Automatic driving vehicle cooperative lane changing strategy considering dynamic risk information

Abstract

The invention provides an automatic driving vehicle collaborative lane changing strategy considering dynamic risk information, and the method is mainly divided into three parts: firstly, constructing a queue collaborative lane changing control architecture, taking a queue collaborative lane changing decision making method considering the dynamic risk information as a decision making layer, and taking queue collaborative lane changing control based on a four-element method as a control layer; and the two are cooperated to complete decision making and control of queue vehicles and lane changing vehicles. Then, by analyzing the motion states of the lane changing vehicles and the queue vehicles, the relative positions and the relative speeds serve as main influence factors, and an optimal cut-in position determining method and lane changing allowing conditions of the lane changing vehicles are given; a risk field with road traffic participants as a field source is designed based on the field theory, the risk field is used as a decision factor of a transverse lane changing track of a lane changing vehicle, and a transverse lane changing track dynamic planning model is constructed. Finally, on the basis of the basic structure of the four-element model, the linear two-degree-of-freedom dynamic model serves as transverse dynamics of vehicle nodes, the characteristics of the queue vehicles compared with single vehicles are considered, and the mass-spring-damper model serves as longitudinal dynamics of the vehicle nodes; the queue expected configuration selects a constant time headway strategy; taking a PLF structure as an information flow topological structure of a queue; and respectively designing a transverse controller based on MPC and a longitudinal controller based on linear control.

Description

Technical Field

[0001] The present invention relates to the field of queue cooperative lane changing control in an autonomous driving environment, and in particular to an autonomous driving vehicle cooperative lane changing strategy that considers dynamic risk information. Background Art

[0002] With the promotion and application of technologies such as the Internet of Things, big data, and artificial intelligence in the field of transportation, the intelligent vehicle-road cooperative system, which was previously characterized by intelligence, collaboration, and coordination, has become an important means to improve traffic efficiency and ensure traffic safety. It is foreseeable that when the autonomous driving technology develops to a certain extent, vehicles in the form of queues will become a common form of traffic flow organization. Therefore, making full use of the advantages of intelligent connected vehicles and vehicle-road cooperative environments and proposing a queue cooperative lane changing strategy is of great significance to improving traffic efficiency and ensuring traffic safety.

[0003] At present, most of the research on cooperative lane-changing strategies focuses on the scenario of a free single-vehicle lane-changing vehicle entering a queue. Elham divides the scenario of a single-vehicle entering a queue into three stages: synchronizing vehicle speed, creating gaps, and merging. First, the lane-changing vehicle adjusts its speed to the speed of the target lane fleet and sends a merge request. After receiving the request, the vehicle behind the target lane starts to slow down to create a gap. At the same time, the lane-changing vehicle also slows down to create a gap with the vehicle in front. When the vehicle behind determines that a safe gap has been reached, it sends a merge signal to the lane-changing vehicle. After receiving the signal, the lane-changing vehicle checks whether a safe gap has been reached with the vehicle in front. If it is safe, it actually merges. Zhu Xu divides the entire single-lane queue expansion process into two stages: reserved space and lane fusion. On this basis, he designs a distributed model predictive controller to collaboratively realize the expansion of single-lane queues. Some scholars also focus on the queue merging scenario. Hsu et al. designed a collaborative strategy for all vehicles to merge into the queue at the same time for the queue merging scenario. When the distance between queues is less than the safe distance, the distance between vehicles is first expanded to reach the safe distance, then all vehicles change lanes at the same time, and finally the queue restores the initial distance between vehicles.

[0004] The existing research on cooperative lane-changing strategies has been relatively complete, but in the design of the model, safety is rarely considered. To address this problem, this paper proposes a cooperative lane-changing strategy for autonomous driving vehicles that considers dynamic risk information to improve the safety and efficiency of lane-changing behavior. Summary of the invention

[0005] The present invention is proposed based on the above-mentioned situation of the prior art. The technical problem to be solved by the present invention is to provide a cooperative lane changing strategy for autonomous driving vehicles taking into account dynamic risk information, so as to improve the efficiency and safety of autonomous driving vehicles when changing lanes.

