A hierarchical AEB braking method based on game theory in scenarios where other vehicles cut in

Abstract

This invention discloses a game-theory-based AEB graded braking method for scenarios involving other vehicles entering the vehicle's path. The method includes the following steps: Step 1: Acquiring state information data about the vehicle and its surroundings, determining the heading angle of vehicles in adjacent lanes and the distance from the adjacent lane vehicle to the lane line adjacent to the vehicle; Step 2: Predicting the probability of lateral behavior intentions of vehicles in adjacent lanes based on a Gaussian mixture hidden Markov model; Step 3: Determining whether to establish a longitudinal safety distance model based on the state of the vehicles in the adjacent lanes; Step 4: Establishing a payoff matrix based on a non-cooperative Nash game; Step 5: Designing a payoff function and deriving an optimal strategy solution based on a Nash equilibrium. The AEB system then performs graded braking according to the optimal strategy; Step 6: Upon determining that there is no potential danger between the vehicle and the adjacent lane vehicle, the AEB system deactivates. This method avoids potential risks and collisions with vehicles in adjacent lanes, improving driving safety and comfort, and resolving the issues of untimely braking or false triggering in conventional AEB systems.

Description

Technical Field

[0001] The present invention belongs to the technical field of driving safety, and in particular relates to an AEB graded braking method in a scenario where other vehicles cut in based on game theory. Background Art

[0002] Autonomous Emergency Braking (AEB), a representative of advanced automotive safety assistance systems, uses onboard sensing devices such as LiDAR and cameras to identify obstacles and other vehicles posing potential risks ahead, thus applying emergency braking. The timely intervention of the AEB system can significantly improve driving safety.

[0003] Lane changing is an essential and potentially dangerous driving behavior in road traffic. However, due to numerous factors, including the complexity of road conditions, the uncertainty of human drivers' psychological states, and the diversity of their driving styles, human drivers exhibit significant variability in their lane change behavior. For example, on urban expressways, a vehicle in an adjacent lane may shift from the side of a cut-off vehicle to its lane at a higher speed and a lower heading angle, or may forcefully change lanes to the side of a cut-off vehicle at a lower speed and a higher heading angle. In these situations, the cut-off vehicle, acting as the cut-off vehicle, must predict the driving intentions of the vehicle in the adjacent lane. Furthermore, drivers may experience brief lane departures and unclear lane-changing behaviors due to factors such as driver distraction and nervousness.

[0004] These complex lane-changing behaviors significantly interfere with the vehicle's safety system's prediction of the lateral intentions of vehicles in adjacent lanes, easily leading to delayed or inadvertent braking by the AEB system and diminishing trust in the vehicle's safety system. Furthermore, the AEB system's single emergency braking strategy cannot meet the safety requirements of various approach scenarios, significantly reducing passenger comfort. In severe cases, it can even prevent the vehicle behind from reacting, resulting in an unnecessary rear-end collision and negatively impacting driving safety.

[0005] In summary, the lane-changing process of another vehicle occurs through a complex interaction, game-playing, and coupling process between the vehicle being cut off and the vehicle cutting in. However, existing AEB methods are unable to effectively predict the lateral intentions of vehicles in adjacent lanes and implement corresponding collision avoidance strategies. Therefore, estimating the driving intentions of surrounding vehicles during driving, thereby enabling the AEB system to accurately and promptly apply braking measures, is crucial for improving vehicle safety and occupant comfort. To address these existing deficiencies, the present invention proposes a graded AEB braking method for the scenario of another vehicle cutting in, based on game theory. Summary of the Invention

[0006] Purpose of the invention: The present invention proposes an AEB graded braking method in scenarios where other vehicles cut in based on game theory, aiming to achieve the avoidance of dangerous working conditions during vehicle driving, so as to improve the safety in scenarios where other vehicles cut in and the AEB graded braking strategy in different cut-in scenarios; the state information of vehicles in adjacent lanes during driving is obtained through on-board sensors, and the lateral intentions of surrounding vehicles are probabilistically predicted using a hidden Markov model based on a Gaussian mixture, and the probability of lateral behavior intentions is converted into coefficient weights. According to the AEB graded braking strategy of the vehicle itself and the lateral intentions of vehicles in adjacent lanes, a non-cooperative Nash game framework is constructed, and the benefit function is designed considering the driving safety benefits, passenger comfort benefits and negative benefits of risk factors, the benefit matrix is ​​calculated, and the Nash equilibrium is solved to obtain the optimal AEB graded braking strategy.

[0007] Technical solution: A game theory-based AEB graded braking method for scenarios involving other vehicles, including the following steps:

[0008] Step 1: Obtain the status information data of the ego vehicle and the surrounding environment, and calculate the heading angle of the vehicle in the adjacent lane and the distance from the vehicle in the adjacent lane to the lane line on the adjacent side of the ego vehicle;

[0009] Step 2: Based on the obtained data, the lateral maneuver intention of vehicles in adjacent lanes is predicted using a Gaussian mixture hidden Markov model (GMM-HMM). The lateral maneuver characteristics are classified into three types: lane keeping, lane departure, and lane cutting. The lateral maneuver intention predicted by the GMM-HMM is expressed as a probability.

