A lane changing method and system, computer device, and storage medium

Abstract

The application relates to the field of driving assistance, in particular to a lane changing method and system, a computer device and a storage medium. The lane changing method comprises the following steps: receiving a lane changing instruction and generating a lane changing intention; screening a safe gap that can be used for merging in a target lane according to the motion state of a vehicle in the target lane; predicting the reaction of an interactive object of the target lane based on an interactive model of the interactive object and the ego vehicle in the case that the ego vehicle executes any initial lane changing scheme, and generating a corresponding interactive scene; obtaining a risk utility value of each interactive scene, determining a target lane changing scheme, generating and executing a lane changing action. The application establishes an interactive model of the ego vehicle and the interactive object, predicts the possible reaction of the interactive object of the target lane, determines the target lane changing scheme according to the quantitative risk utility value, prevents the decision from being too aggressive or conservative, realizes the balance between safety and efficiency, and greatly improves the traffic efficiency while ensuring safety.

Description

Technical Field

[0001] This application relates to the field of driving assistance technology, specifically to a lane-changing method and system, computer equipment, and storage medium. Background Technology

[0002] With the exponential growth of urban traffic flow and the accelerated commercialization of autonomous driving technology, the dynamic interaction of vehicles in congested scenarios has become a key bottleneck restricting the development of intelligent transportation systems. Especially in typical high-density scenarios such as ramp merging, lane changing in construction zones, and intersection passage, connected and autonomous vehicles (CAVs) not only have to deal with the path planning challenges of complex geometric topologies, but also face the dual challenges posed by human-driven vehicles (HDVs) in mixed traffic flows. The random driving behaviors exhibited by human drivers, such as asymmetric following, forced lane changes, and game-like acceleration, result in traffic flow parameters exhibiting strong time-varying characteristics.

[0003] Existing intelligent driving strategies are constrained by the rigid requirement of safety first. In balancing risk prediction and traffic efficiency, they often fall into the dilemma of "over-defense". This manifests as frequent conservative braking, i.e., "awkward stopping", unnecessary yielding and other non-optimal decisions, resulting in system-level traffic efficiency loss and traffic difficulties. Summary of the Invention

[0004] The purpose of the embodiments of this application is to provide a lane-changing method and system, computer equipment, and storage medium to solve the technical problem in the prior art where autonomous driving is difficult to balance safety and traffic efficiency, resulting in traffic difficulties.

[0005] To address the aforementioned technical problems, embodiments of this application disclose the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a lane-changing method, including:

[0007] In response to a received lane change instruction, a lane change intention is generated based on the lane change instruction, the lane change instruction including the target lane;

[0008] Based on the movement state of vehicles in the target lane, safe gaps in the target lane that can be used for merging are selected, and lane-changing trajectories are planned based on the safe gaps to obtain several initial lane-changing schemes.

[0009] Based on the interaction model between the vehicle and the interaction object in the target lane, the reaction of the interaction object is predicted when the vehicle executes any of the initial lane-changing schemes, and a corresponding interaction scenario is generated.

[0010] Obtain the risk utility value for each of the aforementioned interaction scenarios, and determine the target lane-changing scheme from the initial lane-changing scheme based on the risk utility value;

[0011] Based on the target lane-changing scheme, a longitudinal velocity and acceleration adapted to the lateral movement are planned, and a lane-changing action is generated and executed.

[0012] In one embodiment of this application, generating a lane-changing intention based on the lane-changing command includes:

[0013] Extract lane change information from the lane change command, the lane change information including the direction of the target lane, the purpose of lane change, and the lane change constraints;

[0014] Based on the lane change information, a lane change intention is generated that includes a target lane identifier and an initial assessment of the timing of the lane change operation.

[0015] In one embodiment of this application, detecting the motion state of vehicles in the target lane corresponding to the lane-changing intention and filtering out safe gaps in the target lane that can be used for merging includes:

[0016] Obtain the motion states of at least two vehicles within the target lane;

[0017] Calculate the physical clearance between adjacent vehicles within the target lane;

[0018] Each physical gap is evaluated using preset safety indicators, and the physical gap that meets the safety indicators is selected as the safe gap.

[0019] In one embodiment of this application, planning a lane-changing trajectory based on the safety gap includes:

[0020] A parametric curve generation method is used to generate a trajectory connecting the current state of the vehicle with the target point in the selected safety gap within the target lane, thus obtaining the lane-changing trajectory.

[0021] In one embodiment of this application, the method further includes: verifying the obtained lane-changing trajectory, and selecting the verified lane-changing trajectory as the initial lane-changing scheme;

[0022] Verification of the obtained lane-changing trajectory includes:

[0023] Verify whether the lane-changing trajectory meets the vehicle's motion restrictions; and / or;

[0024] Verify whether the required lateral and longitudinal accelerations meet the vehicle's dynamic limits when traveling along the lane-changing trajectory at the desired speed profile.

[0025] In one embodiment of this application, the interaction model adopts a Level-K cognitive hierarchy model, in which the vehicle is the decision-maker at the Kth level, used to predict vehicles in the target lane; where K is a positive integer.

[0026] In one embodiment of this application, the longitudinal velocity and acceleration adapted to the lateral movement are planned according to the target lane-changing scheme, including:

[0027] The desired target speed during the lane-changing process is determined based on the traffic flow speed of the target lane, the dynamic change rate of the safety gap, and the lane-changing strategy.

