Vehicle lane-changing trajectory planning method on highways based on natural driver characteristics
By obtaining the historical driving trajectory information of natural drivers and conducting statistical analysis, a lane change trajectory that conforms to the characteristics of natural drivers is solved, and the problem that autonomous driving vehicles fail to consider the characteristics of natural drivers during the freeway lane change is improved. Computational efficiency and ride comfort.
Patent Information
- Application Number
- CN202210557257.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-19
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-05-19
AI Technical Summary
Existing autonomous driving vehicles fail to fully consider the characteristics of natural drivers in the highway lane change trajectory planning, affecting the comfort of drivers and passengers and reducing their acceptance of autonomous driving systems.
By obtaining the highway historical driving trajectory information of natural drivers, extracting the lane-changing trajectory fragments, using one-dimensional and two-dimensional Gaussian functions for statistical analysis, generating a trajectory primitive library, and matching the optimal trajectory primitives according to the vehicle's longitudinal driving speed, generating Lane change track that conforms to the characteristics of natural drivers.
The calculation efficiency and generalization ability are improved, and the generated lane change trajectory is more in line with the driving habits of natural drivers, and the driving comfort and acceptance of autonomous driving systems are improved.
Smart Images

Figure CN115092138B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of autonomous driving vehicle trajectory planning, and in particular to a vehicle highway lane changing trajectory planning method based on natural driver characteristics. Background Art
[0002] Trajectory planning technology takes over the decision-making and motion control technologies of autonomous driving. Its trajectory planning results are directly related to driving safety and comfort. It is one of the key technologies for vehicles to achieve autonomous driving, and it is also a hot spot and difficulty in autonomous driving research. Currently, commonly used trajectory planning methods can be roughly divided into artificial potential field method, graph search method, random sampling method, parameter curve method and numerical optimization method.
[0003] Chinese patent application CN112414419A discloses a vehicle lane change path planning method and related devices. The method uses the ram's horn spiral to plan the lane change path based on the lane change starting point and lane change end point information; Chinese patent CN110244713A discloses an intelligent vehicle lane change trajectory planning system and method based on the artificial potential field method. It draws on the obstacle repulsion field model in the artificial potential field method, considers the maximum lateral safety acceleration and road curvature constraints, and plans the lane change path. However, the above trajectory planning methods mostly consider environmental constraints, vehicle dynamics constraints, etc., and rarely consider the natural driver characteristics; with the continuous improvement of the level of autonomous driving, trajectory planning technology that only considers efficiency and safety can no longer meet the needs of drivers and passengers.
[0004] The highway lane-changing scenario is one of the most common driving scenarios. If the lane-changing trajectory planning of the autonomous vehicle cannot fully consider the characteristics of the natural driver, it will inevitably affect the comfort of the driver and passengers, and reduce their acceptance of the autonomous driving system and increase their psychological panic. Therefore, how to extract useful information from the historical driving trajectory information of the vehicle on the highway based on the natural driver and plan a highway lane-changing trajectory that conforms to the characteristics of the natural driver is still in urgent need of research. Summary of the invention
[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a vehicle highway lane changing trajectory planning method based on natural driver characteristics with high computational efficiency, strong generalization ability and good driving comfort.
