V2X-based cooperative lane change path planning method on highways

Abstract

The present invention discloses a collaborative lane change path planning method on a highway based on V2X, comprising the following steps: Step 1: When a lane-changing vehicle plans to change lanes, data is collected; Step 2: The trajectory points of adjacent vehicles within the next 5-10 seconds are output; Step 3: A longitudinal displacement-time curve function is constructed to obtain a longitudinal path plan; Step 4: A transverse plan is performed based on the longitudinal plan, and a cost function of the transverse path is calculated to obtain a transverse path plan; Step 5: The optimal solution for the lane-changing vehicle path plan is obtained; Step 6: The lane-changing vehicle sends its lane-changing intention and planned path information to adjacent vehicles; Step 7: The adjacent vehicles obtain the longitudinal path plan; Step 8: The calculated minimum cost is compared with a pre-set threshold, and when the cost is less than the threshold, the lane-changing vehicle's lane-changing request is approved. The present invention is implemented through an algorithm with low time complexity, which can provide a relatively fast response and better lane-changing path planning.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle autonomous driving, and in particular relates to a path planning method in a V2X-based cooperative lane change scenario on a highway. Background Art

[0002] Most existing vehicle lane-changing algorithms are based on the vehicle's analysis of road conditions and then planning the path for lane change based on information such as obstacles.

[0003] For example, patent application 202210960590.7 discloses a path selection method, device and electronic device for an autonomous driving vehicle, the method including: obtaining the current positioning data of the autonomous driving vehicle and the corresponding path planning results, the path planning results including multiple alternative lane-level paths; based on the current positioning data of the autonomous driving vehicle, obtaining the predicted positioning data of the autonomous driving vehicle and obstacles; determining the driving cost of each alternative lane-level path according to the predicted positioning data of the autonomous driving vehicle and obstacles, the driving cost is used to characterize the congestion level of the alternative lane-level path; determining the target lane-level path according to the driving cost of each alternative lane-level path.

[0004] Another example is a path planning method, device, storage medium, chip, and vehicle disclosed in patent application 202210834671.2, which relates to the field of autonomous driving. Image information of a target area to be detected is acquired through an image acquisition device; obstacle parameters of the obstacle and lane line parameters of the lane line are acquired based on the image information; a target pitch angle of the vehicle is determined based on the obstacle parameters and lane line parameters; and a first target lane where the obstacle is located is determined from the lane line based on the obstacle parameters, lane line parameters, and target pitch angle, and path planning is performed based on the first target lane.

[0005] However, existing technologies for path planning largely rely solely on traffic information collected by the vehicle's sensors, neglecting coordination within traffic conditions. Furthermore, obstacle recognition algorithms, which rely on information collected by lidar or cameras, often have high time complexity, resulting in long execution times. Lane changes on highways require rapid algorithmic response. Existing technologies do not leverage existing historical trajectory information to predict future trajectories, instead directly planning based on current conditions, potentially introducing errors in the planned route. Summary of the Invention

[0006] Based on this, the primary purpose of the present invention is to provide a V2X-based collaborative lane change path planning method on highways, which solves the problems of vehicle lane change decision-making and trajectory planning in collaborative lane change scenarios on highways for intelligent connected vehicles.

[0007] Another object of the present invention is to provide a V2X-based collaborative lane change path planning method on highways. The method is implemented through an algorithm with low time complexity, which can give a relatively fast response and better plan the lane change path.

[0008] To achieve the above object, the technical solution of the present invention is:

[0009] A V2X-based cooperative lane change path planning method on a highway, characterized by comprising the following steps:

[0010] Step 1: When a lane-changing vehicle plans to change lanes, it collects the current position, speed, acceleration and other data of the vehicle and surrounding vehicles in the recent time period.

[0011] The following sub-steps are included:

[0012] 1.1. When a lane-changing vehicle plans to change lanes, it collects its own motion data through its own vehicle sensors to obtain the position, speed, acceleration, and other data of the lane-changing vehicle at the current time t0.

[0013] 1.2. The lane-changing vehicle receives the V2X broadcast data from surrounding vehicles during the time period [t0-3, t0] through the OBU (On Board Unit), which includes the current position, speed, acceleration, and other information of the surrounding vehicles.