[0028] A reference longitudinal acceleration profile is generated to guide the vehicle's speed to the desired target speed.

[0029] In one embodiment of this application, the method further includes monitoring the vehicle's environment and avoiding obstacles;

[0030] The monitoring of the vehicle's environment and obstacle avoidance includes:

[0031] The reference longitudinal acceleration profile is used as the target acceleration input for the solver.

[0032] Projecting a spatiotemporal diagram onto the obstacles generates corresponding constraint boundaries, which serve as the longitudinal boundaries for the solution.

[0033] Weights are assigned based on the interaction scenario, and acceleration optimization calculations are performed using the solver.

[0034] In one embodiment of this application, generating and executing a lane-changing action includes:

[0035] The vehicle status and environmental perception information are acquired, the current lane-changing trajectory is optimized, and a target execution trajectory synchronized with the currently planned longitudinal velocity profile is generated.

[0036] The target execution trajectory is decomposed into vehicle control commands for the vehicle itself;

[0037] The lane-changing action is executed according to the vehicle control command, the actual execution trajectory of the vehicle is tracked, and the vehicle control command is adjusted to compensate for the deviation between the actual execution trajectory and the target execution trajectory.

[0038] The system continuously monitors the vehicle's status and the environmental perception information, and aborts the lane-changing action if the lane-changing conditions are not met.

[0039] Secondly, embodiments of this application provide a lane-changing system, including:

[0040] The instruction parsing module is used to receive lane change instructions and generate lane change intentions based on the lane change instructions; the lane change instructions include the target lane.

[0041] The trajectory planning module, connected to the instruction parsing module, is used to filter out safe gaps in the target lane that can be used for merging based on the movement state of vehicles in the target lane, plan lane-changing trajectories based on the safe gaps, and obtain several initial lane-changing schemes.

[0042] The interaction prediction module, connected to the trajectory planning module, is used to predict the reaction of the interaction object when the vehicle executes any of the initial lane change schemes based on the interaction model between the vehicle and the interaction object in the target lane, and generate the corresponding interaction scenario.

[0043] The risk assessment module, connected to the interaction prediction module, is used to obtain the risk utility value of each interaction scenario and determine the target lane change scheme from the initial lane change scheme based on the risk utility value.

[0044] The lane change execution module is connected to the risk assessment module. Based on the target lane change plan, it plans the longitudinal speed and acceleration that are adapted to the lateral movement, and generates and executes the lane change action.

[0045] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the lane-switching method.

[0046] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the lane-switching method.

[0047] The beneficial effects of this application are: Based on the real-time motion state of vehicles in the target lane, this application selects safe gaps, predicts the possible reactions of the interactive objects in the target lane by establishing an interaction model between the vehicle and the interactive object, generates risk utility values, and determines the target lane-changing scheme based on the quantified risk utility values, preventing decisions from being too aggressive or conservative, achieving a balance between safety and efficiency, and greatly improving traffic efficiency while ensuring safety. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1This is a schematic diagram of the lane-changing method according to an embodiment of this application;

[0050] Figure 2 This is a flowchart of a lane-changing method according to an embodiment of this application;

[0051] Figure 3 This is a schematic diagram illustrating the steps of generating a lane-changing intention according to an embodiment of this application;

[0052] Figure 4 This is a schematic diagram illustrating the steps of screening safety gaps according to an embodiment of this application;

[0053] Figure 5 This is a schematic diagram illustrating the steps of performing a lane-changing operation according to an embodiment of this application;

[0054] Figure 6 This is an architectural diagram of the lane-changing system according to an embodiment of this application.

[0055] Explanation of reference numerals in the attached figures:

[0056] 1. Instruction parsing module; 2. Trajectory planning module; 3. Interactive prediction module; 4. Risk assessment module; 5. Lane change execution module. Detailed Implementation

[0057] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. In addition, it should be understood that the specific embodiments described herein are only for illustration and explanation of this application and are not intended to limit this application. In this application, unless otherwise stated, directional terms such as "up," "down," "left," and "right" generally refer to up, down, left, and right in the actual use or working state of the device, specifically the drawing directions in the accompanying drawings.

[0058] In this application, unless otherwise expressly specified and limited, the terms "connected," "linked," "stacked," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two elements or the interaction between two elements. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0059] The specific implementation methods of this application are illustrated below through examples:

[0060] One lane-changing method, such as Figure 1 , Figure 2As shown, it includes:

[0061] Step S1: In response to the received lane change instruction, generate a lane change intention based on the lane change instruction, which includes the target lane.

[0062] Step S2: Based on the movement status of vehicles in the target lane, select safe gaps in the target lane that can be used for merging, plan lane-changing trajectories based on the safe gaps, and obtain several initial lane-changing schemes.

[0063] Step S3: Based on the interaction model between the vehicle and the target lane's interaction object, predict the reaction of the interaction object when the vehicle executes any initial lane change plan, and generate the corresponding interaction scenario.

[0064] Step S4: Obtain the risk utility value for each interaction scenario, and determine the target lane-changing scheme from the initial lane-changing scheme based on the risk utility value.