[0006] In order to solve this technical problem, the technical solution provided by the present invention includes: 1. A coordinated lane-changing strategy for autonomous driving vehicles considering dynamic risk information, comprising the following steps:

[0007] Step 1: Build a platoon cooperative lane-changing control architecture, with the platoon cooperative lane-changing decision method that considers dynamic risk information as the decision-making layer, and the platoon cooperative lane-changing control based on the four-element method as the control layer. The two work together to complete the decision-making and control of platoon vehicles and lane-changing vehicles.

[0008] Step 2: Based on the driving status of the lane-changing vehicle and the platoon vehicles, determine the optimal entry position of the lane-changing vehicle and give the lane-changing permission conditions, and perform dynamic trajectory planning for the lane-changing vehicle. When planning the lane-changing trajectory, fully consider the dynamic risks in the traffic environment, quantify the risk value faced by the lane-changing vehicle based on the risk field theory, and apply it to the optimization of trajectory parameters.

[0009] Step 3: Design a platoon cooperative lane-changing control method based on the four-element model, fully utilizing the characteristics of vehicle following behavior when the platoon is used as the traffic flow organization form, select the vehicle node dynamics, the information flow topology and expected configuration of the platoon, and design a distributed controller for all platoon vehicles.

[0010] The above technical solution quantifies the dynamic risks faced by lane-changing vehicles during the lane-changing process, and based on the dynamic path planning and collaborative control considering the dynamic risk information according to the vehicle's motion state, realizes the collaborative lane changing of the queue while ensuring safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is the overall idea frame diagram of the present invention;

[0012] Figure 2 is a schematic diagram of a cooperative lane-changing scenario;

[0013] Figure 3 is a schematic diagram of the initial cut-in position;

[0014] Figure 4 It is a schematic diagram of the best entry position;

[0015] Figure 5 It is a flow chart of the risk field discretization method;

[0016] Figure 6 It is a schematic diagram of the risk field discretization method;

[0017] Figure 7 It is a schematic diagram of the information flow topology. DETAILED DESCRIPTION

[0018] The present invention is described in detail below with reference to the accompanying drawings and implementation schemes. It should be understood that this example is only used to illustrate the present invention and is not used to limit the scope of the present invention. The overall idea of ​​the autonomous driving vehicle cooperative lane changing strategy considering dynamic risk information provided by the present invention is as follows: Figure 1 As shown in Figure 1, its application scenario is a bicycle changing lanes and entering a queue on a city road. Figure 2 As shown, the specific implementation method includes the following steps:

[0019] Step 1: The lane-changing vehicle sends a lane-changing request to the lead vehicle in the target queue.

[0020] Step 2: After receiving the lane change request from the lane-changing vehicle, the platoon leader vehicle gives the best entry position for the lane-changing vehicle based on the current driving status of the lane-changing vehicle and the platoon vehicle. The specific implementation steps of step 2 are as follows:

[0021] Step 21 Calculate the vehicle number closest to the lane-changing vehicle

[0022] In real road scenarios, in order to improve lane-changing efficiency and reduce the risks faced during the lane-changing process, drivers generally choose the gap between the nearest vehicles as the lane-changing entry point. The advantage of this selection method is that the speed fluctuation is small during lane changing, and the time required for the queue to return to stability is short. Based on this, first calculate the number of the vehicle closest to the lane-changing vehicle as i, and the calculation method is:

[0023]

[0024] like Figure 3 Vehicle k-1 is the vehicle closest to the lane-changing vehicle C in the longitudinal direction, so i=k-1 in this scenario.

[0025] Step 22 Determine the best entry position

[0026] The relative position, speed and acceleration of the lane-changing vehicle and vehicle i determine whether the lane-changing vehicle cuts in front of or behind vehicle i. Figure 4 As shown in the figure, t and T are the lane change request time and lane change time step respectively. If the lane-changing vehicle can drive in front of vehicle i at time t+T, it should choose to cut in front of vehicle i; otherwise, it should cut in behind vehicle i. The queue vehicle number in front of the best cut-in position is I, which satisfies:

[0027]

[0028] where x(t), v(t) and a(t) are the longitudinal position, velocity and acceleration of the vehicle, respectively. The subscripts c and i represent the lane-changing vehicle and the i-th vehicle in the queue, respectively.