[0014] Step 2: The lane-changing vehicle encodes the collected information and passes it into the dynamic graph neural network model, which calculates and outputs the possible trajectory points of the adjacent vehicle in the next 5-10 seconds.

[0015] The following sub-steps are included:

[0016] 2.1. Divide the collected data within the time period into historical trajectory point data with an interval of 0.1s, and construct a graph data for each moment.

[0017] The method of constructing graph data is as follows: there are two types of nodes. One is to regard the lane line vector as a node, and the other is to regard the vehicle trajectory vector within every 0.1 second as a node. The node representation is:

[0018]

[0019] in, Represents the coordinates of the starting and ending nodes of the vector, type is the node type, and attr is the attribute information of the node. When the node type is a lane line, the attribute information is the road speed limit. In the constructed graph data, all nodes are fully connected.

[0020] 2.2. Input the constructed graph data into the state graph neural network model. The model is designed based on the encoder-decoder structure. The encoder uses a graph convolutional neural network model to learn the structural information in the graph data, and the decoder uses an LSTM network to learn the temporal information of the historical trajectory.

[0021] 2.3. The model outputs the predicted trajectory points of adjacent vehicles in the next 5-10 seconds.

[0022] Step 3: Construct a longitudinal displacement-time curve function, and solve the longitudinal displacement-time curve function by calculating the cost function of the longitudinal path to obtain the longitudinal path planning;

[0023] This step is implemented by the following sub-steps:

[0024] 3.1. Lane change completion time t f Sampling is performed in a certain range (t∈[4.5s, 7.0s], δ=0.1s, the expected lane change time range is 4.5s to 7.0s, and the sampling step is 0.1s), and the termination conditions of the lane-changing vehicle longitudinal displacement-time curve are obtained: time, velocity, acceleration where t f Indicates the lane change time. Express the speed and assume that with the future t f The speeds of adjacent vehicles are equal at all times. Indicates acceleration.

[0025] 3.2. Based on the Frenet coordinate system, the longitudinal displacement-time curve of the lane-changing vehicle is expressed as follows:

[0026] s(t)=[a0,a1,...,a5][1,t,t 2 ,...t 5 ] T

[0027] Among them, a i is the parameter of the polynomial, and t represents time.

[0028] The termination condition and the position, velocity, and acceleration information of the lane-changing vehicle at the current time t0 obtained in step 1 are introduced as constraints into the quintic polynomial above, and the quintic polynomial s(t) is converted into a function form that is only related to a5:

[0029]

[0030] Among them, f1(t) and f2(t) are polynomials about t, t∈[t0, t f ]. t0 is the time when the lane-changing vehicle decides to change lanes, tf is the lane change completion time of the lane-changing vehicle, s0, is the position, velocity, and acceleration of the lane-changing vehicle at time t0, t f The speed and acceleration of the vehicle changing lanes at any moment.

[0031] 3.3, by calculating t f The positions of adjacent vehicles at the moment are satisfied s(t f ) The range s(t) that does not conflict with the projection area of ​​adjacent vehicles f )∈[s fmin ,s fmax ], thus obtaining the value range of a5:

[0032]

[0033] Calculate t from the longitudinal displacement curve of the adjacent vehicle f At the position of adjacent vehicles at the moment, calculate the situation where the lane-changing vehicle chooses to overtake or follow the vehicle, and satisfies s(t f ) The range s(t) that does not conflict with the projection area of ​​adjacent vehicles f )∈[s fmin ,s fmax ], thus obtaining the value range of a5:

[0034] 3.4. Solve the minimum value of the cost function by the value range of a5:

[0035]

[0036] Among them, w i is a pre-set weight; when a5 is determined, in [t0, t f ] range is determined, and finally, the longitudinal displacement-time curve is taken as the longitudinal path.

[0037] Step 4: Perform lateral planning based on the longitudinal planning, calculate the cost function of the lateral path, and obtain the lateral path planning;

[0038] This step is implemented by the following sub-steps:

[0039] 4.1, by [t0, t f ] range (sampling step δ = 0.05s), obtaining a discretized point set representing the set of points that a lane-changing vehicle may reach at time t. By constraining the upper and lower boundaries of each point (the upper bound is the position of the adjacent vehicle, and the lower bound is the position of the lane-changing vehicle), the range of the point set is narrowed.