[0010] Step 3: If there is a risk factor for the vehicle in the adjacent lane, that is, the sum of the lane departure probability and the lane cutting probability is greater than the lane keeping probability, a longitudinal safety distance model is established based on the relevant information of the vehicle and the vehicle in the adjacent lane. If there is no risk factor, steps 1 and 2 are repeated.

[0011] Step 4: Establish a non-cooperative Nash game framework, construct the strategy space of the ego vehicle and vehicles in adjacent lanes, and establish a payoff matrix based on the strategy combinations of both parties.

[0012] Step 5: Design a payoff function, assigning weights to the payoff coefficients based on the safety of each strategy. Substitute the resulting values ​​into the payoff matrix and determine the optimal response strategy based on the Nash equilibrium. The AEB system then performs graded braking based on the optimal strategy.

[0013] Step 6: After determining that there is no potential danger between the vehicle and the vehicles in the adjacent lane, the AEB system exits operation.

[0014] The step 1 is specifically as follows:

[0015] The vehicle status and surrounding environment information in step 1 are obtained through GPS and vehicle body sensors;

[0016] The vehicle status information includes: vehicle position information, vehicle speed information and vehicle acceleration information;

[0017] The surrounding environment information includes: adjacent lane vehicle position information, adjacent lane vehicle information, adjacent lane vehicle acceleration information, adjacent lane vehicle heading angle information and lane line position information;

[0018] Among them, the heading angle information of the vehicle in the adjacent lane can be obtained from the speed information, and the expression is:

[0019]

[0020] Where, is the heading angle of the vehicle in the adjacent lane, is the lateral speed of vehicles in the adjacent lane, is the longitudinal speed of vehicles in the adjacent lane;

[0021] Assuming the vehicle is in the middle of the lane, the distance from the vehicle in the adjacent lane to the lane line on the adjacent side of the vehicle can be expressed as:

[0022]

[0023] Where, is the distance from the adjacent lane vehicle to the lane line on the adjacent side of the vehicle, is the lane width, is the lateral position of the vehicle in the adjacent lane, is the lateral position of the vehicle, The width of vehicles in adjacent lanes.

[0024] The step 2 is specifically as follows:

[0025] In step 2, a Gaussian mixture hidden Markov model is used to predict the lateral intention of vehicles in the adjacent lane, and the internal latent state is identified using the external observation sequence. Based on the obtained data, the speed, acceleration, heading angle and distance from the lane line on the adjacent side of the vehicle are selected as the observation state, and the three lateral intentions of vehicles in the adjacent lane (LK), lane departure (LD) and lane cut (LC) are selected as the latent state. The random distribution function of the lateral intention of vehicles in the adjacent lane is established using HMM, which is expressed as a five-tuple. :

[0026]

[0027] The parts of HMM are defined as follows:

[0028] Represents a set of hidden states ,in is the number of states, which represents the set of unobservable states of the lateral intentions of vehicles in adjacent lanes;

[0029] Represents a set of observable sequences ,in is the number of different observations that can be output from each state;

[0030] is the state transition matrix ,in , indicating that the system consists of the state Transfer to The probability of for The state of the moment, , ;

[0031] is the observation probability matrix ,express The probability of the corresponding observation value output in the state, where , for Observe events at all times, , ;

[0032] The initial hidden state is The probability of initializing the state distribution , , ;

[0033] The driving state and driving behavior of the vehicle and the vehicles in the adjacent lanes are continuous time behaviors. The Gaussian mixture model GMM is combined with the hidden Markov model HMM to transform the discrete observation values ​​into continuous states, thereby outputting the probabilities of lane keeping, lane departure and lane cutting of the vehicles in the adjacent lanes. That is, the probability density function of the observation state is represented by the Gaussian mixture model. The state transition matrix in the HMM is: Convert from discrete quantity to continuous quantity through probability density function; is an observable sequence, is an unobservable sequence, then The expected hidden state conditional probability distribution of a sample is:

[0034]

[0035] Add constraints to the above formula , each hidden state can be obtained separately The Gaussian mixture hidden Markov model GMM-HMM can be expressed as:

[0036]

[0037] Then the observation probability matrix Transformed into a probability density function of a set of observations, then given the state Observed values The probability density of is:

[0038]

[0039] The mathematical expression is:

[0040]

[0041]

[0042] Where, for The multi-dimensional Gaussian density function output in the state, is the mean vector, is the covariance matrix, is the data dimension, is the number of Gaussian mixtures, is the Gaussian mixing coefficient weight, and , for In the state The mean of a Gaussian random number generator.