[0006] The purpose of the present invention can be achieved by the following technical solutions:
[0007] A vehicle highway lane-changing trajectory planning method based on natural driver characteristics, the lane-changing trajectory planning method comprising:
[0008] Step 1: Obtain the historical driving trajectory information of the vehicle on the highway based on the natural driver;
[0009] Step 2: Segment the historical driving trajectory of the vehicle based on the vehicle heading angle information and extract the vehicle lane change trajectory segment information;
[0010] Step 3: classifying the vehicle lane-changing trajectory segments according to the longitudinal speed of the vehicle lane-changing trajectory segments at the start time;
[0011] Step 4: Perform statistical analysis on the lane-changing trajectory segments in each category in step 3 using the one-dimensional Gaussian function maximum likelihood estimation to obtain the expected lane-changing time in a statistical sense;
[0012] Step 5: Scaling the lane-changing trajectory segments in each category according to the corresponding expected lane-changing time obtained in step 4;
[0013] Step 6: Use the two-dimensional Gaussian function maximum likelihood estimation to perform statistical analysis on different lateral and longitudinal positions at the same time to obtain the expected lateral and longitudinal positions in a statistical sense, thereby obtaining the teaching lane change trajectory under the longitudinal vehicle speed classification;
[0014] Step 7: Use the improved trajectory primitive algorithm to learn the lane change teaching trajectory obtained in step 6 to obtain a trajectory primitive library;
[0015] Step 8: According to the longitudinal speed of the vehicle, the optimal trajectory primitive is matched, and the lane changing trajectory based on the natural driver characteristics is generated by using the optimal trajectory primitive according to the target position.
[0016] Preferably, the historical driving trajectory information in step 1 includes the vehicle's lateral position, longitudinal position, speed, acceleration and timestamp.
[0017] Preferably, the step 2 is specifically:
[0018] Step 2-1: Calculate the vehicle heading angle from the vehicle position parameters;
[0019] Step 2-2: Traverse the heading angle from the lane change point in a forward direction to obtain the lane change end point, and then traverse the heading angle in a reverse direction to obtain the lane change start point;
[0020] The lane change point is specifically: the intersection of the vehicle trajectory and the lane line;
[0021] Step 2-3: Extract the vehicle lane-changing trajectory segment information based on the lane-changing starting point and end point.
[0022] More preferably, the method for calculating the vehicle heading angle is specifically as follows:
[0023]
[0024] Among them, θ (t) is the heading angle at time t; x (t)is the longitudinal position of the vehicle at time t; y (t) is the lateral position of the vehicle at time t.
[0025] Preferably, the step 3 is specifically:
[0026] The lane-changing trajectory segments are divided according to the longitudinal speed of the vehicle at the starting moment of the lane-changing trajectory segment. The speed division range is specifically: from 60km / h to 130km / h with a division step of 5km / h, and a total of 14 categories.
[0027] Preferably, the one-dimensional Gaussian function maximum likelihood estimation in step 4 is specifically:
[0028]
[0029] Where T is the time elapsed of the lane-changing trajectory segment; μ T The expected lane changing time for this type of lane changing trajectory segment; is the variance of the elapsed time of the lane-changing trajectory segments.
[0030] Preferably, the step 5 is specifically:
[0031] Step 5-1: Obtain the longitudinal speed curve of the lane change trajectory and scale it;
[0032] Specifically:
[0033]
[0034] in, is the scaled longitudinal vehicle speed; v x is the longitudinal vehicle speed before scaling;
[0035] Step 5-2: Using one-dimensional linear interpolation, sample the scaled longitudinal vehicle speed N times at equal intervals based on the time dimension to obtain a longitudinal vehicle speed sequence at the expected time, and combine it with the expected lane change time to obtain an expected longitudinal position sequence;
[0036] The specific number N is:
[0037]
[0038] Where, Δt is the sampling interval;
[0039] The longitudinal vehicle speed sequence under the expected time is as follows:
[0040]
[0041] The expected vertical position sequence is:
[0042]
[0043] Step 5-3: Based on the expected longitudinal position sequence obtained in step 5-2, combined with the lane-changing trajectory, the expected lateral position is obtained by using one-dimensional linear interpolation to complete the scaling of the lane-changing trajectory on a time scale;
[0044] The expected horizontal position sequence is as follows:
[0045]
[0046] Preferably, the two-dimensional Gaussian function maximum likelihood estimation in step 6 is specifically:
[0047]
[0048] Among them, σ x is the vertical position standard deviation; σ y is the standard deviation of the lateral position; ρ is the correlation coefficient between x and y; μ x is the desired longitudinal position; μ y is the desired lateral position.