[0040] 4.2. Determine the d(t) curve at this point by calculating the cost function:

[0041]

[0042] Among them, q i is the set weight, d ir Represents t i Reference curve at time d iu and d il Represents d i The upper and lower boundaries of the discrete point curve; δ d is a smaller value set in advance. are the upper and lower boundary conditions set in advance.

[0043] Finally, according to the minimum value of the result of the cost function, the discretized d(t) curve is obtained as the lateral path.

[0044] Step 5: Obtain the optimal solution for lane-changing vehicle path planning by calculating the sum of the cost function of the longitudinal path and the cost function of the lateral path;

[0045] This step is implemented by the following sub-steps:

[0046] 5.1. By adding the cost functions constructed in step 3 and step 4, the current sample t can be obtained f Total cost function formula:

[0047]

[0048] Calculate the sum of the cost function of the longitudinal path and the cost function of the transverse path;

[0049] 5.2, repeat steps 3, 4 and 5.1 multiple times (the number depends on the number of samples, for each sample t f All must be calculated), each time the current sampling t f The total cost function is calculated.

[0050] 5.3. Finally, by comparing the total cost of each time, the minimum cost is obtained, that is, the optimal solution for lane-changing vehicle path planning is obtained.

[0051] Step 6: The lane-changing vehicle sends its lane-changing intention and planned path information to the adjacent vehicle;

[0052] The lane-changing vehicle sends its lane-changing intention to the neighboring vehicles in the target lane via its OBU. The neighboring vehicles' OBUs receive the data packet sent by the lane-changing vehicle and parse it to obtain the lane-changing intention, which includes the short-term trajectory plan obtained in step 5.

[0053] Step 7. Construct a longitudinal displacement-time curve function for adjacent vehicles and solve the longitudinal displacement-time curve function by calculating the cost function of the longitudinal path to obtain the longitudinal path plan;

[0054] This step is implemented by the following sub-steps:

[0055] 7.1. Assuming that the longitudinal displacement of adjacent vehicles can be expressed as a quintic polynomial, the longitudinal displacement-time curve of adjacent vehicles is expressed as follows based on the Frenet coordinate system:

[0056] s(t)=[a0,a1,...,a5][1,t,t 2 ,...,t 5 ] T

[0057] Among them, a i is the parameter of the polynomial, and t represents time.

[0058] 7.2. Set the lane-changing time t of the lane-changing vehicle f The speed and acceleration of the vehicle are used as the termination conditions, and the position, speed, and acceleration information of the adjacent vehicle at the current time t0 are used as constraints to be introduced into the quintic polynomial assumed in 7.1. The quintic polynomial s(t) can be converted into a function form that is only related to a5:

[0059]

[0060] Among them, f1(t) and f2(t) are polynomials about t, t∈[t0, t f ], t0 is the time when the adjacent vehicle decides to change lanes, t f is the lane change completion time of the lane-changing vehicle, s0, is the position, velocity, and acceleration of the adjacent vehicle at time t0, t f The speed and acceleration of the vehicle changing lanes at any moment.

[0061] 7.3, taking the maximum comfortable acceleration as the upper bound and the minimum vehicle speed as the lower bound, the end time is sampled s(t f )∈[s fmin , s fmax ], thus obtaining the value range of a5:

[0062]

[0063] 7.4. Calculate the cost function:

[0064]

[0065]

[0066] And in [a 5min ,a 5max ] area, we can find the minimum value in [t0,t f ] range, the longitudinal path planning of the adjacent vehicle is obtained. Finally, the longitudinal displacement-time curve is used as the longitudinal path of the adjacent vehicle.

[0067] Step 8. Compare the calculated minimum cost with a preset threshold. When the cost is less than the threshold, the lane change request of the lane-changing vehicle is approved.