[0043] The step 3 is specifically as follows:

[0044] In step 3, based on the lateral intentions of the vehicles in the adjacent lanes predicted in steps 1 and 2, if the sum of the probabilities of lane departure and lane cutting is greater than the probability of lane keeping, that is, the lateral behavior of the vehicles in the adjacent lanes poses a certain risk to the ego vehicle, the ego vehicle will only move in a straight line within its lane under the action of the AEB system. A safety distance model is established, which is expressed as:

[0045]

[0046] Where, 、 are the distances traveled by the vehicle and the adjacent lane vehicles during the interactive game, 、 are the initial velocities of the vehicle and the adjacent lane vehicles, is the lane-changing time for interactive game, is the acceleration of the vehicle in the adjacent lane, The braking deceleration provided by the vehicle's AEB system;

[0047] The longitudinal distance between the vehicle and the adjacent lane vehicle can be expressed as:

[0048]

[0049] Where, is the longitudinal distance between the two vehicles at the initial moment of lane change;

[0050] Introducing the Berkeley longitudinal safety distance model

[0051]

[0052] Where, is the minimum braking distance, is the relative speed of the two vehicles, is the AEB system response time, is the braking system response time, is the maximum braking deceleration of the two vehicles;

[0053] The longitudinal distance and minimum braking distance between the vehicle and the adjacent lane vehicle should meet the following requirements:

[0054]

[0055] Where, To reserve a safe distance.

[0056] The step 4 is specifically as follows:

[0057] In step 4, based on the non-cooperative Nash game, an interactive game model between the vehicle and the adjacent lane is established under different cut-in scenarios:

[0058]

[0059] Where, Indicates the number of participants. In this invention, the vehicle and the cutting vehicle are considered as game participants. For participants The decision set, For participants The profit function of

[0060] The braking deceleration provided by the ego vehicle's AEB system is selected as the strategy control variable. The interactive game elements between the ego vehicle and the vehicle in the adjacent lane in the lane cutting or lane departure scenario are as follows:

[0061] Participants:

[0062]

[0063] Strategy Space:

[0064] ,

[0065] Profit function:

[0066]

[0067] balanced:

[0068]

[0069] Where, The strategy space consists of all strategy combinations of participants, are the game strategy sets of the interactive game participants, Represent the possible actions of the participants, is the profit function of the set of actions taken by the vehicle for all other participating vehicles in the adjacent lanes, is the Nash equilibrium strategy solution of the interactive game, Represent their respective optimal strategies;

[0070] definition and are the strategy sets for the vehicle and the adjacent lane cutting vehicle, ={no braking, mild braking, emergency braking}, ={lane keeping, lane departure, lane cutting}, and the benefits of the vehicle under the three strategies are expressed as , the benefits of vehicles in adjacent lanes under the three strategies are expressed as The three typical strategies of both parties in the game form nine decision combinations, and the payoff matrix of the ego vehicle and the vehicles in the adjacent lane is established: the payoff matrix of the ego vehicle when the ego vehicle does not brake and the vehicles in the adjacent lane keep the lane is The payoff matrix when the ego vehicle brakes gently and the adjacent lane vehicles keep lane is: ; The payoff matrix of the emergency braking strategy of the ego vehicle and the lane keeping of the adjacent lane vehicle is: ; The payoff matrix when the ego vehicle does not brake and the adjacent lane vehicle deviates is ; The payoff matrix when the ego vehicle brakes gently and the adjacent lane vehicle deviates from the lane is: ; The payoff matrix when the ego vehicle brakes suddenly and the adjacent lane vehicle deviates is: ; The payoff matrix when the vehicle in the adjacent lane cuts in without braking is ; The payoff matrix when the vehicle in the adjacent lane cuts into the lane during gentle braking of the ego vehicle is: ; The payoff matrix when the vehicle in the adjacent lane cuts in during emergency braking is: .

[0071] The step 5 is specifically as follows:

[0072] In step 5, the safety benefit, comfort benefit and loss benefit of dangerous factors of vehicles in adjacent lanes are considered. and the benefit of vehicles cutting into the adjacent lane The current state of the benefits is calculated, including:

[0073] Security benefits The safety benefit takes into account the relative distance and relative speed between the two vehicles, which can be expressed as:

[0074]

[0075] Negative benefit of lateral hazard intention of vehicles in adjacent lanes Taking into account the negative impact of the lateral intention of the vehicle in the adjacent lane on the vehicle, the benefit is designed as follows based on the predicted lateral intention behavior probability:

[0076]

[0077] Where, is the lateral intention probability of vehicles in adjacent lanes predicted by the Gaussian mixture hidden Markov model, Calculate the coefficient for the heading angle of vehicles in adjacent lanes.