[0029] Step 3 After the platoon leader vehicle determines the best entry position for the lane-changing vehicle, it determines whether the lane-changing vehicle can start the lane-changing behavior based on the relative position and speed of the lane-changing vehicle and the platoon vehicle. This patent mainly considers the safe vehicle spacing as a key parameter of the lane-changing permission condition. Specifically, the safe vehicle spacing is the product of the safety factor λ and the vehicle body length L, and the lane-changing permission condition C is defined as:

[0030]

[0031] When C is 1, the lane-changing vehicle can start the lane-changing operation under the control of the lateral and longitudinal control strategies; otherwise, the lane-changing vehicle needs to wait for the queue to provide it with sufficient lane-changing gap.

[0032] Step 4: When the platoon leader determines the best entry position for the lane-changing vehicle and the lane-changing permission condition C is 1, the lane-changing vehicle starts to execute the lane-changing operation. The platoon leader updates the platoon ID and dynamically plans its lane-changing trajectory during the lane-changing process. The specific implementation steps of step 4 are as follows:

[0033] Step 41 Quantify dynamic risk

[0034] Consider each vehicle in the queue as a risk source, and its spatial position as the origin of the coordinate system of the risk field it generates. Then the width is Wk, the length is Lk, and the speed is vk ,lon (t), acceleration is ak ,lon The risk value RV corresponding to the risk field generated by vehicle k at time t at (x, y) ,k The calculation method of (x,y,t) is:

[0035]

[0036] where δk ,lon (x,y,t) and δk ,lat (x, y, t) are the longitudinal and lateral attenuation coefficients of the risk field, α, β and γ are the sensitivity coefficients of the risk field to the speed, position and acceleration of the source vehicle, the subscript k represents the kth vehicle, lon and lat represent the longitudinal and lateral directions, and the calibration method can be found in the paper "Risk Field Model of Driving and Its Application in Modeling Car-Following Behavior" by Tan et al. For the global coordinate system, the risk value of any point should be the superposition of the risk values ​​of all risk fields at that point.

[0037] In order to reduce the computational load in practical applications, the continuous risk field is discretized. Figure 5As shown in the figure, first determine the calculation range as the minimum outer frame rectangle of the expected trajectory; initialize the entire calculation range with the length and width of the lane-changing vehicle body as the length and width of the initial cell; calculate the risk value of the center point of each cell according to the above calculation method, and use it to represent the risk value of the entire cell; check whether the risk value of each cell is less than the risk threshold R0∈(0,1], if it is less than, the cell is processed, otherwise it is divided into 2×2 sub-cells along the horizontal and vertical symmetry axes of the cell, and the calculation and check are repeated until the risk values ​​of all cells are less than the risk threshold or the minimum cell size has been reached. Figure 6 For example, assuming that the risk threshold is set to 0.8, the risk value of the upper right corner cell in the bottom layer is calculated to be 0.89, which is greater than the set risk threshold. At this time, it is divided into 2×2 sub-cells along its horizontal and vertical axes of symmetry, which are the cells in the top layer of the figure, and the process of "calculation-checking-re-division" is repeated until the iterative stop condition of the algorithm is reached.

[0038] Step 42: Calculate the longitudinal length of the optimal expected trajectory

[0039] The longitudinal length of the trajectory is modeled as a parameter optimization problem, expressed as follows:

[0040]

[0041] Among them, R(L t ) indicates that the longitudinal length is L t The risk value of the cells that the trajectory passes through is summed up to represent the risk value of the trajectory, which is used to characterize safety; the longitudinal length represents the lane-changing efficiency; both are normalized by comparing them with their value ranges, ω is the weight parameter, and the objective function is the weighted average of trajectory safety and lane-changing efficiency. The constraints mainly constrain the range of the heading angle of the lane-changing vehicle by limiting the value of the first-order derivative of the trajectory function.

[0042] Step 43: Solve for the expected trajectory parameters

[0043] The fifth-order polynomial curve is used as the preset form of the trajectory curve, and the relevant parameters are solved according to the state of the vehicle at the beginning and end of the lane change process. The expected trajectory planned at time t is defined as:

[0044]

[0045] Where (x0, y0) is the position of the vehicle at time t, a i is the curve coefficient. At the start and end of the lane change, the vehicle travels in the positive longitudinal direction of the ground coordinate system and has no lateral acceleration. tWhen x = x0, the slope and curvature of the curve should be calculated based on the lateral and longitudinal speeds and lateral acceleration of the vehicle at time t; x = x0 + L t When , the slope and curvature of the trajectory curve are both 0. Therefore, the value of the curve parameter can be solved and expressed by the following formula:

[0046]

[0047] Where w is the lane width, v x,t and v y,t are the longitudinal speed and lateral speed of the lane-changing vehicle at time t, a y,t is the lateral acceleration of the lane-changing vehicle at time t.