[0068] By comparing the optimal value of the adjacent vehicle's cost function in step 7 with a preset threshold, a decision can be made as to whether to allow the lane-changing vehicle to change lanes. If the cost of the planned longitudinal path is less than the predetermined threshold, the lane-changing vehicle's lane-change request is approved, and the adjacent vehicle will then follow the planned longitudinal path (since the adjacent vehicle is not involved in a lane change, no lateral path planning is required for the adjacent vehicle). Otherwise, the lane-changing vehicle's lane-change request is deemed unreasonable, rejected, and the vehicle continues driving in its current state.

[0069] The present invention is implemented using an algorithm. Since the algorithm has a very low time complexity, it can respond quickly in the lane change scenario on a highway. It also utilizes information exchange based on intelligent networking, takes into account information such as the driving intentions of the two vehicles, and can better plan the lane change path. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flow chart of trajectory planning implemented by the present invention.

[0071] Figure 2 This is a schematic diagram of the projected area of ​​the vehicle that does not conflict after EV1 changes lanes according to the present invention.

[0072] Figure 3 This is a schematic diagram of the non-conflicting projection area when EV1 chooses to overtake after changing lanes according to the present invention.

[0073] Figure 4 This is a flowchart of EV2 judging whether to agree with EV1 to change lanes implemented in the present invention. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0075] The method implemented in this invention is a collaborative lane-change planning algorithm for intelligent connected vehicles on highways, optimizing lane-change timing, location, and speed. This invention encompasses only the short-term path planning for EV1's lane change process after it decides to change lanes, and EV2's decision to approve EV1's lane-change request based on its own driving state (for simplicity, the lane-changing vehicle will be referred to as EV1, and the vehicle in the adjacent lane, i.e., the adjacent vehicle, will be referred to as EV2). Figure 1 As shown, the method includes the following steps:

[0076] Step 1: When EV1 plans to change lanes, it collects the current position, speed, acceleration and other data of the vehicle and surrounding vehicles in the recent time period, including the following sub-steps:

[0077] 1.1. When EV1 plans to change lanes, it collects its own motion data through its own vehicle sensors and obtains the position, speed, acceleration and other data of the lane-changing vehicle EV1 at the current time t0.

[0078] 1.2. EV1 obtains the V2X broadcast data of surrounding vehicles in the time period [t0-3, t0] through the OBU (On Board Unit), which includes the current position, speed, acceleration and other information of surrounding vehicles.

[0079] Step 2: EV1 encodes the collected information and passes it into the dynamic graph neural network model, which calculates and outputs the possible trajectory points of EV2 in the next 7 seconds.

[0080] The following sub-steps are included:

[0081] 2.1. Divide the collected data into historical trajectory point data at intervals of 0.1s, and construct a graph for each moment. The construction method is as follows: There are two types of nodes: one that considers the lane line vector as a node, and the other that considers the vehicle trajectory vector within every 0.1 second as a node. The node representation is:

[0082]

[0083] in, and Represents the coordinates of the starting and ending nodes of the vector, type is the node type, and attr is the attribute information of the node. When the node type is a lane line, the attribute information is the road speed limit. In the constructed graph data, all nodes are fully connected.

[0084] 2.2. Input the constructed graph data into the state graph neural network model. The model is designed based on the encoder-decoder structure. The encoder uses a graph convolutional neural network model to learn the structural information in the graph data, and the decoder uses an LSTM network to learn the temporal information of the historical trajectory.

[0085] 2.3. The model outputs the predicted trajectory points of EV2 in the next 7 seconds.

[0086] Step 3: Construct a longitudinal displacement-time curve function and solve it by calculating the cost function of the longitudinal path to obtain the longitudinal path planning. This step is implemented by the following sub-steps:

[0087] 3.1. Time t at which EV1 completes lane change f Sampling is performed in a certain range (t∈[4.5s,7.0s], δ=0.1s, the expected lane change time range is 4.5s to 7.0s, and the sampling step is 0.1s), and the termination conditions of the EV1 longitudinal displacement-time curve are obtained: time, velocity, acceleration where t f Indicates the lane change time. Express the speed and assume that with the future t f At this moment, the speed of EV2 is equal. Indicates acceleration.

[0088] 3.2. Based on the Frenet coordinate system, the longitudinal displacement-time curve of EV1 is expressed as follows:

[0089] s(t)=[a0,a1,...,a5][1,t,t 2 ,...,t 5 ] T

[0090] Among them, a i is the parameter of the polynomial, and t represents time.