[0078] Comfort benefits Use the acceleration change during the game to represent the comfort gain

[0079]

[0080] The total benefits of the interactive game are formed by combining and weighting the ego vehicle and vehicles in adjacent lanes;

[0081]

[0082] Where, The weighted coefficients corresponding to safety benefit, negative benefit of dangerous intention and comfort benefit respectively;

[0083] With the help of the probability estimation and benefit function of the lateral intention of vehicles in adjacent lanes in the above steps, we can get

[0084]

[0085] The ego vehicle evaluates the relative benefits of each strategy in its own strategy set and selects the current optimal strategy. At this time, the AEB system executes the corresponding optimal strategy.

[0086] The step 6 is specifically as follows:

[0087] When a vehicle in an adjacent lane has dangerous lateral behavior that triggers the AEB system, if it is detected that the vehicle's speed is lower than the speed of the oncoming vehicle, the oncoming vehicle's deceleration is lower than the vehicle's AEB mild braking deceleration, and the relative distance of the oncoming vehicle in the direction of the vehicle's lane is greater than the preset AEB brake release distance, the vehicle will be controlled to exit the AEB braking mode and the game interaction process will end.

[0088] Beneficial effects: The present invention addresses the problem that the lateral driving intentions of vehicles in adjacent lanes with similar distances are difficult to predict due to the complexity of human drivers' vehicle psychological factors, differences in driving styles and different road environments, which causes the AEB braking system to respond untimely or be falsely triggered. The method uses a Gaussian mixture hidden Markov model for classification and prediction, and classifies the lateral behavior of vehicles in adjacent lanes into three states and expresses them with probability.

[0089] The present invention establishes the interactive behavior between the vehicle and vehicles in the adjacent lane when cutting in or leaving the lane as a non-cooperative Nash game, designs a profit function and establishes a profit matrix, so that the AEB system can implement the optimal graded braking strategy to improve passenger comfort and driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0091] Figure 1 Flow chart of the method of the present invention.

[0092] Figure 2 Schematic diagram of the scenario where a vehicle cuts into an adjacent lane.

[0093] Figure 3 Schematic diagram of the lane departure scenario of vehicles in adjacent lanes.

[0094] Figure 4 Schematic diagram of the established safety distance model. DETAILED DESCRIPTION

[0095] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0096] In the description of the present invention, it should be understood that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0097] In the present invention, unless otherwise expressly specified or limited, a first feature being "above" or "below" a second feature may include the first and second features being in direct contact, or may include the first and second features being in contact not directly but through another feature between them. Furthermore, a first feature being "above," "above," and "above" a second feature may include the first feature being directly above or obliquely above the second feature, or may simply mean that the first feature is higher in level than the second feature. A first feature being "below," "below," and "below" a second feature may include the first feature being directly below or obliquely below the second feature, or may simply mean that the first feature is lower in level than the second feature.

[0098] like Figure 1 As shown, a game theory-based AEB graded braking method in a scenario where another vehicle cuts in includes the following steps:

[0099] Step 1: Obtain the status information data of the ego vehicle and the surrounding environment, and calculate the heading angle of the vehicle in the adjacent lane and the distance from the vehicle in the adjacent lane to the lane line on the adjacent side of the ego vehicle; specifically:

[0100] The vehicle status and surrounding environment information are obtained through GPS and vehicle body sensors;

[0101] The vehicle status information includes: vehicle position information, vehicle speed information and vehicle acceleration information;

[0102] The surrounding environment information includes: adjacent lane vehicle position information, adjacent lane vehicle information, adjacent lane vehicle acceleration information, adjacent lane vehicle heading angle information and lane line position information;

[0103] Among them, the heading angle information of the vehicle in the adjacent lane can be obtained from the speed information, and the expression is:

[0104]

[0105] Where, is the heading angle of the vehicle in the adjacent lane, is the lateral speed of vehicles in the adjacent lane, is the longitudinal speed of vehicles in the adjacent lane;

[0106] Assuming the vehicle is in the middle of the lane, the distance from the vehicle in the adjacent lane to the lane line on the adjacent side of the vehicle can be expressed as:

[0107]

[0108] Where, is the distance from the adjacent lane vehicle to the lane line on the adjacent side of the vehicle, is the lane width, is the lateral position of the vehicle in the adjacent lane, is the lateral position of the vehicle, The width of vehicles in adjacent lanes.