[0065] Step S5: Based on the target lane change plan, plan the longitudinal velocity and acceleration that are compatible with the lateral movement, generate and execute the lane change action.

[0066] In one alternative embodiment, such as Figure 3 As shown, step S1 of this application generates a lane-changing intention based on the lane-changing instruction, including:

[0067] Step S11: Extract lane change information from the lane change command. The lane change information includes the direction of the target lane, the purpose of the lane change, and the lane change constraints.

[0068] Step S12: Based on the lane change information, generate a lane change intention that includes the target lane identifier and a preliminary assessment of the timing of the lane change operation.

[0069] Specifically, the lane-changing process in this application is triggered by a lane-changing command issued by a navigation system or an advanced driver assistance system. The lane-changing command includes target lane information and geographical or temporal constraints for the lane-changing operation, such as the distance to the target exit or a suggested execution time window.

[0070] More specifically, lane change information is extracted from the lane change instruction, including the direction of the target lane, the reason or purpose of the lane change, and the constraints for performing the lane change operation; the direction of the target lane is specifically, for example, left or right, the reason or purpose of the lane change is specifically, for example, entering a ramp or avoiding an obstacle, and the constraints for performing the lane change operation are specifically, for example, "starting at a certain distance of a certain meter" or "completing within a certain number of seconds".

[0071] Step S12 generates an initial lane change intention based on the parsed lane change information. The lane change intention includes the target lane identifier and a preliminary timing assessment of the lane change operation. The lane change intention is used to indicate that attention resources should be preferentially allocated to the target lane direction.

[0072] In an optional embodiment, step S2 detects the motion state of vehicles in the target lane corresponding to the lane-changing intention and filters out safe gaps in the target lane that can be used for merging, such as... Figure 4 As shown, it includes:

[0073] Step S21: Obtain the motion status of at least two vehicles in the target lane;

[0074] Step S22: Calculate the physical gap between adjacent vehicles in the target lane;

[0075] Step S23: Evaluate each physical gap using preset safety indicators, and select the physical gap that meets the safety indicators as the safe gap.

[0076] Step S2 of this application uses onboard sensors, specifically such as lidar, millimeter-wave radar, cameras, and their fusion systems, to accurately perceive and model the target lane and its surrounding environment in order to identify and assess feasible lane change gaps.

[0077] Within the target lane object sequence, identify all potential physical gaps. A physical gap is defined as the longitudinal distance between the rear of the vehicle in front and the front of the vehicle behind in the target lane, minus the length of the vehicle itself, expressed as:

[0078] ;

[0079] Among them, S gap Indicates physical gap, x rear This represents the head-on coordinates of the vehicle behind in the target lane, x. front This indicates the coordinates of the rear of the vehicle in front in the target lane, and PFL indicates the length of the vehicle itself.

[0080] Compared to schemes that rely on fixed rules, this application detects the motion state of vehicles in the target lane in step S2. The safety gap obtained by filtering is not a fixed value, but is obtained based on the real-time motion state of vehicles in the target lane. The safety gap adapts to the time-varying characteristics of traffic flow, avoiding conservative decisions caused by static parameters. This avoids the problem of excessive defense such as "awkward stopping" and "unnecessary yielding" that easily triggers when faced with the random behavior of human driving vehicles, as is the case with fixed rule schemes that use conservative parameters. This leads to a loss of traffic efficiency.

[0081] In one alternative embodiment, the security indicators include spatial indicators, temporal indicators, and disturbance indicators;

[0082] The space index is used to assess whether the physical clearance meets the space safety requirements for vehicle merging. The space safety requirements of this application are to determine whether the physical clearance meets the vehicle length plus a preset minimum safety margin. If the physical clearance is greater than or equal to the vehicle length plus the preset minimum safety margin, it is determined that the space index assessment is met; if the physical clearance is less than the vehicle length plus the preset minimum safety margin, it is determined that the space index assessment is not met, and the physical clearance is discarded.

[0083] The time metric is used to assess whether the distance between the vehicle and the vehicle in front and / or behind in the target lane meets the time safety requirements for collision avoidance. The time metric judges from a time perspective whether there is enough reaction time between the vehicle and the vehicle in the target lane to deal with emergencies and avoid instantaneous collisions caused by excessive relative speed. The parameters of the time metric include the time interval and the collision time. The time interval is divided into the time interval with the vehicle in front and the time interval with the vehicle behind, which is calculated by the formula of distance to the vehicle in front or behind divided by the vehicle speed. The collision time refers to the time required for the vehicle to collide with the vehicle in the target lane if it maintains its current state of motion. The time metric requires the collision time to be greater than a preset safety threshold. If it does not meet the threshold, the physical clearance is discarded.

[0084] The disturbance index is used to assess whether the disturbance caused by a merging vehicle to vehicles in the target lane meets the disturbance safety requirements. It considers the current speed and acceleration of the target lane vehicle and assesses whether the merging vehicle will cause it to decelerate drastically or result in an unsafe interaction. If the vehicle behind in the target lane needs to decelerate significantly, exceeding a preset deceleration threshold, to avoid a collision, the dynamic acceptability of the physical clearance is low. If, after the merging vehicle, vehicles in the target lane do not need to significantly adjust their driving status to maintain a safe distance, the acceptability is high.