[0091] Substitute the above termination conditions and the position, velocity, and acceleration information of EV1 at the current time t0 obtained in step 1 as constraints into the assumed quintic polynomial, and convert the quintic polynomial s(t) into a function form that is only related to a5:

[0092]

[0093] Among them, f1(t) and f2(t) are polynomials about t, t∈[t0, t f ], t0 is the time when EV1 decides to change lanes, t f is the time when EV1 completes lane change, s0, is the position, velocity, and acceleration of EV1 at time t0, t f The speed and acceleration of EV1 at this moment.

[0094] 3.3, by calculating t f The position of EV2 at the moment satisfies s(t f ) does not conflict with the EV2 projection area s(t f )∈[s fmin , s fmax ], thus obtaining the value range of a5:

[0095]

[0096] The details of this sub-step are as follows:

[0097] According to step 2, the predicted driving trajectory of EV2 before and after EV1 changes lanes can be obtained. The longitudinal displacement curve of EV2 can be expressed as , and project the longitudinal displacement-time curve of EV2 into the same Frenet coordinate system as EV1. Since the longitudinal displacement-time curves of EV1 and EV2 are mapped into the same coordinate system, it is only necessary to ensure that at the time t f Afterwards, the longitudinal displacement-time curves of EV1 and EV2 do not overlap, which ensures that the planned longitudinal displacement will not cause a collision between the two vehicles.

[0098] like Figure 2-Figure 3 As shown, the box area represents the longitudinal displacement-time curve of EV2, where the upper line is the longitudinal displacement-time curve of the front of the vehicle, and the lower line is the longitudinal displacement-time curve of the rear of the vehicle. The curve represents the longitudinal displacement-time curve of EV1, where the upper line is the non-conflict range s(t f ) is the upper bound, and the lower line is the non-conflict range s(t f ) is the lower bound of . Figure 2 This represents the situation where EV1 chooses to follow the car after changing lanes. At this time, the upper limit of the non-conflict range is determined by the projection area of ​​EV2, and the lower limit can be set manually in advance. Figure 3 This represents the situation where EV1 chooses to overtake after changing lanes. At this time, the lower limit of the non-conflict range is determined by the projection area of ​​EV2, and the upper limit can be set manually in advance.

[0099] Calculate t from the longitudinal displacement curve of EV2 f At the moment of EV2's location, calculate the situation where EV1 chooses to overtake or follow the car, satisfying s(t f ) does not conflict with the EV2 projection area s(t f )∈[s fmin , s fmax], thus obtaining the value range of a5.

[0100] 3.4. Solve the minimum value of the cost function by the value range of a5:

[0101]

[0102] Among them, w i is a pre-set weight. When a5 is determined, in [t0, t f ]The s(t) within the range is determined.

[0103] Finally, the longitudinal displacement-time curve is taken as the longitudinal path.

[0104] Step 4: Perform lateral planning based on the longitudinal planning, calculate the cost function of the lateral path, and obtain the lateral path planning. This step is implemented by the following sub-steps:

[0105] 4.1, by [t0, t f ] range (sampling step δ = 0.05s) to obtain a discretized point set, which represents the set of points that EV1 may reach at time t. By constraining the upper and lower boundaries of each point (the upper bound is the position of EV2, and the lower bound is the position of EV1), the range of the point set is narrowed.

[0106] 4.2. Determine the d(t) curve at this point by calculating the cost function:

[0107]

[0108] Among them, q i is the set weight, d ir Represents t i Reference curve at time d iu and d il Represents d i The upper and lower boundaries of the discrete point curve. d A smaller value that is set in advance. are the upper and lower boundary conditions set in advance.

[0109] Finally, according to the minimum value of the result of the cost function, the discretized d(t) curve is obtained as the lateral path.

[0110] Step 5: Obtain the optimal solution for EV1 path planning by calculating the sum of the cost function of the longitudinal path and the cost function of the lateral path. This step is implemented by the following sub-steps:

[0111] 5.1. By adding the cost functions constructed in step 3 and step 4, the current sample t can be obtained f Total cost function formula:

[0112]

[0113] Calculate the sum of the cost function of the longitudinal path and the cost function of the transverse path;

[0114] 5.2, repeat steps 3, 4 and 5.1 multiple times (the number depends on the number of samples, for each sample t f All must be calculated), each time the current sampling t f The total cost function is calculated.