[0109] Step 2: Based on the obtained data, the lateral behavior intention of vehicles in adjacent lanes is predicted using the Gaussian Mixture Hidden Markov Model (GMM-HMM). The lateral behavior characteristics are summarized into three situations: lane keeping, lane departure, and lane cutting. The lateral intention predicted by the GMM-HMM is expressed as a probability. Specifically:

[0110] In step 2, a Gaussian mixture hidden Markov model is used to predict the lateral intention of vehicles in the adjacent lane, and the internal latent state is identified using the external observation sequence. Based on the obtained data, the speed, acceleration, heading angle and distance from the lane line on the adjacent side of the vehicle are selected as the observation state, and the three lateral intentions of vehicles in the adjacent lane (LK), lane departure (LD) and lane cut (LC) are selected as the latent state. The random distribution function of the lateral intention of vehicles in the adjacent lane is established using HMM, which is expressed as a five-tuple. :

[0111]

[0112] The parts of HMM are defined as follows:

[0113] Represents a set of hidden states ,in is the number of states, which represents the set of unobservable states of the lateral intentions of vehicles in adjacent lanes;

[0114] Represents a set of observable sequences ,in is the number of different observations that can be output from each state;

[0115] is the state transition matrix ,in , indicating that the system consists of the state Transfer to The probability of for The state of the moment, , ;

[0116] is the observation probability matrix ,express The probability of the corresponding observation value output in the state, where , for Observe events at all times, , ;

[0117] The initial hidden state is The probability of initializing the state distribution , , ;

[0118] The driving state and driving behavior of the vehicle and the vehicles in the adjacent lanes are continuous time behaviors. The Gaussian mixture model GMM is combined with the hidden Markov model HMM to transform the discrete observation values ​​into continuous states, thereby outputting the probabilities of lane keeping, lane departure and lane cutting of the vehicles in the adjacent lanes. That is, the probability density function of the observation state is represented by the Gaussian mixture model. The state transition matrix in the HMM is: Convert from discrete quantity to continuous quantity through probability density function; is an observable sequence, is an unobservable sequence, then The expected hidden state conditional probability distribution of a sample is:

[0119]

[0120] Add constraints to the above formula , each hidden state can be obtained separately The Gaussian mixture hidden Markov model GMM-HMM can be expressed as:

[0121]

[0122] Then the observation probability matrix Transformed into a probability density function of a set of observations, then given the state Observed values The probability density of is:

[0123]

[0124] The mathematical expression is:

[0125]

[0126]

[0127] Where, for The multi-dimensional Gaussian density function output in the state, is the mean vector, is the covariance matrix, is the data dimension, is the number of Gaussian mixtures, is the Gaussian mixing coefficient weight, and , for In the state The mean of a Gaussian random number generator.

[0128] Step 3: If there is a risk factor for the vehicle in the adjacent lane, that is, the sum of the lane departure probability and the lane cutting probability is greater than the lane keeping probability, a longitudinal safety distance model is established based on the relevant information of the vehicle and the vehicle in the adjacent lane. If there is no risk factor, steps 1 and 2 are repeated repeatedly. Specifically:

[0129] like Figures 2 to 4 As shown in Figure 3, based on the lateral intentions of the vehicles in the adjacent lanes predicted in steps 1 and 2, if the sum of the probabilities of lane departure and lane cutting is greater than the probability of lane keeping, that is, the lateral behavior of the vehicles in the adjacent lanes poses a certain risk to the ego vehicle, the ego vehicle will only move in a straight line within its lane under the action of the AEB system. A safety distance model is established, which is expressed as:

[0130]

[0131] Where, 、 are the distances traveled by the vehicle and the adjacent lane vehicles during the interactive game, 、 are the initial velocities of the vehicle and the adjacent lane vehicles, is the lane-changing time for interactive game, is the acceleration of the vehicle in the adjacent lane, The braking deceleration provided by the vehicle's AEB system;

[0132] The longitudinal distance between the vehicle and the adjacent lane vehicle can be expressed as:

[0133]

[0134] Where, is the longitudinal distance between the two vehicles at the initial moment of lane change;

[0135] Introducing the Berkeley longitudinal safety distance model

[0136]

[0137] Where, is the minimum braking distance, is the relative speed of the two vehicles, is the AEB system response time, is the braking system response time, is the maximum braking deceleration of the two vehicles;

[0138] The longitudinal distance and minimum braking distance between the vehicle and the adjacent lane vehicle should meet the following requirements:

[0139]

[0140] Where, To reserve a safe distance.

[0141] Step 4: Establish a non-cooperative Nash game framework, construct the strategy space of the ego vehicle and vehicles in adjacent lanes, and establish a payoff matrix based on the strategy combinations of both parties. Specifically:

[0142] In step 4, based on the non-cooperative Nash game, an interactive game model between the vehicle and the adjacent lane is established under different cut-in scenarios:

[0143]

[0144] Where, Indicates the number of participants. In this invention, the vehicle and the cutting vehicle are considered as game participants. For participants The decision set, For participants The profit function of

[0145] The braking deceleration provided by the ego vehicle's AEB system is selected as the strategy control variable. The interactive game elements between the ego vehicle and the vehicle in the adjacent lane in the lane cutting or lane departure scenario are as follows:

[0146] Participants:

[0147]

[0148] Strategy Space:

[0149] ,

[0150] Profit function:

[0151]

[0152] balanced:

[0153]

[0154] Where, The strategy space consists of all strategy combinations of participants, are the game strategy sets of the interactive game participants, Represent the possible actions of the participants, is the profit function of the set of actions taken by the vehicle for all other participating vehicles in the adjacent lanes, is the Nash equilibrium strategy solution of the interactive game, Represent their respective optimal strategies;