[0085] This application evaluates the above three safety indicators based on the multi-criteria decision method (MCDM) or preset hard safety constraints, and selects one or more physical gaps that meet the requirements as safety gaps.

[0086] In one optional embodiment, planning a lane-changing trajectory based on a safety gap includes: using a parametric curve generation method to generate a trajectory connecting the current state of the vehicle with a target point in a selected safety gap within the target lane, thereby obtaining the lane-changing trajectory.

[0087] Specifically, step S2 generates a geometrically smooth trajectory that connects the current state of the vehicle with the target lane gap for the selected safety gap. This application uses parametric curves such as fifth-order polynomials, B-splines, and spirals. These parametric curves have high-order continuity, which can avoid severe vehicle shaking or difficulty in handling caused by abrupt trajectory changes. For example, a fifth-order polynomial is used to describe the trajectory through the functional relationship between lateral displacement y (unit: meters) and longitudinal displacement x (unit: meters), the expression of which is:

[0088] ;

[0089] The coefficients c0, c1, c2, c3, c4, and c5 are determined by boundary conditions and must satisfy the lateral position, lateral velocity, and lateral acceleration constraints of the current lane's vehicle position and the safety gap. For example, the current lane's lateral velocity is 0, and the safety gap's lateral velocity matches the traffic flow direction of the target lane to ensure a smooth transition of the lane-changing trajectory from the current state to the target state. Specific data examples:

[0090] Scenario assumption:

[0091] Longitudinal lane change distance L1 = 40 meters. Lateral lane change distance L2 = 3.5 meters.

[0092] To ensure a smooth transition of the trajectory vehicle from its current state to the target location, this embodiment sets the following boundary conditions:

[0093] At the starting position, the vehicle's lateral and longitudinal displacements are both zero. , Furthermore, the vehicle's lateral velocity and lateral acceleration are both zero. , At the finish line, the vehicle's lateral displacement was 3.5 meters and its longitudinal displacement was 40 meters. Furthermore, the vehicle's lateral velocity and lateral acceleration are both zero. , .

[0094] Solving for the coefficients: From the initial conditions, we get C0 = C1 = C2 = 0. The remaining coefficients are obtained from the final conditions.

[0095] ;

[0096] ;

[0097] ;

[0098] Final trajectory equation:

[0099] .

[0100] In an optional embodiment, step S2 further includes: verifying the obtained lane-changing trajectory and selecting the verified lane-changing trajectories as the initial lane-changing scheme; wherein, verifying the obtained lane-changing trajectory includes:

[0101] Verify whether the lane-changing trajectory meets the vehicle's motion restrictions; and / or;

[0102] Verify whether the required lateral and longitudinal accelerations meet the vehicle's dynamic limits when traveling along the lane-changing trajectory at the desired speed profile.

[0103] Specifically, motion limits include, for example, the maximum steering wheel angle and the minimum turning radius. Verifying whether the lane-changing trajectory meets the vehicle's motion limits means verifying whether the vehicle can complete the geometric motion of the trajectory, that is, verifying whether the lane-changing trajectory conforms to the physical limits of the vehicle's steering and motion mechanisms, and avoiding planning lane-changing trajectories that are actually impossible to achieve.

[0104] More specifically, the rotation angle of the vehicle's steering motor has a physical upper limit. If the steering wheel angle corresponding to the curvature of a point on the lane-changing trajectory exceeds this upper limit, the lane-changing trajectory is determined to be invalid. Correspondingly, due to wheelbase and steering structure limitations, the vehicle has a minimum turning radius. If the curvature of the lane-changing trajectory exceeds the curvature corresponding to the minimum turning radius, the vehicle will be unable to travel along the trajectory due to insufficient steering. This application can calculate the curvature and required steering wheel angle of each point on the lane-changing trajectory and compare them with the vehicle motion limitation parameters. If all trajectory points meet the motion limitations, the lane-changing trajectory passes the motion limitation verification.

[0105] Specifically, the power limits include, for example, tire-road adhesion limits and engine / motor output capabilities. Verifying whether the required lateral and longitudinal accelerations meet the vehicle's power limits when traveling along the lane-changing trajectory at the desired speed profile is to verify whether the vehicle can travel along the trajectory at the desired speed, i.e., to verify whether the force and acceleration required for the lane-changing trajectory are within the capabilities of the vehicle's power system. This application combines the desired speed profile, such as the expected speed through the safety gap, to calculate the lateral / longitudinal acceleration and rate of change of acceleration at each point when traveling along the trajectory, and compares it with vehicle dynamic parameters, such as adhesion limits and power output. If all parameters meet the power limits, then the lane-changing trajectory passes the power limit verification.

[0106] More specifically, this application, while conducting power limitation verification, also considers ride comfort, calculating the lateral / longitudinal acceleration and rate of change of acceleration at each point along the trajectory. In addition to comparing with vehicle dynamics parameters, it also compares with comfort thresholds. If the comfort threshold is not met, it is considered that the power limitation verification has not been passed. This application verifies whether the force and acceleration required for the trajectory are within the capabilities of the vehicle's power system while taking into account passenger comfort.