[0115] 5.3. Finally, by comparing the total cost of each time, the minimum cost is obtained, that is, the optimal solution for EV1 path planning is obtained.

[0116] Step 6: EV1 sends its lane-changing intention and planned path information to EV2.

[0117] EV1 sends its lane change intention to EV2, the vehicle in the target lane, via its onboard unit (OBU). EV2's OBU receives the data packet sent by EV1 and parses it to determine EV1's lane change intention, which includes EV1's short-term trajectory plan obtained in step 5.

[0118] Step 7.EV2 constructs the longitudinal displacement-time curve function and solves it by calculating the cost function of the longitudinal path to obtain the longitudinal path plan. This step is implemented by the following substeps:

[0119] 7.1. Based on the Frenet coordinate system, the longitudinal displacement-time curve of EV2 is expressed as follows:

[0120] s(t)=[a0,a1,...,a5][1,t,t 2 ,...,t 5 ] T

[0121] Among them, a i is the parameter of the polynomial, and t represents time.

[0122] 7.2. Set EV1’s lane-changing time t f By bringing the velocity and acceleration of EV2 as the termination conditions and the position, velocity, and acceleration of EV2 at the current time t0 as the constraints into the quintic polynomial assumed in 7.1, the quintic polynomial s(t) can be converted into a function form that is only related to a5:

[0123]

[0124] Among them, f1(t) and f2(t) are polynomials about t, t∈[t0, tf ]. t0 is the time when EV2 decides to change lanes, t f is the time when EV1 completes lane change, s0, is the position, velocity, and acceleration of EV2 at time t0, t f The speed and acceleration of EV1 at this moment.

[0125] 7.3, taking the maximum comfortable acceleration as the upper bound and the minimum vehicle speed as the lower bound, the end time is sampled s(t f )∈[s fmin , s fmax ], thus obtaining the value range of a5:

[0126]

[0127] 7.4. Calculate the cost function:

[0128]

[0129]

[0130] And in [a 5min ,a 5max ] area, we can find the minimum value in [t0, t f ] range, and the longitudinal path planning of EV2 is obtained. Finally, the longitudinal displacement-time curve is used as the longitudinal path of EV2.

[0131] Step 8. Compare the calculated minimum cost with a preset threshold. When the cost is less than the threshold, the lane change request of EV1 is approved.

[0132] By comparing the optimal value of EV2's cost function in step 7 with a preset threshold, it can be decided whether to allow the lane-changing vehicle to change lanes. If the cost of the planned longitudinal path is less than the predetermined threshold, the lane-changing request of the lane-changing vehicle EV1 can be approved, and EV2 will then travel along the planned longitudinal path (since EV2 does not involve lane changes, there is no need to plan a transverse path for EV2). Otherwise, EV1's lane-changing request is considered unreasonable, and the request is rejected, and the vehicle continues to drive according to the current state. The process of EV2 judging whether to approve EV1's lane change is as follows: Figure 4 shown.

[0133] In terms of trajectory planning, it can be done by using reinforcement learning, particle swarm learning, etc. This is an existing technology and will not be described in detail here.

[0134] Therefore, the time complexity of the algorithm of the present invention is very low. In the lane-changing scenario on a highway, the algorithm of this solution can respond relatively quickly, and utilizes the information exchange of intelligent networking, taking into account information such as the driving intentions of the two vehicles, and can better plan the lane-changing path.

[0135] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A V2X-based cooperative lane change path planning method on a highway, characterized by The steps include: Step 1: When a lane-changing vehicle plans to change lanes, it collects the current position, speed, and acceleration data of the vehicle and surrounding vehicles in the recent time period; Step 2: The lane-changing vehicle encodes the collected information and passes it into the dynamic graph neural network model, which calculates and outputs the trajectory points of the adjacent vehicle within the next 5-10 seconds. Step 3: Construct a longitudinal displacement-time curve function, and solve the longitudinal displacement-time curve function by calculating the cost function of the longitudinal path to obtain the longitudinal path planning; This is achieved by the following sub-steps: 3.