[0048] After the relevant parameters and longitudinal length of the desired trajectory are determined, the desired lane-changing trajectory of the lane-changing vehicle is completely determined.

[0049] Step 5: After completing the ID update of all vehicles and the trajectory planning of the lane-changing vehicle, the platoon vehicles and the lane-changing vehicles begin to cooperate with each other to complete the lane-changing operation under the control of the collaborative control strategy. The specific implementation steps of step 5 are as follows:

[0050] Step 51 Determine vehicle node dynamics

[0051] Node dynamics is a module that describes the dynamics of each vehicle object in the scene. Each vehicle is regarded as a node and its dynamic characteristics are described in mathematical language. The linear two-degree-of-freedom dynamics model is used as the lateral dynamics of the vehicle node, and its dynamic state space model is expressed by the following formula:

[0052]

[0053] Among them, v x and y are the longitudinal speed and lateral displacement of the vehicle, respectively; m is the vehicle mass; δ is the front wheel steering angle; θ is the vehicle yaw angle; k1 and k2 are the front and rear axle cornering stiffnesses (positive values); L f and L r are the distances from the front and rear axles to the center of mass of the vehicle, I Z is the vehicle's moment of inertia about the Z axis.

[0054] Considering the interaction between vehicles in the queue from the driver's perspective, the main factors affecting the driver's acceleration and deceleration strategies are the relative position and speed of the vehicle and the front vehicle. The present invention regards the queue as a mass-spring-damper system, and the spring coefficient and the damping coefficient of the damper are used to characterize the influence of relative position and relative speed on the acceleration strategy. The state space equation corresponding to the longitudinal dynamics of the vehicle node is expressed by the following formula:

[0055]

[0056] where x i (t) is the longitudinal position of vehicle i at time t, m i is the mass of vehicle i, u i (t) is the control value of vehicle i at time t, k i and i are the influence coefficients of relative position and relative velocity respectively.

[0057] Step 52: Determine the desired configuration of the queue

[0058] The constant headway strategy is used as the desired configuration of the queue. When the queue is stable, the vehicles in the queue should maintain the same speed and headway, that is:

[0059]

[0060] where v0 and v i are the speeds of the platoon leader and the i-th following car, x0 and x i are the longitudinal positions of the platoon leader and the i-th following vehicle, L is the vehicle length, gap is the minimum distance between vehicles that should be maintained when the platoon is stationary, and h des is the expected headway.

[0061] Step 53: Determine the information flow topology

[0062] The basic way to characterize the topological structure of information flow is to transform the information interaction mode of the queue from an algebraic graph into a matrix. Figure 7 For example, the platoon includes 1 leading car and 4 following cars. The leading car-leading car following type is used as the information flow topology of the platoon. The related information flow topology matrix is ​​given by the following formula:

[0063]

[0064] Wherein, A4 is the adjacency matrix, D4 is the Laplace matrix, and P4 is the pulling matrix. The information flow topology structure matrix will be used in the distributed controller design in step 54.

[0065] Step 54 determines the distributed controller

[0066] The lateral error state of the lane-changing vehicle is defined as

[0067]

[0068] where x des is the expected lateral state quantity, which is calculated by the expected lane-changing trajectory curve function given by the lane-changing vehicle dynamic trajectory planning part of the present invention. Then the prediction model of the model predictive control of the lateral control is expressed by the following formula:

[0069]

[0070] in, and δ(i|k) represent the error state and control quantity predicted i steps backward in the kth step; N P and N C are respectively the prediction time domain and the control time domain, and satisfy N P ≥U C .

[0071] The control goal of the lateral controller is to make the lane-changing trajectory of the lane-changing vehicle as close to the expected lane-changing trajectory as possible, that is, the error between the two is as small as possible. Based on this, the MPC optimization problem is expressed as follows:

[0072]

[0073] in, and are weight matrices, δ min and δ max is the boundary value of the vehicle's front wheel turning angle, Δδ min and Δδ max is the boundary value of the rate of change of the vehicle's front wheel steering angle; the constraint condition ensures that the control amount is within the reachable range and that the change in the control amount before and after a single sampling interval meets the vehicle performance limit.