[0107] In an optional embodiment, the interaction model in step S3 adopts a Level-K cognitive hierarchy model, where the vehicle is the decision-maker at level K (K≥1), used to predict vehicles in the target lane. The vehicle can autonomously infer the behavioral logic of the interaction object, and the inference depth increases with the value of K. The interaction object is regarded as a Level-(K-1) or Level-0 decision-maker: Level-0 means that the interaction object's behavior is random, such as driving according to fixed rules, without any intention to infer; Level-(K-1) means that the interaction object will infer the vehicle's behavior with a cognition level lower than that of the vehicle. Through hierarchical game theory, the vehicle can predict the possible reaction of the interaction object to its lane-changing action, such as whether the following vehicle will slow down and give way because the vehicle cuts in, avoiding decision-making bias caused by ignoring the randomness of human driving, and achieving formal safety in extremely small spaces. Compared with reinforcement learning technology, it can perform stably in sparse or complex scenarios, balancing safety and efficiency. Under the premise of flexible planning results, it can stably control the performance in congested road conditions, avoiding overly aggressive or overly conservative situations. Furthermore, the calculation method and process presented in this application have low implementation costs and do not require a large amount of computing power and data to be wasted.

[0108] More specifically, this application predicts the behavioral response of interactive objects, such as vehicles behind in the target lane, based on the assumption that the vehicle is executing a certain lane-changing trajectory. The prediction results can be obtained through a predefined driving behavior model library, such as the typical reaction pattern of human drivers when they are cut off, or through online learning, such as updating the behavioral preferences of interactive objects in real time.

[0109] Possible responses from the interactive object include, but are not limited to:

[0110] Slow down and yield: The interactive object reduces its speed and increases the safety gap;

[0111] Maintain speed: The interactive object maintains its current driving state and does not actively avoid or stop other objects;

[0112] Accelerated blocking: Interactive objects are accelerated, and safety gaps are compressed.

[0113] In one optional embodiment, step S3 evaluates the merits of different lane-changing schemes for each interaction scenario in which the vehicle executes a certain lane-changing trajectory by quantifying risk-utility values, and finally selects the optimal target lane-changing scheme.

[0114] The risk utility value, which is the quantitative value used in this application to balance safety and efficiency, includes: collision probability, degree of safety margin violation, lane change completion time, impact on traffic flow, and control input smoothness, among which:

[0115] Collision probability represents the likelihood of a collision between the vehicle and an interactive object;

[0116] The degree of safety margin violation represents the quantification of the degree of violation when the distance between the vehicle and the interaction object after merging is less than the safety margin.

[0117] Lane change completion time indicates the duration from the start of the lane change to fully entering the target lane;

[0118] The impact on traffic flow is represented by the quantitative result of the merging of vehicles causing the target lane's traffic flow to slow down and congestion to worsen.

[0119] Control input smoothness represents the quantified result of the control smoothness of vehicle actions during lane changing.

[0120] This application selects the optimal action. The goal is to maximize expected utility or minimize expected risk / cost, expressed as:

[0121] ;

[0122] in, Let A represent the optimal action, E represent the set of possible actions for the vehicle, and P(θ) represent the expected action. -i |s,H) represents the current state And historical observations H for other agent behavior models θ -i The probability distribution, U(s,a,θ) -i ) represents the utility function, used to quantify the actions of the vehicle (s) and other agents (θ) in the current state (s), including the vehicle's action (a) and the behavioral model of the other agents. -i In the scenario, the overall benefit of the lane-changing behavior of the autonomous vehicle is represented by argmax, which represents maximizing utility. In addition to using maximizing utility argmax to select the optimal action, this application can also replace it with minimizing cost argmin for calculation. The actions that the autonomous vehicle can perform include specific trajectory, adjusting parameters, and abandoning lane changing. Based on the evaluation results, this application selects the lane-changing scheme with acceptable risk and optimal utility, or decides to suspend / delay lane changing.

[0123] In one specific embodiment, a specific scenario of this application is as follows:

[0124] Current status of the vehicle This includes the vehicle's position and speed. The vehicle's position is located at the center line of the lane, meaning its lateral displacement is 0 meters. The vehicle's speed is 10 meters per second, and its longitudinal distance from the vehicle behind it in the target lane is 5 meters.

[0125] Historical data shows that, based on the current status of the vehicle, there is a 70% probability that the driver will slow down and yield when a vehicle behind cuts in (behavior θ). -i,1 ), 30% probability of accelerating prevention (behavior θ) -i,2 ), that is, P(θ) -i,1 |s,H)=0.7,P(θ) -i,2|s,H)=0.3.

[0126] In the set of optional actions A, the vehicle has two actions 'a' to perform: rapid cut-in a1 and slow cut-in a2; this is achieved by defining the utility function U(s,a, θ). -i This application integrates the comprehensive benefits from three dimensions: safety, efficiency, and comfort, each with different weights. The weight distribution is assumed to be: safety 0.5, efficiency 0.3, and comfort 0.2. For each action, the utility is calculated under two scenarios: the following vehicle slows down to yield, and the following vehicle accelerates to stop. The expected utility is then calculated using probability weighting. The expected utility of the three actions is compared, and the action with the highest expected utility is selected as the optimal action. .