[0155] definition and are the strategy sets for the vehicle and the adjacent lane cutting vehicle, ={no braking, mild braking, emergency braking}, ={lane keeping, lane departure, lane cutting}, where the corresponding braking deceleration of the self-vehicle strategy is the no-braking strategy, the mild braking strategy, and the emergency braking strategy. , and ; Expressed in matrix form, establish the payoff matrix between the vehicle and the vehicles in the adjacent lanes:

[0156]

[0157] in: They represent the benefits of the vehicle and the vehicle cutting into the adjacent lane respectively. The three typical strategies of both parties in the game constitute nine decision combinations.

[0158] Step 5: Design a payoff function, assigning a weight to the payoff coefficient considering the different safety characteristics of each strategy. Substitute the solved value into the payoff matrix, and use the Nash equilibrium to find the optimal response strategy. The AEB system then performs graded braking based on the optimal strategy. Specifically:

[0159] In step 5, the safety benefit, comfort benefit and loss benefit of dangerous factors of vehicles in adjacent lanes are considered. and the benefit of vehicles cutting into the adjacent lane The current state of the benefits is calculated, including:

[0160] Security benefits The safety benefit takes into account the relative distance and relative speed between the two vehicles, which can be expressed as:

[0161]

[0162] Negative benefit of lateral hazard intention of vehicles in adjacent lanes Taking into account the negative impact of the lateral intention of the vehicle in the adjacent lane on the vehicle, the benefit is designed as follows based on the predicted lateral intention behavior probability:

[0163]

[0164] Where, is the lateral intention probability of vehicles in adjacent lanes predicted by the Gaussian mixture hidden Markov model, Calculate the coefficient for the heading angle of vehicles in adjacent lanes.

[0165] Comfort benefits Use the acceleration change during the game to represent the comfort gain

[0166]

[0167] The total benefits of the interactive game are formed by combining and weighting the ego vehicle and vehicles in adjacent lanes;

[0168]

[0169] Where, The weighted coefficients corresponding to safety benefit, negative benefit of dangerous intention and comfort benefit respectively;

[0170] With the help of the probability estimation and benefit function of the lateral intention of vehicles in adjacent lanes in the above steps, we can get

[0171]

[0172] The ego vehicle evaluates the relative benefits of each strategy in its own strategy set and selects the current optimal strategy. At this time, the AEB system executes the corresponding optimal strategy.

[0173] Step 6: After determining that there is no potential danger between the vehicle and the vehicles in the adjacent lane, the AEB system will exit.

[0174] When a vehicle in an adjacent lane has dangerous lateral behavior that triggers the AEB system, if it is detected that the vehicle's speed is lower than the speed of the oncoming vehicle, the oncoming vehicle's deceleration is lower than the vehicle's AEB mild braking deceleration, and the relative distance of the oncoming vehicle in the direction of the vehicle's lane is greater than the preset AEB brake release distance, the vehicle will be controlled to exit the AEB braking mode and the game interaction process will end.