1. Lane change completion time t f Sampling is performed in a certain range, t∈[4.5s, 7.0s], δ=0.1s, the expected lane change time range is 4.5s to 7.0s, the sampling step is 0.1s, and the termination condition of the lane change vehicle longitudinal displacement-time curve is obtained: time, speed, acceleration (t f ,s″ f ,s″ f ), where t f Indicates the lane change time. Express the speed and assume that with the future t f The speeds of adjacent vehicles are equal at any given moment. represents acceleration; 3.

2. Based on the Frenet coordinate system, the longitudinal displacement-time curve of the lane-changing vehicle is expressed as follows: s(t)=[a0,a1,...,a5][1,t,t 2 ,...,t 5 ] T Among them, a i is the parameter of the polynomial, t represents time; Substitute the above termination condition and the position, velocity, and acceleration information of the lane-changing vehicle at the current time t0 obtained in step 1 into the above formula as constraints, and convert the quintic polynomial s(t) into a function form that is only related to a5: Among them, f1(t) and f2(t) are polynomials about t, t∈[t0, t f ], t0 is the time when the lane-changing vehicle decides to change lanes, t f is the lane change completion time of the lane-changing vehicle, s0, is the position, velocity, and acceleration of the lane-changing vehicle at time t0, t f The speed and acceleration of the vehicle changing lanes at any moment; 3.3, by calculating t f The positions of adjacent vehicles at the moment are satisfied s(t f ) The range s(t) that does not conflict with the projection area of ​​adjacent vehicles f )∈[s fmin , s fmax ], thus obtaining the value range of a5: Calculate t from the longitudinal displacement curve of the adjacent vehicle f At the position of adjacent vehicles at the moment, calculate the situation where the lane-changing vehicle chooses to overtake or follow the vehicle, and satisfies s(t f ) The range s(t) that does not conflict with the projection area of ​​adjacent vehicles f )∈[s fmin , s fmax ], thus obtaining the value range of a5: 3.

4. Solve the minimum value of the cost function by the value range of a5: Among them, w i is a pre-set weight; when a5 is determined, in [t0,t f ] is determined; finally, the longitudinal displacement-time curve is used as the longitudinal path; Step 4: Perform lateral planning based on the longitudinal planning, calculate the cost function of the lateral path, and obtain the lateral path planning; This step is implemented by the following sub-steps: 4.1, by [t0, t f ] range of t is sampled with a sampling step of δ = 0.05s to obtain a discretized point set, which represents the set of points that the lane-changing vehicle may reach at time t; the range of the point set is narrowed by constraining the upper and lower boundaries of each point, where the upper bound is the position of the adjacent vehicle and the lower bound is the position of the lane-changing vehicle; 4.

2. Determine the d(t) curve at this point by calculating the cost function: Among them, q i is the set weight, d ir Represents t i Reference curve at time d iu and d il Represents d i The upper and lower boundaries of the discrete point curve, δ d is a smaller value set in advance. are the upper and lower boundary conditions set in advance; Finally, according to the minimum value of the result of the cost function, the discretized d(t) curve is obtained as the lateral path; Step 5: Obtain the optimal solution for lane-changing vehicle path planning by calculating the sum of the cost function of the longitudinal path and the cost function of the lateral path; Step 6: The lane-changing vehicle sends its lane-changing intention and planned path information to the adjacent vehicle; Step 7. Construct a longitudinal displacement-time curve function for adjacent vehicles and solve the longitudinal displacement-time curve function by calculating the cost function of the longitudinal path to obtain the longitudinal path plan; Step 8. Compare the calculated minimum cost with a preset threshold. When the cost is less than the threshold, the lane change request of the lane-changing vehicle is approved.

2. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 1, characterized in that Step 1 includes the following sub-steps: 1.

1. When a lane-changing vehicle plans to change lanes, it collects its own motion data through its own vehicle sensors to obtain the position, velocity, and acceleration data of the lane-changing vehicle at the current time t0; 1.