[0127] To be more specific, let's assume that for a rapid cut into lane a1, the lane change time is 3 seconds and the lateral acceleration is 0.8 m / s². 2 If, after changing lanes, the distance between you and the following vehicle is 8 meters (with the following vehicle slowing down), which is greater than or equal to the safe distance of 5 meters, the safety score is 10 points, and the efficiency score is 10 points, assuming the lateral acceleration exceeds 0.5 m / s². 2 After that, for every additional 0.1 m / s 2 Deduct 2 points, the comfort score is 4 points, and after weighting, the utility U value is 8.8 points;

[0128] With the following vehicle accelerating, the distance between the two vehicles after lane change is 3 meters, which is less than the safe distance of 5 meters. The safety score is 6 points. Since the lane change time and lateral acceleration remain unchanged, the efficiency score is still 10 points, the comfort score is still 4 points, and the utility score is 6.8 points. Therefore, the expected utility for the action of quickly cutting into a1 is 8.2.

[0129] For a slow incursion into lane a2, the lane change time is 6 seconds and the lateral acceleration is 0.3 m / s². 2 If the following vehicle decelerates, the distance between the two vehicles after lane changing is 8 meters; if the following vehicle accelerates, the distance is 3 meters. Using the same calculation process as the rapid lane-changing action a1, the expected utility is calculated to be 8.8. Therefore, the slow lane-changing action a2 is selected as the optimal action. .

[0130] In one optional embodiment, step S4 plans a longitudinal velocity and acceleration adapted to the lateral movement according to the target lane-changing scheme, including:

[0131] The desired target speed during the lane-changing process is determined based on the traffic flow speed in the target lane, the dynamic rate of change of the safety gap, and the lane-changing strategy.

[0132] Generate a reference longitudinal acceleration profile to guide the vehicle's speed to the desired target speed.

[0133] Specifically, this application determines the desired target speed or speed range during the lane-changing process based on the traffic flow speed of the target lane, the dynamic characteristics of the selected gap, and the lane-changing strategy, such as smooth merging or merging after accelerating overtaking.

[0134] By generating a reference longitudinal acceleration profile, the vehicle's speed is guided to smoothly transition to the target speed, ensuring a safe relative position with vehicles in the target lane. This application employs an Intelligent Driver Model (IDM) or other advanced car-following models and optimization algorithms to calculate the reference longitudinal acceleration profile. A typical formula for IDM is:

[0135] ;

[0136] in, This represents the calculated longitudinal reference acceleration (unit: meters per second). 2 ); This indicates the vehicle's maximum acceleration (unit: meters per second). 2 The value range is 1.0 to 3.0; Indicates the vehicle's current speed (unit: meters per second); Indicates the desired cruising speed (unit: meters per second); This represents the acceleration exponent (dimensionless), which is usually a fixed value, such as 4; Indicates the actual net distance to the vehicle in front (unit: meters); This represents the minimum parking distance (in meters), with a value ranging from 1.5 to 5.0. This represents the safe interval (unit: seconds), with a value range of 1.0 to 2.2. Indicates comfortable deceleration (unit: meters per second) 2 The value range is 1.5 to 3.0; This indicates the speed difference between the vehicle and the vehicle in front (unit: meters per second).

[0137] To take current speed into account Speed ​​difference with the vehicle in front The expected safe distance (in meters) is greater than or equal to the minimum parking distance. Its specific form is as follows.

[0138] ;

[0139] The calculated acceleration profile will be used as an input reference for longitudinal control.

[0140] In one specific embodiment, assume the vehicle's current speed Speed ​​of the vehicle ahead The actual net distance between the vehicle and the vehicle in front acceleration index 4; Minimum parking distance ; Safety time interval Comfort deceleration Expected cruising speed Maximum vehicle acceleration .

[0141] Calculation steps:

[0142] Step 1: Calculate the speed difference between your vehicle and the vehicle in front. ;

[0143] ;

[0144] Step 2: Calculate the desired dynamic safety distance ;

[0145] ;

[0146] Step 3: Calculate the final longitudinal acceleration ;

[0147] .

[0148] Based on the given operating conditions (vehicle speed 20 m / s, preceding vehicle speed 18 m / s, actual distance between vehicles 35 m), this application uses the IDM model to calculate a recommended longitudinal acceleration of -1.27 m / s². 2 The results show that, in typical close-range following scenarios, this application can generate accurate, reasonable, and safe longitudinal control strategies that balance safety and comfort, effectively simulating the decision-making of excellent human drivers.

[0149] In one alternative embodiment, the lane-changing method further includes monitoring the vehicle's environment and avoiding obstacles;

[0150] Monitor the vehicle's environment and avoid obstacles, including:

[0151] Use the reference longitudinal acceleration profile as the target acceleration input for the solver;

[0152] Projecting a spatiotemporal diagram onto the obstacles generates corresponding constraint boundaries, which serve as the longitudinal boundaries for the solution.

[0153] Weights are assigned based on the interaction scenario, and acceleration optimization calculations are performed using a solver.

[0154] Specifically, step S4, while performing longitudinal velocity planning, also continuously monitors the wide-area environment to ensure that collisions with dynamic or static obstacles are avoided under any circumstances. This application has planned and obtained a reference longitudinal acceleration profile to guide the smooth change of vehicle speed, but in actual road conditions, there are dynamic or stationary obstacles, requiring additional optimization.

