CN-122009218-A - Individualized man-machine co-driving vehicle lane change track planning method in complex environment

CN122009218ACN 122009218 ACN122009218 ACN 122009218ACN-122009218-A

Abstract

The invention discloses a personalized man-machine co-driving vehicle lane change track planning method in a complex environment, and belongs to the technical field of intelligent vehicle control. The method comprises the following steps of generating an initial channel change reference track according to a basic personalized cost function, extracting a corresponding theoretical control sequence based on the initial channel change reference track, carrying out deviation recognition on an input control signal and the theoretical control sequence, calculating a target parameter set capable of minimizing the difference between the theoretical recommended track and an observed sequence, and carrying out online correction on the basic personalized cost function by utilizing the target parameter set. The method and the device solve the problem that the personalized driving model solidification in the prior art cannot adapt to the dynamic change state and the instant intention of the driver in real time by identifying the deviation of the driver operation and the system planning track in real time and reversely solving the implicit driving preference parameters so as to correct the decision model on line.

Inventors

NI JIE
WANG YU
GUO YIWEN
DONG FEI

Assignees

常州江苏大学工程技术研究院
江苏大学

Dates

Publication Date: 20260512
Application Date: 20260401

Claims (8)

1. The personalized path-changing track planning method for the man-machine co-driving vehicle in the complex environment is characterized by comprising the following steps of: acquiring historical driving data of a driver, constructing a reference personalized cost function, and generating an initial lane change reference track according to the reference personalized cost function; collecting a driver input control signal in the running process of the vehicle in real time, and extracting a corresponding theoretical control sequence based on an initial lane change reference track; Performing deviation recognition on the input control signal and the theoretical control sequence, and inputting the input control signal as an observation sequence into a reverse reasoning process when the existence of unplanned behaviors is judged; Performing reverse optimization in a parameter space of a reference personalized cost function, and calculating a target parameter set capable of minimizing differences between a theoretical recommended track and an observation sequence; and carrying out online correction on the reference personalized cost function by utilizing the target parameter set, and generating a re-planning lane change track which accords with the instant intention of the driver by combining the current environment perception data.
2. The method for personalized human-machine co-driving vehicle lane-change trajectory planning in a complex environment according to claim 1, wherein the step of obtaining driver historical driving data and constructing a reference personalized cost function comprises: Collecting lane change behavior data of a target driver in a preset history period through a vehicle-mounted sensor cluster; Constructing a reference personalized cost function, wherein the reference personalized cost function is formed by weighting and summing a safety sub-function, a comfort sub-function and an efficiency sub-function; The safety sub-function quantifies collision risks of vehicles and road boundaries and surrounding vehicles by establishing a potential field model; the comfort sub-function evaluates the running stability by calculating the curvature change rate and the longitudinal impact of the track; the efficiency sub-function measures the passing efficiency by calculating the deviation between the lane change completion time and the expected speed; And extracting historical average preference weights of the driver on safety, comfort and efficiency from the lane change behavior data through a statistical analysis algorithm, and taking the historical average preference weights as initial weight vectors of a basic personalized cost function so as to construct a decision reference reflecting the long-term driving style of the driver.
3. The method for planning a lane-changing trajectory of a personalized man-machine co-driving vehicle in a complex environment according to claim 1, wherein the step of collecting a control signal input by a driver during the running of the vehicle in real time and extracting a corresponding theoretical control sequence based on an initial lane-changing reference trajectory comprises: continuously reading the physical quantity of the driver acting on the vehicle actuating mechanism in a fixed time step to form an input control signal; synchronously acquiring current yaw rate, longitudinal speed and lateral acceleration motion state parameters of the vehicle, and constructing a vehicle dynamics model; Inputting an initial lane change reference track into the vehicle dynamics model, executing feedforward control simulation calculation, and deducing a theoretical control sequence; And (3) strictly aligning the acquired input control signals with the calculated theoretical control sequences on a time axis to form a two-channel data stream containing the measured value and the theoretical value.
4. The method for planning a lane-change trajectory of a personalized man-machine co-driving vehicle in a complex environment according to claim 1, wherein the step of performing deviation recognition on the input control signal and the theoretical control sequence comprises: Setting a sliding time window, and extracting an input control signal subset and a corresponding theoretical control sequence subset in the sliding time window in each sampling period; Calculating the space-time track distance between the input control signal subset and the theoretical control sequence subset; Introducing a dynamic deviation threshold constructed based on the driver's historical operating variance, the dynamic deviation threshold being formulated as: ; In the formula, Representing a dynamic deviation threshold value of the dynamic deviation, Indicating the expected value of the control deviation in the history of normal driving conditions, The standard deviation of the historical deviation is indicated, Representing a sensitivity adjustment coefficient; And when the calculated space-time track distance exceeds the dynamic deviation threshold value in a plurality of continuous sampling periods, judging that the current driving expectation of the driver deviates from the historical average style, namely identifying that one unplanned action occurs.
5. The method for planning a lane-change trajectory of a personalized man-machine co-driving vehicle in a complex environment according to claim 1, wherein the step of performing reverse optimization in a parameter space of a reference personalized cost function and calculating a target parameter set capable of minimizing a difference between a theoretical recommended trajectory and an observed sequence comprises: Defining the parameter vector to be optimized of the reference personalized cost function as The safety weight coefficient, the comfort weight coefficient and the efficiency weight coefficient are included; constructing a reverse optimization objective function, wherein the objective function is used for measuring the fitting degree between a predicted control sequence generated under a specific parameter vector and an observation sequence actually executed by a driver; and searching a group of optimal solutions in a preset parameter vector search space by adopting a Bayesian inference method, so that a minimum value is obtained by a reverse optimization objective function under the solutions.
6. The personalized man-machine co-driving vehicle lane change track planning method according to claim 1, wherein the specific mathematical implementation process of the reverse optimization is as follows: For each candidate parameter vector within the search space Calling a track generation operator to calculate a corresponding simulated track curve ; Simulation of trajectory curves by inverse kinetic model Conversion to analog control sequences ; Calculating a simulated control sequence and an observation sequence using a least squares criterion The sum of squares of the residuals between them is given by: ; Wherein, the In order to observe the length of time of the sequence, Is a discrete point in time; by performing multiple rounds of iterative optimization, find a cause Minimum target parameter set 。
7. The method for planning a lane-change trajectory of a personalized man-machine co-driving vehicle in a complex environment according to claim 1, wherein the step of online correcting the reference personalized cost function by using the target parameter set comprises the steps of: fusing the solved target parameter set with the original reference parameters by adopting a weighted fusion algorithm to generate instant cost function parameters; the weighted fusion algorithm introduces an intention confidence factor, and dynamically adjusts the fusion duty ratio of the target parameter set according to the length of the observation sequence and the confidence level of the deviation recognition stage; and replacing the weight coefficient vector in the reference personalized cost function with the instant cost function parameter to complete the online reconstruction of the cost function model.
8. The method for planning a lane-change trajectory of a personalized man-machine co-driving vehicle in a complex environment according to claim 1, wherein the step of generating a re-planned lane-change trajectory in accordance with the instant intention of a driver by combining current environmental awareness data comprises: Synchronously calling real-time environment data sensed by fusion of the millimeter wave radar and the camera, and acquiring obstacle positions, speeds and acceleration vectors of a target lane and adjacent lanes; Taking the reference personalized cost function after on-line correction as an optimization target, and taking the current position, the gesture and the speed of the vehicle as initial boundary conditions to construct a nonlinear programming problem; Adding road geometric topological limit, vehicle maximum lateral acceleration limit and actuator response delay compensation into constraint conditions, and solving the nonlinear programming problem in millisecond time by using a sequence quadratic programming algorithm; The output series of space-time path points is the re-planning track change track.

