multiple first candidate lane-changing unmanned vehicles, and generates the unmanned driving lane change decision.

6. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 5, characterized in that: The second-layer decision network makes a balanced decision on the lane change impact according to the plurality of first candidate lane-changing unmanned vehicles, and generates the unmanned driving lane change decision, including: According to the independent lane change impact indicators corresponding to the multiple first lane change candidate unmanned driving vehicles, the multiple first lane change candidate unmanned driving vehicles are sorted in descending order to generate a candidate lane change vehicle sequence; The candidate lane-changing vehicle sequence is traversed to perform a coordinated lane-changing impact superposition and equilibrium analysis of multiple first candidate lane-changing unmanned driving vehicles to generate the unmanned driving lane-changing decision.

7. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 6, characterized in that: Traversing the candidate lane-changing vehicle sequence to perform a coordinated lane-changing impact superposition and equilibrium analysis of multiple first candidate lane-changing unmanned driving vehicles to generate the unmanned driving lane-changing decision, including: Extracting a first candidate lane-changing vehicle that is at the top of the candidate lane-changing vehicle sequence, performing lane-changing fitting and lane-changing impact analysis based on the fitted mixed traffic flow, and generating a first mixed traffic flow fitting update result and a first lane-changing impact index; Extracting a second candidate lane-changing vehicle adjacent to the first candidate lane-changing vehicle from the candidate lane-changing vehicle sequence, performing lane-changing fitting and lane-changing impact analysis based on the first mixed traffic flow fitting update result, and generating a second mixed traffic flow fitting update result and a second lane-changing impact index; Determine whether the second lane change impact index is greater than or equal to the first lane change impact index; if not, continue to obtain a third candidate lane change vehicle, and return to the first mixed traffic flow fitting update result to perform lane change fitting and lane change impact analysis, and generate a third mixed traffic flow fitting update result and a third lane change impact index; Repeat the analysis until the candidate lane-changing vehicle sequence is traversed to generate a final mixed traffic flow fitting update result; The unmanned driving lane change decision is generated based on the final mixed traffic flow fitting update result.

8. The coordinated lane change decision method for mixed traffic flow with no people or people as claimed in claim 1, characterized in that: Performing lane change monitoring and reminder on the first unmanned vehicle according to the first lane change constraint information includes: Monitoring the driving state of the first unmanned vehicle to obtain first monitoring data; determining whether the first unmanned vehicle satisfies the first lane change constraint information according to the first monitoring data; If yes, determining whether the current road section is suitable for lane change based on the distance between the front and rear vehicles of the first unmanned vehicle; A lane change instruction is issued to the first unmanned vehicle according to the determination result.

Images