[0155] In addition to ensuring that collisions with obstacles are avoided, the optimization direction of this application is to flexibly adjust the longitudinal control strategy in complex scenarios, such as narrow spaces and densely packed obstacles, to prioritize passage while trying to match the reference acceleration as closely as possible.

[0156] Specifically, the solver in this application employs a quadratic programming (QP) solver. The essence of longitudinal control is optimizing acceleration / velocity under constraints. As a mathematical optimization tool, QP can solve for the optimal longitudinal control quantity and output the actual acceleration under constraints such as no collision and maximum acceleration. The reference objective of the QP optimization in this application is to make the actual control as close as possible to the acceleration curve of the preceding programming, that is, to minimize the deviation between the actual acceleration and the reference acceleration, while satisfying other constraints.

[0157] Specifically, this application projects a spatiotemporal diagram (ST diagram) onto other obstacles to generate corresponding soft and hard boundaries, which serve as the upper and lower boundaries of the longitudinal axis for solving. The spatiotemporal diagram associates the position (space, S-axis) of the obstacle with time (T-axis), predicts the trajectory of the obstacle over a period of time, and projects it onto the spatiotemporal diagram to form hard boundaries that the vehicle cannot enter and soft boundaries that the vehicle should try to avoid entering.

[0158] This application differentiates between scenarios and assigns weights to different scenarios. In this case, it tends to assign a higher weight to the reference longitudinal acceleration profile to reduce the penalty for exceeding the soft boundary and ensure that longitudinal driving space can be obtained in a narrow space.

[0159] In one alternative embodiment, a lane-changing action is generated and executed, such as... Figure 5 As shown, it includes:

[0160] Step S41: Obtain vehicle status and environmental perception information, optimize the current lane-changing trajectory, and generate a target execution trajectory that is synchronized with the currently planned longitudinal velocity profile.

[0161] Step S42: Decompose the target execution trajectory into vehicle control commands for the vehicle.

[0162] Step S43: Execute lane-changing action according to vehicle control command, track the actual execution trajectory of the vehicle, and adjust vehicle control command to compensate for deviation based on the deviation between the actual execution trajectory and the target execution trajectory.

[0163] Step S44: Continuously monitor the vehicle's status and environmental perception information. If the lane-changing conditions are not met, stop the lane-changing action.

[0164] Specifically, provided that all prerequisites are met and safety is guaranteed, the final lane-changing action is generated and executed in step S4.

[0165] First, based on the latest vehicle status and environmental perception information, especially the final confirmation of the target gap, the selected initial trajectory is adjusted and optimized as necessary, or a final execution trajectory that is precisely synchronized with the currently planned longitudinal velocity profile is regenerated.

[0166] The final spatiotemporal trajectory planned by the high-level system, including the changes in lateral position, longitudinal position / velocity, and orientation over time, is decomposed into executable low-level control commands for the vehicle's underlying actuators. Specific examples of these underlying actuators include the steering motor, drive motor, and braking system. Specific examples of these low-level control commands include the target steering wheel angle and the target drive / braking torque.

[0167] More specifically, robust lateral and longitudinal controllers are employed, such as Model Predictive Controller (MPC), Linear Quadratic Regulator (LQR), and Proportional-Integral-Derivative (PID) controllers, to accurately track the generated trajectory and velocity profile, compensate for external disturbances and model uncertainties, and achieve smooth and accurate lane changes.

[0168] During lane change operations, the system continuously monitors vehicle status and the external environment. If a sudden hazard occurs or lane change conditions are no longer met, the system should be able to safely abort the lane change operation (e.g., return to the original lane) or execute other preset emergency plans.

[0169] This application also provides a lane-changing system, such as... Figure 6 As shown, it includes:

[0170] Instruction parsing module 1 is used to receive lane change instructions and generate lane change intentions based on the lane change instructions; the lane change instructions include the target lane;

[0171] The trajectory planning module 2, connected to the instruction parsing module 1, is used to filter out safe gaps in the target lane that can be used for merging based on the movement state of vehicles in the target lane, plan lane-changing trajectories based on the safe gaps, and obtain several initial lane-changing schemes.

[0172] Interaction prediction module 3, connected to trajectory planning module 2, is used to predict the reaction of the interaction object when the vehicle executes any initial lane change plan based on the interaction model between the vehicle and the target lane, and generate the corresponding interaction scenario.

[0173] Risk assessment module 4, connected to interaction prediction module 3, is used to obtain the risk utility value of each interaction scenario and determine the target lane change plan from the initial lane change plan based on the risk utility value.

[0174] The lane change execution module 5 is connected to the risk assessment module 4. Based on the target lane change plan, it plans the longitudinal speed and acceleration that are adapted to the lateral movement, and generates and executes the lane change action.

[0175] This application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a lane-switching method.

[0176] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a lane-switching method.