2. The lane-changing vehicle obtains the V2X broadcast data of the surrounding vehicles in the time period [t0-3, t0] through the OBU, which includes the current position, speed, and acceleration information of the surrounding vehicles.

3. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 1, characterized in that Step 2 includes the following sub-steps: 2.

1. Divide the collected data within the time period into historical trajectory point data with 0.1s intervals, and construct a graph data for each moment; 2.

2. Input the constructed graph data into the state graph neural network model. The model is designed based on the encoder-decoder structure. The encoder uses a graph convolutional neural network model to learn the structural information in the graph data, and the decoder uses an LSTM network to learn the temporal information of the historical trajectory. 2.

3. The model outputs the predicted trajectory points of adjacent vehicles in the next 5-10 seconds.

4. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 3, characterized in that In step 2.1, the method for constructing graph data is as follows: there are two types of nodes. One is to regard the lane line vector as a node, and the other is to regard the vehicle trajectory vector within every 0.1 second as a node. The node representation is: in,, and They represent the coordinates of the starting and ending nodes of the vector respectively, type is the type of the node, attr is the attribute information of the node. When the node type is a lane line, the attribute information is the road speed limit. All nodes in the reconstructed graph data are fully connected.

5. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 1, characterized in that In step 5, this step is implemented by the following sub-steps: 5.

1. By adding the cost functions constructed in step 3 and step 4, the current sample t can be obtained f Total cost function formula: Calculate the sum of the cost function of the longitudinal path and the cost function of the transverse path; 5.

2. Repeat steps 3, 4 and 5.1 multiple times, each time for the current sample t f The total cost function is calculated; 5.

3. Finally, by comparing the total cost of each time, the minimum cost is obtained, that is, the optimal solution for lane-changing vehicle path planning is obtained.

6. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 1, characterized in that In step 7, this step is implemented by the following sub-steps: 7.

1. Based on the Frenet coordinate system, the longitudinal displacement-time curve of adjacent vehicles is expressed as follows: s(t)=[a0,a1,...,α5][1,t,t 2 ,...,t 5 ] T Among them, a i is the parameter of the polynomial, t represents time; 7.

2. Set the lane-changing time t of the lane-changing vehicle f The speed and acceleration of the vehicle are used as the termination conditions, and the position, speed, and acceleration information of the adjacent vehicle at the current time t0 are used as constraints to introduce them into the quintic polynomial in step 7.

1. The quintic polynomial s(t) is converted into a function form that is only related to a5: Among them, f1(t) and f2(t) are polynomials about t, t∈[t0,t f ], t0 is the time when the adjacent vehicle decides to change lanes, t f is the lane change completion time of the lane-changing vehicle, s0, is the position, velocity, and acceleration of the adjacent vehicle at time t0, t f The speed and acceleration of the vehicle changing lanes at any moment; 7.3, taking the maximum comfortable acceleration as the upper bound and the minimum vehicle speed as the lower bound, the end time is sampled s(t f )∈[s fmin ,s fmax ], thus obtaining the value range of a5: 7.

4. Calculate the cost function: And in [a 5min , a 5max ] area, we can find the minimum value in [t0, t f ] range, the longitudinal path planning of the adjacent vehicle is obtained, and finally, the longitudinal displacement-time curve is used as the longitudinal path of the adjacent vehicle.

7. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 1, characterized in that In step 6, the lane-changing vehicle sends its lane-changing intention to the adjacent vehicles in the target lane through its OBU. The OBU of the adjacent vehicles receives the data packet sent by the lane-changing vehicle and parses the data packet to obtain the lane-changing intention of the lane-changing vehicle, which includes the short-term trajectory planning of the lane-changing vehicle obtained in step 5.

8. The V2X-based cooperative lane change path planning method on a highway as claimed in claim 1, characterized in that In step 8, the optimal value of the cost function of the adjacent vehicle in step 7 is compared with a preset threshold to determine whether to allow the lane-changing vehicle to change lanes. If the cost of the planned longitudinal path is less than the predetermined threshold, the lane-changing vehicle's lane-changing request is approved, and the adjacent vehicle will then travel along the planned longitudinal path. Otherwise, the lane-changing vehicle's lane-changing request is considered unreasonable, rejected, and the vehicle continues to travel in its current state.

Images