[0175] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0176] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A game theory-based AEB graded braking method for scenarios involving other vehicles, characterized in that: The steps include: Step 1: Obtain the status information data of the ego vehicle and the surrounding environment, and calculate the heading angle of the vehicle in the adjacent lane and the distance from the vehicle in the adjacent lane to the lane line on the adjacent side of the ego vehicle; Step 2: Based on the obtained data, the lateral maneuver intention of vehicles in adjacent lanes is predicted using a Gaussian mixture hidden Markov model (GMM-HMM). The lateral maneuver characteristics are classified into three types: lane keeping, lane departure, and lane cutting. The lateral maneuver intention predicted by the GMM-HMM is expressed as a probability. Step 3: If there is a risk factor for the vehicle in the adjacent lane, that is, the sum of the lane departure probability and the lane cutting probability is greater than the lane keeping probability, a longitudinal safety distance model is established based on the relevant information of the vehicle and the vehicle in the adjacent lane. If there is no risk factor, steps 1 and 2 are repeated. Step 4: Establish a non-cooperative Nash game framework, construct the strategy space of the ego vehicle and vehicles in adjacent lanes, and establish a payoff matrix based on the strategy combinations of both parties. Step 5: Design a payoff function, assigning weights to the payoff coefficients based on the safety of each strategy. Substitute the resulting values ​​into the payoff matrix and determine the optimal response strategy based on the Nash equilibrium. The AEB system then performs graded braking based on the optimal strategy. Step 6: After determining that there is no potential danger between the vehicle and the vehicles in the adjacent lane, the AEB system exits operation; The step 5 is specifically as follows: In step 5, the safety benefit, comfort benefit and loss benefit of dangerous factors of vehicles in adjacent lanes are considered. and the benefit of vehicles cutting into the adjacent lane The current state of the benefits is calculated, including: Security benefits , the safety benefit takes into account the relative distance and relative speed between the two vehicles, which can be expressed as: ; Where, is the longitudinal distance between the vehicle and the vehicle in the adjacent lane, 、 are the initial velocities of the vehicle and the adjacent lane vehicles, is the interactive game duration, i.e., lane-changing time, The braking deceleration provided by the AEB system of the vehicle itself, is the acceleration of vehicles in adjacent lanes; Negative benefit of lateral hazard intention of vehicles in adjacent lanes , considering the negative impact of the lateral intention of the vehicle in the adjacent lane on the vehicle, combined with the predicted lateral intention behavior probability, this part of the benefit is designed as: ; Where, is the lateral intention probability of vehicles in adjacent lanes predicted by the Gaussian mixture hidden Markov model, Calculate the heading angle coefficient for vehicles in adjacent lanes, is the heading angle of the vehicle in the adjacent lane; Comfort benefits , using the acceleration change during the game to represent the comfort benefit ; The total benefits of the interactive game are formed by combining and weighting the ego vehicle and vehicles in adjacent lanes; ; Where, The weighted coefficients corresponding to safety benefit, negative benefit of dangerous intention and comfort benefit respectively; With the help of the probability estimation of the lateral intention of the vehicles in the adjacent lanes and the benefit function in the above steps, the benefits of the vehicle under different strategies can be obtained. , the benefits of vehicles in adjacent lanes under different strategies They are: ; Where, They correspond to the three different strategies that the vehicle may execute: {no braking, mild braking, emergency braking}. Indicates the three different strategic behaviors that vehicles in adjacent lanes may have: {lane keeping, lane departure, lane cutting}; The ego vehicle evaluates the relative benefits of each strategy in its own strategy set and selects the current optimal strategy. At this time, the AEB system executes the corresponding optimal strategy.

2. The AEB graded braking method in the scenario of other vehicles cutting in based on game theory according to claim 1 is characterized in that: The step 1 is specifically as follows: The vehicle status and surrounding environment information in step 1 are obtained through GPS and vehicle body sensors; The vehicle status information includes: vehicle position information, vehicle speed information and vehicle acceleration information; The surrounding environment information includes: adjacent lane vehicle position information, adjacent lane vehicle information, adjacent lane vehicle acceleration information, adjacent lane vehicle heading angle information and lane line position information; The heading angle information of the vehicle in the adjacent lane can be obtained from the speed information, and the expression is: ; Where, is the heading angle of the vehicle in the adjacent lane, is the lateral speed of vehicles in the adjacent lane, is the longitudinal speed of vehicles in the adjacent lane; Assuming the vehicle is in the middle of the lane, the distance from the vehicle in the adjacent lane to the lane line on the adjacent side of the vehicle can be expressed as: ; Where, is the distance from the adjacent lane vehicle to the lane line on the adjacent side of the vehicle, is the lane width, is the lateral position of the vehicle in the adjacent lane, is the lateral position of the vehicle, The width of vehicles in adjacent lanes.

3. The AEB graded braking method in the scenario of other vehicles cutting in based on game theory according to claim 2 is characterized in that: The step 2 is specifically as follows: In step 2, a Gaussian mixture hidden Markov model is used to predict the lateral intention of vehicles in the adjacent lane, and the internal latent state is identified using the external observation sequence. Based on the obtained data, the speed, acceleration, heading angle and distance from the lane line on the adjacent side of the vehicle are selected as the observation state, and the three lateral intentions of vehicles in the adjacent lane (LK), lane departure (LD) and lane cut (LC) are selected as the latent state. The random distribution function of the lateral intention of vehicles in the adjacent lane is established using HMM, which is expressed as a five-tuple. : ; The parts of HMM are defined as follows: Represents a set of hidden states ,in is the number of states, which represents the set of unobservable states of the lateral intentions of vehicles in adjacent lanes; Represents a set of observable sequences ,in is the number of different observations that can be output from each state; is the state transition matrix ,in Indicates that the system consists of the state Transfer to The probability of for The state of the moment, , ; is the observation probability matrix ,express The probability of the corresponding observation value output in the state, where , for Observe events at all times, , ; The initial hidden state is The probability of initializing the state distribution , , ; The driving state and driving behavior of the vehicle and the vehicles in the adjacent lanes are continuous time behaviors. The Gaussian mixture model GMM is combined with the hidden Markov model HMM to transform the discrete observation values ​​into continuous states, thereby outputting the probabilities of lane keeping, lane departure and lane cutting of the vehicles in the adjacent lanes. That is, the probability density function of the observation state is represented by the Gaussian mixture model. The state transition matrix in the HMM is: Convert from discrete quantity to continuous quantity through probability density function; is an observable sequence, is an unobservable sequence, then The expected hidden state conditional probability distribution of a sample is: ; Add constraints to the above formula , each hidden state can be obtained separately The Gaussian mixture hidden Markov model GMM-HMM can be expressed as: ; Then the observation probability matrix Transformed into a probability density function of a set of observations, then given the state Observed values The probability density of is: ; The mathematical expression is: ; ; Where, for The multi-dimensional Gaussian density function output in the state, is the mean vector, is the covariance matrix, is the data dimension, is the number of Gaussian mixtures, is the Gaussian mixing coefficient weight, and , for In the state The mean of a Gaussian random number generator.