[0049] Preferably, the improved trajectory primitive algorithm in step 7 is:
[0050]
[0051] Where τ is the time scaling factor; α z and β z is the system parameter; g is the target position; y0 is the system starting position; f is the forcing function; y, and are the system displacement, velocity and acceleration respectively; x is the system independent variable;
[0052] The forcing function f is specifically:
[0053]
[0054] Where Ψ(x) is the Gaussian basis function; ω is the basis function weight; M is the number of basis functions;
[0055] x is a function of time t, specifically:
[0056]
[0057] Among them, α x >0, it is the system constant;
[0058] The step 7 is specifically as follows:
[0059] Step 7-1: Calculate the target forcing function based on the teaching trajectory;
[0060] Specifically:
[0061]
[0062] Among them, y demo , and are the displacement, velocity and acceleration of the teaching trajectory respectively;
[0063] Step 7-2: Construct a loss function and use the local weighted regression algorithm to solve the weight value corresponding to the basis function;
[0064] The loss function is specifically:
[0065]
[0066] The method for solving the weight value corresponding to the basis function by the local weighted regression algorithm is as follows:
[0067]
[0068]
[0069]
[0070]
[0071] Preferably, the step 8 is specifically:
[0072] Step 8-1: Based on the longitudinal speed at the start of the teaching trajectory and the minimum distance principle, select the teaching trajectory closest to the longitudinal speed of the vehicle, and then match the optimal trajectory primitive;
[0073] Step 8-2: Calculate and generate a lane-changing trajectory based on the driver's characteristics using the optimal trajectory primitive according to the target lane-changing position;
[0074] Specifically, the trajectory primitive algorithm is used to calculate and obtain the displacement, velocity and acceleration information of the lane-changing trajectory;
[0075] The trajectory primitive algorithm is as follows:
[0076] Compared with the prior art, the present invention has the following beneficial effects:
[0077] 1. High computational efficiency and strong generalization ability: After obtaining the trajectory primitive library, the vehicle highway lane change trajectory planning method of the present invention matches the optimal trajectory primitive according to the longitudinal driving speed of the vehicle. When generating the lane change trajectory, only iterative calculation is required, and no numerical solution and optimization problems are involved. Therefore, the computational efficiency is high; and according to different target lane change positions, multiple lane change trajectories that meet the characteristics of natural drivers can be quickly generalized.
[0078] 2. It fully considers the natural driver characteristics during lane changing and the driving comfort is highly acceptable: The trajectory primitive library obtained by the vehicle highway lane changing trajectory planning method in the present invention is based on the driving data of natural drivers, and the lane changing trajectory obtained by generalization is more in line with the natural driver characteristics, so it is highly accepted by drivers and passengers. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] Figure 1 It is a flow chart of the vehicle highway lane change trajectory planning method in the present invention;
[0080] Figure 2 Schematic diagram of the probability distribution of the horizontal and vertical positions of lane change trajectories in the vehicle speed range of 85km / h to 90km / h in an embodiment of the present invention;
[0081] Figure 3 Schematic diagram of lane change teaching trajectory in the vehicle speed range of 85km / h to 90km / h in an embodiment of the present invention;
[0082] Figure 4 Schematic diagram of the generalized lane changing trajectory in the horizontal and vertical directions in an embodiment of the present invention. DETAILED DESCRIPTION
[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0084] A vehicle highway lane-changing trajectory planning method based on natural driver characteristics, the process is as follows Figure 1 As shown, including:
[0085] Step 1: Obtain the historical driving trajectory information of the vehicle on the highway based on the natural driver;
[0086] The historical highway driving trajectory information of the vehicle based on the natural driver includes the vehicle's lateral and longitudinal position, speed, acceleration and timestamp;
[0087] Step 2: Segment the historical driving trajectory of the vehicle based on the vehicle heading angle information and extract the vehicle lane change trajectory segment information;
[0088] Step 3: classifying the vehicle lane-changing trajectory segments according to the longitudinal speed of the vehicle lane-changing trajectory segments at the start time;
[0089] Step 4: Perform statistical analysis on the lane-changing trajectory segments in each category in step 3 using the one-dimensional Gaussian function maximum likelihood estimation to obtain the expected lane-changing time in a statistical sense;
[0090] The maximum likelihood estimation of one-dimensional Gaussian function is specifically:
[0091]
[0092] Where T is the time elapsed for the lane-changing trajectory segment, μ T is the expected lane-changing time for this type of lane-changing trajectory segment, is the variance of the elapsed time of the lane-changing trajectory segments.