[0074] The longitudinal controller based on linear control is expressed as follows:

[0075]

[0076] Assuming that the queue is homogeneous, that is, the vehicles in the queue have the same dynamic characteristics, and ignoring the communication delay, the control law can be simplified to:

[0077]

[0078] The aggregate state quantity and aggregate control quantity of the queue are defined as follows:

[0079]

[0080] U(t)=[u1(t) u2(t) … u N (t)] T ∈R N×1

[0081] The control rate of the queue is simplified to After the lane-changing behavior begins, the lane-changing vehicle should be considered as a member of the platoon, and the control law of the platoon vehicles should be calculated according to the new information flow topology to achieve coordinated lane changing.

[0082] The above steps describe the implementation process of the present invention in detail, but the present invention is not limited to the specific details of the above implementation. Anything within the scope of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A cooperative lane-changing strategy for autonomous driving vehicles considering dynamic risk information, characterized in that: The steps include: Step 1: Build a platoon cooperative lane-changing control architecture, with the platoon cooperative lane-changing decision method that considers dynamic risk information as the decision-making layer, and the platoon cooperative lane-changing control based on the four-element method as the control layer. The two work together to complete the decision-making and control of platoon vehicles and lane-changing vehicles. Step 2: Based on the driving status of the lane-changing vehicle and the platoon vehicles, determine the optimal entry position of the lane-changing vehicle and give the lane-changing permission conditions, and perform dynamic trajectory planning for the lane-changing vehicle. When planning the lane-changing trajectory, fully consider the dynamic risks in the traffic environment, quantify the risk value faced by the lane-changing vehicle based on the risk field theory, and apply it to the optimization of trajectory parameters. Step 3: Design a platoon cooperative lane-changing control method based on the four-element model, fully utilizing the characteristics of vehicle following behavior when the platoon is used as the traffic flow organization form, select the vehicle node dynamics, the information flow topology and expected configuration of the platoon, and design a distributed controller for all platoon vehicles.

2. The autonomous driving vehicle cooperative lane changing strategy considering dynamic risk information according to claim 1, characterized in that: The method to quantify dynamic risk is: Consider each vehicle in the queue as a risk source and its spatial position as the origin of the coordinate system of the risk field it generates. The width is W k , length is L k , speed is v k,lon (t), acceleration is a k,lon The risk value of the risk field generated by vehicle k at time t at (x, y) is where δ k,lon (x,y,t) and δ k,lat (x, y, t) are the longitudinal and lateral attenuation coefficients of the risk field, α, β and γ are the sensitivity coefficients of the risk field to the speed, position and acceleration of the source vehicle, the subscript k represents the kth vehicle, lon and lat represent the longitudinal and lateral directions, respectively. For the global coordinate system, the risk value of any point should be the superposition of the risk values ​​of all risk fields at that point. In the design of the risk field function, the influence of vehicle position, speed and acceleration are comprehensively considered, so that the calculated risk field is asymmetric in both the horizontal and vertical directions, which is more in line with the psychological expectations and actual conditions of the risk field.

3. The autonomous driving vehicle cooperative lane changing strategy considering dynamic risk information according to claim 1, characterized in that: The method to determine the optimal expected trajectory longitudinal length is: The longitudinal length of the trajectory is modeled as a parameter optimization problem, expressed as follows: Among them, R(L t ) indicates that the longitudinal length is L t The risk value of the trajectory is summed up in the discrete risk field, which is used as the risk value of the trajectory to characterize safety; the second term uses the longitudinal length to represent the lane-changing efficiency; both are normalized by comparing them with their value ranges, ω is the weight parameter, and the objective function is the weighted average of trajectory safety and lane-changing efficiency. The constraints mainly constrain the range of the heading angle of the lane-changing vehicle by limiting the value of the first-order derivative of the trajectory function. The safety of the desired trajectory is calculated by the risk quantification method described in claim 2, and the objective function of the optimization problem is designed to be a weighted average of trajectory safety and lane-changing efficiency, taking into account both the safety of the desired trajectory and the lane-changing efficiency.

Images