[0177] The above provides a detailed description of a lane-switching method and system, computer equipment, and storage medium provided by this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A lane-changing method, characterized in that, include: In response to a received lane change instruction, a lane change intention is generated based on the lane change instruction, the lane change instruction including the target lane; Based on the movement state of vehicles in the target lane, safe gaps in the target lane that can be used for merging are selected, and lane-changing trajectories are planned based on the safe gaps to obtain several initial lane-changing schemes. Based on the interaction model between the vehicle and the target lane, the system predicts the reaction of the interaction object when the vehicle executes any of the initial lane-changing schemes, and generates a corresponding interaction scenario. The interaction model adopts the Level-K cognitive hierarchy model, in which the vehicle is the decision-maker at the Kth level, used to predict the vehicles in the target lane. The autonomous vehicle, through hierarchical game theory, predicts the possible reactions of the interacting objects to its lane-changing actions and infers the behavioral logic of the interacting objects; where K is a positive integer; Obtain the risk utility value for each of the aforementioned interaction scenarios, and determine the target lane-changing scheme from the initial lane-changing scheme based on the risk utility value; Based on the target lane-changing scheme, a longitudinal velocity and acceleration adapted to the lateral movement are planned, and a lane-changing action is generated and executed.

2. The lane-changing method according to claim 1, characterized in that, Generate a lane change intention based on the lane change command, including: Extract lane change information from the lane change command, the lane change information including the direction of the target lane, the purpose of lane change, and the lane change constraints; Based on the lane change information, a lane change intention is generated that includes a target lane identifier and an initial assessment of the timing of the lane change operation.

3. The lane-changing method according to claim 1, characterized in that, Detect the motion state of vehicles in the target lane corresponding to the lane-changing intention, and filter out safe gaps in the target lane that can be used for merging, including: Obtain the motion states of at least two vehicles within the target lane; Calculate the physical clearance between adjacent vehicles within the target lane; Each physical gap is evaluated using preset safety indicators, and the physical gap that meets the safety indicators is selected as the safe gap.

4. The lane-changing method according to claim 1, characterized in that, Planning a lane-changing trajectory based on the aforementioned safety gap includes: A parametric curve generation method is used to generate a trajectory connecting the current state of the vehicle with the target point in the selected safety gap within the target lane, thus obtaining the lane-changing trajectory.

5. The lane-changing method according to claim 4, characterized in that, It also includes: verifying the obtained lane-changing trajectory, and selecting the verified lane-changing trajectory as the initial lane-changing scheme; Verification of the obtained lane-changing trajectory includes: Verify whether the lane-changing trajectory meets the vehicle's motion restrictions; and / or; Verify whether the required lateral and longitudinal accelerations meet the vehicle's dynamic limits when traveling along the lane-changing trajectory at the desired speed profile.

6. The lane-changing method according to claim 1, characterized in that, Based on the target lane-changing scheme, the longitudinal velocity and acceleration adapted to the lateral movement are planned, including: The desired target speed during the lane-changing process is determined based on the traffic flow speed of the target lane, the dynamic change rate of the safety gap, and the lane-changing strategy. A reference longitudinal acceleration profile is generated to guide the vehicle's speed to the desired target speed.

7. The lane-changing method according to claim 6, characterized in that, The method also includes monitoring the vehicle's environment and avoiding obstacles; The monitoring of the vehicle's environment and obstacle avoidance includes: The reference longitudinal acceleration profile is used as the target acceleration input for the solver. Projecting a spatiotemporal diagram onto the obstacles generates corresponding constraint boundaries, which serve as the longitudinal boundaries for the solution. Weights are assigned based on the interaction scenario, and acceleration optimization calculations are performed using the solver.

8. The lane-changing method according to claim 1, characterized in that, Generate and execute lane-changing actions, including: The vehicle status and environmental perception information are acquired, the current lane-changing trajectory is optimized, and a target execution trajectory synchronized with the currently planned longitudinal velocity profile is generated. The target execution trajectory is decomposed into vehicle control commands for the vehicle itself; The lane-changing action is executed according to the vehicle control command, the actual execution trajectory of the vehicle is tracked, and the vehicle control command is adjusted to compensate for the deviation between the actual execution trajectory and the target execution trajectory. The system continuously monitors the vehicle's status and the environmental perception information, and aborts the lane-changing action if the lane-changing conditions are not met.

9. A lane-changing system, characterized in that, include: The instruction parsing module is used to receive lane change instructions and generate lane change intentions based on the lane change instructions; The lane change instruction includes the target lane; The trajectory planning module, connected to the instruction parsing module, is used to filter out safe gaps in the target lane that can be used for merging based on the movement state of vehicles in the target lane, plan lane-changing trajectories based on the safe gaps, and obtain several initial lane-changing schemes. The interaction prediction module, connected to the trajectory planning module, is used to predict the reaction of the interaction object when the vehicle executes any of the initial lane-changing schemes, based on the interaction model between the vehicle and the interaction object in the target lane, and generate the corresponding interaction scenario; the interaction model adopts the Level-K cognitive hierarchy model, in which the vehicle is the decision-maker at the Kth level, and is used to predict the vehicles in the target lane. The autonomous vehicle, through hierarchical game theory, predicts the possible reactions of the interacting objects to its lane-changing actions and infers the behavioral logic of the interacting objects; where K is a positive integer; The risk assessment module, connected to the interaction prediction module, is used to obtain the risk utility value of each interaction scenario and determine the target lane change scheme from the initial lane change scheme based on the risk utility value. The lane change execution module is connected to the risk assessment module. Based on the target lane change plan, it plans the longitudinal speed and acceleration that are adapted to the lateral movement, and generates and executes the lane change action.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the lane-changing method as described in any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the lane-changing method as described in any one of claims 1 to 8.

Images