[0093] Step 5: Scaling the lane-changing trajectory segments in each category according to the corresponding expected lane-changing time obtained in step 4;
[0094] Step 6: Use the two-dimensional Gaussian function maximum likelihood estimation to perform statistical analysis on different lateral and longitudinal positions at the same time to obtain the expected lateral and longitudinal positions in a statistical sense, thereby obtaining the teaching lane change trajectory under the longitudinal vehicle speed classification, such as Figure 2 , Figure 3 As shown;
[0095] The maximum likelihood estimation of the two-dimensional Gaussian function is specifically:
[0096]
[0097] Among them, σ x is the vertical position standard deviation, σ y is the standard deviation of the lateral position, ρ is the correlation coefficient between x and y, μ x is the desired longitudinal position, μ y is the desired lateral position.
[0098] Step 7: Use the improved trajectory primitive algorithm to learn the lane change teaching trajectory obtained in step 6 to obtain a trajectory primitive library;
[0099] Step 8: According to the longitudinal speed of the vehicle, the optimal trajectory primitive is matched, and the lane-changing trajectory based on the driver's characteristics is generated using the optimal trajectory primitive according to the target position, such as Figure 4 shown.
[0100] Step 2 is as follows:
[0101] Step 2-1: Calculate the vehicle heading angle from the vehicle position parameters;
[0102] Step 2-2: Traverse the heading angle from the lane change point in a forward direction to obtain the lane change end point; traverse the heading angle in a reverse direction to obtain the lane change start point, where the lane change point is defined as the intersection of the vehicle trajectory and the lane line;
[0103] Step 2-3: Extract the vehicle lane-changing trajectory segment information based on the lane-changing starting point and end point.
[0104] The specific method for calculating the vehicle heading angle is:
[0105]
[0106] Among them, θ (t) is the heading angle at time t, x (t) is the longitudinal position of the vehicle at time t, y (t) is the lateral position of the vehicle at time t.
[0107] Step 3 is specifically as follows: the lane changing trajectory segments are divided according to the longitudinal speed of the vehicle at the starting time of the lane changing trajectory segments, and the speed division interval is specifically from 60km / h to 130km / h with a division step of 5km / h, and is divided into 14 categories in total.
[0108] Step 5 is as follows:
[0109] Step 5-1: Obtain the longitudinal speed curve of the lane change trajectory and scale it. The specific formula is:
[0110]
[0111] in, is the scaled longitudinal vehicle speed, v x is the longitudinal vehicle speed before scaling.