4. The AEB graded braking method in the scenario of other vehicles cutting in based on game theory according to claim 3 is characterized in that: The step 3 is specifically as follows: In step 3, based on the lateral intentions of the vehicles in the adjacent lanes predicted in steps 1 and 2, if the sum of the probabilities of lane departure and lane cutting is greater than the probability of lane keeping, that is, the lateral behavior of the vehicles in the adjacent lanes poses a certain risk to the ego vehicle, the ego vehicle will only move in a straight line within its lane under the action of the AEB system. A safety distance model is established, which is expressed as: ; Where, 、 are the distances traveled by the vehicle and the adjacent lane vehicles during the interactive game, 、 are the initial velocities of the vehicle and the adjacent lane vehicles, is the lane-changing time for interactive game, is the acceleration of the vehicle in the adjacent lane, The braking deceleration provided by the vehicle's AEB system; The longitudinal distance between the vehicle and the adjacent lane vehicle can be expressed as: ; Where, is the longitudinal distance between the two vehicles at the initial moment of lane change; Introducing the Berkeley longitudinal safety distance model ; Where, is the minimum braking distance, is the relative speed of the two vehicles, is the AEB system response time, is the braking system response time, is the maximum braking deceleration of the two vehicles; The longitudinal distance and minimum braking distance between the vehicle and the adjacent lane vehicle should meet the following requirements: ; Where, To reserve a safe distance.

5. The AEB graded braking method in the scenario of other vehicles cutting in based on game theory according to claim 4 is characterized in that: The step 4 is specifically as follows: In step 4, based on the non-cooperative Nash game, an interactive game model between the vehicle and the adjacent lane is established under different cut-in scenarios: ; Where, Represents the set of participants, and the vehicle and the cutting vehicle are regarded as the set of game participants. For participants The decision set, For participants The profit function of The braking deceleration provided by the ego vehicle's AEB system is selected as the strategy control variable. The interactive game elements between the ego vehicle and the vehicle in the adjacent lane in the lane cutting or lane departure scenario are as follows: Participants: Strategy Space: ， ; Profit function: ; balanced: ; Where, The strategy space consists of all strategy combinations of participants, are the game strategy sets of the interactive game participants, Represent the possible actions of the participants, is the profit function of the set of actions taken by the vehicle for all other participating vehicles in the adjacent lanes, is the Nash equilibrium strategy solution of the interactive game, Represent their respective optimal strategies; definition and are the strategy sets for the vehicle and the adjacent lane cutting vehicle, ={no braking, mild braking, emergency braking}, ={lane keeping, lane departure, lane cutting}, and the benefits of the vehicle under the three strategies are expressed as , the benefits of vehicles in adjacent lanes under the three strategies are expressed as The three typical strategies of both parties in the game form nine decision combinations, and the payoff matrix of the ego vehicle and the vehicles in the adjacent lane is established: the payoff matrix of the ego vehicle when the ego vehicle does not brake and the vehicles in the adjacent lane keep the lane is The payoff matrix when the ego vehicle brakes gently and the adjacent lane vehicles keep lane is: ; The payoff matrix of the emergency braking strategy of the ego vehicle and the lane keeping of the adjacent lane vehicle is: ; The payoff matrix when the ego vehicle does not brake and the adjacent lane vehicle deviates is ; The payoff matrix when the ego vehicle brakes gently and the adjacent lane vehicle deviates from the lane is: ; The payoff matrix when the ego vehicle brakes suddenly and the adjacent lane vehicle deviates is: ; The payoff matrix when the vehicle in the adjacent lane cuts in without braking is ; The payoff matrix when the vehicle in the adjacent lane cuts into the lane during gentle braking of the ego vehicle is: ; The payoff matrix when the vehicle in the adjacent lane cuts in during emergency braking is: .

6. The AEB graded braking method in the scenario of other vehicles cutting in based on game theory according to claim 5 is characterized in that: The step 6 is specifically as follows: When a vehicle in an adjacent lane has dangerous lateral behavior that triggers the AEB system, if it is detected that the vehicle's speed is lower than the speed of the oncoming vehicle, the oncoming vehicle's deceleration is lower than the vehicle's AEB mild braking deceleration, and the relative distance of the oncoming vehicle in the direction of the vehicle's lane is greater than the preset AEB brake release distance, the vehicle will be controlled to exit the AEB braking mode and the game interaction process will end.

Images