Description

Individualized man-machine co-driving vehicle lane change track planning method in complex environment Technical Field The invention relates to the technical field of intelligent vehicle control, in particular to a personalized man-machine co-driving vehicle lane change track planning method in a complex environment. Background With the development of intelligent driving technology, a man-machine co-driving mode has become an important direction for improving driving safety and comfort. In this mode, how to make the vehicle system understand and adapt to the personalized preferences and real-time intentions of different drivers, so as to realize natural and harmonious cooperative control is a key challenge currently faced. Existing personalized driving assistance systems typically construct mathematical models (e.g., models based on clusters or parameterized style classification) reflecting their long-term average driving style by collecting and analyzing historical operating data of the driver, and generate lane-change decisions and trajectories based thereon. Such systems achieve a degree of personalization by learning the general habits of the driver. The prior art mainly relies on the induction of historical average characteristics, and has the problem that a personalized driving model is solidified and cannot adapt to the dynamic change state and the instant intention of a driver in real time. Because the driver's physiological state, level of attention, and driving goals in a specific scenario (e.g., emergency avoidance or pursuit of efficiency) fluctuate in real time, the cured historical model tends to be disjointed from the actual expectations of the driver. The disjoining directly causes the personalized track planned by the system to frequently conflict with the instant operation intention of the driver, so that the driving experience is reduced, the trust of the driver to the system is further damaged, the user is abandoned with the personalized function for a long time, and the practical benefit is difficult to generate due to the investment of related technologies. Therefore, there is a need for a personalized planning method that can dynamically sense and synchronize the driver's real-time intent. Disclosure of Invention The embodiment of the application solves the problem that the personalized driving model is solidified and the dynamic change state and the instant intention of the driver cannot be adapted in real time in the prior art by providing the personalized man-machine co-driving vehicle lane change track planning method under the complex environment, and realizes that a vehicle system can understand and synchronize the current expectations of the driver in real time like the mercy co-driving, and generates the lane change track which always fits the instant operation intention of the driver. The embodiment of the application provides a personalized man-machine co-driving vehicle lane change track planning method in a complex environment, which comprises the steps of obtaining historical driving data of a driver, constructing a reference personalized cost function, and generating an initial lane change reference track according to the reference personalized cost function; collecting a driver input control signal in the running process of the vehicle in real time, and extracting a corresponding theoretical control sequence based on an initial lane change reference track; Performing deviation recognition on the input control signal and the theoretical control sequence, and inputting the input control signal as an observation sequence into a reverse reasoning process when the existence of unplanned behaviors is judged; Performing reverse optimization in a parameter space of a reference personalized cost function, and calculating a target parameter set capable of minimizing differences between a theoretical recommended track and an observation sequence; and carrying out online correction on the reference personalized cost function by utilizing the target parameter set, and generating a re-planning lane change track which accords with the instant intention of the driver by combining the current environment perception data. Further, the step of obtaining driver historical driving data and constructing a reference personalized cost function includes: Collecting lane change behavior data of a target driver in a preset history period through a vehicle-mounted sensor cluster; Constructing a reference personalized cost function, wherein the reference personalized cost function is formed by weighting and summing a safety sub-function, a comfort sub-function and an efficiency sub-function; The safety sub-function quantifies collision risks of vehicles and road boundaries and surrounding vehicles by establishing a potential field model; the comfort sub-function evaluates the running stability by calculating the curvature change rate and the longitudinal impact of the track; the efficiency sub-func