[0112] Step 5-2: Using one-dimensional linear interpolation, sample the scaled longitudinal vehicle speed N times at equal intervals based on the time dimension to obtain a longitudinal vehicle speed sequence at the expected time, and combine it with the expected lane change time to obtain an expected longitudinal position sequence;
[0113] N is specifically:
[0114]
[0115] Where, Δt is the sampling interval;
[0116] The longitudinal vehicle speed sequence under the expected time is as follows:
[0117]
[0118] The expected vertical position sequence is as follows:
[0119]
[0120] Step 5-3: Based on the expected longitudinal position sequence obtained in step 5-2, combined with the lane-changing trajectory, the expected lateral position is obtained by using one-dimensional linear interpolation to complete the scaling of the lane-changing trajectory on a time scale;
[0121] The expected horizontal position sequence is as follows:
[0122]
[0123] Step 7 is as follows:
[0124] The improved trajectory primitive algorithm can avoid algorithm failure when the target position is close to the starting position. Specifically:
[0125]
[0126] Where τ is the time scaling factor; α z and β z is the system parameter; g is the target position; y0 is the system starting position; f is the forcing function; y, and are the system displacement, velocity and acceleration respectively; x is the system independent variable;
[0127] The forcing function f is specifically:
[0128]
[0129] Where Ψ(x) is the Gaussian basis function; ω is the basis function weight; M is the number of basis functions;
[0130] x is a function of time t, specifically:
[0131]
[0132] Among them, α x >0, it is the system constant;
[0133] The step 7 is specifically as follows:
[0134] Step 7-1: Calculate the target forcing function based on the teaching trajectory;
[0135] Specifically:
[0136]
[0137] Among them, y demo , and are the displacement, velocity and acceleration of the teaching trajectory respectively;
[0138] Step 7-2: Construct a loss function and use the local weighted regression algorithm to solve the weight value corresponding to the basis function;
[0139] The loss function is specifically:
[0140]
[0141] The method for solving the weight value corresponding to the basis function by the local weighted regression algorithm is as follows:
[0142]
[0143]
[0144]
[0145]
[0146] Step 8 is as follows:
[0147] Step 8-1: Based on the longitudinal speed at the start of the teaching trajectory and the minimum distance principle, select the teaching trajectory closest to the longitudinal speed of the vehicle, and then match the optimal trajectory primitive;
[0148] Step 8-2: Calculate and generate a lane-changing trajectory based on the driver's characteristics using the optimal trajectory primitive according to the target lane-changing position;
[0149] The lane-changing trajectory based on the driver's characteristics is calculated and generated by: using the trajectory primitive algorithm to calculate and obtain the displacement, velocity, and acceleration information of the lane-changing trajectory;
[0150] The trajectory primitive algorithm is as follows:
[0151]
[0152] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.
Claims
1. A vehicle highway lane-changing trajectory planning method based on natural driver characteristics, characterized in that: The lane-changing trajectory planning method includes: Step 1: Obtain the historical driving trajectory information of the vehicle on the highway based on the natural driver; Step 2: Segment the historical driving trajectory of the vehicle based on the vehicle heading angle information and extract the vehicle lane change trajectory segment information; Step 3: classifying the vehicle lane-changing trajectory segments according to the longitudinal vehicle speed at the start time of the acquired vehicle lane-changing trajectory segments; Step 4: Perform statistical analysis on the lane-changing trajectory segments in each category in step 3 using the one-dimensional Gaussian function maximum likelihood estimation to obtain the expected lane-changing time in a statistical sense; Step 5: Scaling the lane-changing trajectory segments in each category according to the corresponding expected lane-changing time obtained in step 4; Step 6: Use the two-dimensional Gaussian function maximum likelihood estimation to perform statistical analysis on different lateral and longitudinal positions at the same time to obtain the expected lateral and longitudinal positions in a statistical sense, thereby obtaining the teaching lane change trajectory under the longitudinal vehicle speed classification; Step 7: Use the improved trajectory primitive algorithm to learn the lane change teaching trajectory obtained in step 6 to obtain a trajectory primitive library; Step 8: Match the optimal trajectory primitive according to the longitudinal speed of the vehicle, and use the optimal trajectory primitive to generate a lane change trajectory based on natural driver characteristics according to the target position; The one-dimensional Gaussian function maximum likelihood estimation in step 4 is specifically: Where T is the time elapsed of the lane-changing trajectory segment; μ T The expected lane changing time for this type of lane changing trajectory segment; is the variance of the elapsed time of the lane-changing trajectory segment; The step 5 is specifically as follows: Step 5-1: Obtain the longitudinal speed curve of the lane change trajectory and scale it; Specifically: in, is the scaled longitudinal vehicle speed; v x is the longitudinal vehicle speed before scaling; Step 5-2: Using one-dimensional linear interpolation, sample the scaled longitudinal vehicle speed N times at equal intervals based on the time dimension to obtain a longitudinal vehicle speed sequence at the expected time, and combine it with the expected lane change time to obtain an expected longitudinal position sequence; The specific number N is: Where, Δt is the sampling interval; The longitudinal vehicle speed sequence under the expected time is as follows: The expected vertical position sequence is as follows: Step 5-3: Based on the expected longitudinal position sequence obtained in step 5-2, combined with the lane-changing trajectory, the expected lateral position is obtained by using one-dimensional linear interpolation to complete the scaling of the lane-changing trajectory on a time scale; The expected horizontal position sequence is as follows: The improved trajectory primitive algorithm in step 7 is: Where τ is the time scaling factor; α z and β z is the system parameter; g is the target position; y0 is the system starting position; f is the forcing function; y, and are the system displacement, velocity and acceleration respectively; x is the system independent variable; The forcing function f is specifically: Among them, Ψ(x) is the Gaussian basis function; w i is the basis function weight; M is the number of basis functions; x is a function of time t, specifically: Among them, α x >0, it is the system constant; The step 7 is specifically as follows: Step 7-1: Calculate the target forcing function based on the teaching trajectory; Specifically: Among them, y demo , and are the displacement, velocity and acceleration of the teaching trajectory respectively; Step 7-2: Construct a loss function and use the local weighted regression algorithm to solve the weight value corresponding to the basis function; The loss function is specifically: The method for solving the weight value corresponding to the basis function by the local weighted regression algorithm is as follows:
2. The method for planning lane-changing trajectories on a highway based on natural driver characteristics according to claim 1, characterized in that: The historical driving trajectory information in step 1 includes the vehicle's lateral position, longitudinal position, speed, acceleration and timestamp.
3. The method for planning vehicle lane-changing trajectory on highways based on natural driver characteristics according to claim 1, characterized in that: The step 2 is specifically as follows: Step 2-1: Calculate the vehicle heading angle from the vehicle position parameters; Step 2-2: Traverse the heading angle from the lane change point in a forward direction to obtain the lane change end point, and then traverse the heading angle in a reverse direction to obtain the lane change start point; The lane change point is specifically: the intersection of the vehicle trajectory and the lane line; Step 2-3: Extract the vehicle lane-changing trajectory segment information based on the lane-changing starting point and end point.
4. The method for planning vehicle lane-changing trajectory on highways based on natural driver characteristics according to claim 3, characterized in that: The method for calculating the vehicle heading angle is specifically as follows: Among them, θ (t) is the heading angle at time t; x (t) is the longitudinal position of the vehicle at time t; y (t) is the lateral position of the vehicle at time t.
5. The method for planning vehicle lane-changing trajectory on highways based on natural driver characteristics according to claim 1, characterized in that: The step 3 is specifically as follows: The lane-changing trajectory segments are divided according to the longitudinal vehicle speed at the starting moment of the lane-changing trajectory segment. The speed division interval is specifically: from 60km / h to 130km / h with a division step of 5km / h, and a total of 14 categories.
6. The vehicle highway lane change trajectory planning method based on natural driver characteristics according to claim 1, characterized in that: The two-dimensional Gaussian function maximum likelihood estimation in step 6 is specifically: Among them, σ x is the standard deviation of the longitudinal position; σ y is the standard deviation of the lateral position; ρ is the correlation coefficient between x and y; μ x is the desired longitudinal position; μ y is the desired lateral position.
7. The vehicle highway lane change trajectory planning method based on natural driver characteristics according to claim 1, characterized in that: The step 8 is specifically as follows: Step 8-1: Based on the longitudinal speed at the start of the teaching trajectory and the minimum distance principle, select the teaching trajectory closest to the longitudinal speed of the vehicle, and then match the optimal trajectory primitive; Step 8-2: Calculate and generate a lane-changing trajectory based on the driver's characteristics using the optimal trajectory primitive according to the target lane-changing position; Specifically: the trajectory primitive algorithm is used to calculate and obtain the displacement, velocity, and acceleration information of the lane changing trajectory.
