Dynamic game-based intelligent vehicle lane change decision-making method in foggy days

Abstract

The invention provides a dynamic game-based intelligent vehicle lane change decision-making method in foggy days, a lane change intention model comprises a dissatisfaction model and a safety distance model, and a lane change income model comprises a safety income model, a speed income model and a comfort income model; dissatisfaction information of the vehicle is obtained through the dissatisfaction model, lane changing safety distance information of the vehicle is obtained through the safety distance model, and lane changing intention information is obtained according to the dissatisfaction information and the lane changing safety distance information; after the lane changing intention is generated, speed income information and safety income information of the vehicle and comfort income information of the vehicle and vehicles behind the target lane are obtained respectively; analyzing and acquiring total income information under each strategy combination; and obtaining an optimal strategy combination according to the total income information, and taking the optimal strategy combination as a lane changing decision. According to the method, the optimal lane changing mode is evaluated through the game revenue function, and the problems of safety, real-time performance and efficiency of lane changing decision making under the low-visibility condition are solved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and particularly to an intelligent vehicle lane-changing decision-making method based on dynamic game in foggy weather. Background Art

[0002] In autonomous driving technology, lane-changing decision-making is one of the core issues in complex scenarios. In the prior art, traditional lane-changing strategies are mostly based on fixed rules or single-objective optimization (such as the shortest path, the lowest energy consumption). However, such methods have significant defects in low-visibility environments such as foggy weather: ① Environmental perception defect: Low visibility in foggy weather leads to a decrease in sensor accuracy, making it difficult for vehicles to accurately obtain the dynamic information of surrounding vehicles (such as speed, acceleration), and static rules are difficult to adapt to dynamic interaction scenarios; ② Multi-agent game conflict: Existing methods usually regard other vehicles as passive obstacles and ignore their active decision-making behaviors (such as avoidance or acceleration), resulting in an overly simplified decision-making model and prone to potential collision risks; ③ Conflict between safety and efficiency: Single-objective optimization (such as maximizing speed) may sacrifice the safety distance, while overly conservative decisions (such as forcibly maintaining the original lane) reduce traffic efficiency.

[0003] This application aims to solve the complexity and safety problems of lane-changing decision-making in multi-vehicle interaction scenarios in foggy weather. By constructing a multi-step dynamic game framework and a lane-changing intention probability model, it realizes safe and efficient lane-changing decision-making for autonomous vehicles under low-visibility conditions. Summary of the Invention

[0004] In order to solve the problems in the background art, the present invention proposes an intelligent vehicle lane-changing decision-making method based on dynamic game in foggy weather.

[0005] An intelligent vehicle lane-changing decision-making method based on dynamic game in foggy weather includes the steps of:

[0006] S100. Construct a lane-changing intention model and a lane-changing benefit model. The lane-changing intention model includes a dissatisfaction model and a safety distance model, and the lane-changing benefit model includes a safety benefit model, a speed benefit model, and a comfort benefit model;

[0007] S200. Obtain the dissatisfaction information of the vehicle through the dissatisfaction model, obtain the lane-changing safety distance information of the vehicle through the safety distance model, and obtain the lane-changing intention information according to the dissatisfaction information and the lane-changing safety distance information;

[0008] S300. If a lane-changing intention is generated, obtain the safety benefit information of the vehicle and the vehicle behind in the target lane through the safety benefit model, obtain the speed benefit information of the vehicle through the speed benefit model, and obtain the comfort benefit information of the vehicle and the vehicle behind in the target lane through the comfort benefit model;

[0009] S400. Introduce a multi-step time discount factor and a fog visibility parameter, and perform non-linear normalization analysis based on the obtained speed benefit information, safety benefit information, and comfort benefit information to obtain the total benefit information under each strategy combination.

[0010] S500. Obtain the optimal strategy combination based on the total benefit information, and use the optimal strategy combination as the lane-changing decision.

[0011] Based on the above, the dissatisfaction model M(t) is:

[0012] M(t) = M(t - 1) + ΔM(t)

[0013]

[0014] v d = min{v Tfront , v max × k}

[0015] where M(t) is the dissatisfaction at time t, ΔM(t) is the change in dissatisfaction at time t, ω v is the dissatisfaction weight of speed, is the dissatisfaction weight of the distance between the host vehicle and the vehicle in front on the target lane, v d is the expected vehicle speed, v Tfront is the speed of the vehicle in front on the target lane, v max is the maximum speed limit of the current road section, k is the speed limit coefficient; v c is the current vehicle speed of the host vehicle, d c is the current distance from the host vehicle to the vehicle in front on the target lane; d s is the safe distance between the vehicle and the vehicles in front and behind during normal driving.

[0016] Based on the above, the safety distance model d s is:

[0017] d s = d 0 ρ + t tap v c

[0018] ρ = μl

[0019] where d 0 is the minimum safe distance under clear weather, t tap is the following time interval, ρ is the minimum safe distance parameter under a certain weather condition, μ is the road surface sliding friction coefficient, and l is the visibility.

[0020] Based on the above, in step S200, a preset dissatisfaction threshold is set. If the current dissatisfaction is greater than the dissatisfaction threshold, and the distance from the vehicle in front and behind on the target lane is greater than the safety distance d sWhen this occurs, a lane change intention is generated; otherwise, no lane change intention is generated.

[0021] Based on the above, the safety benefit model includes:

[0022] (1) Lane change of the ego vehicle:

[0023] Under foggy conditions, the current position x of the ego vehicle ego and the current position x of the following vehicle Tback in the target lane Tback follow a normal distribution, and the standard deviation of the positions of the two vehicles is inversely proportional to the visibility:

[0024]

[0025] where is the measured value of the position of the ego vehicle under ideal conditions; is the measured value of the position of the following vehicle Tback in the target lane under ideal conditions; σ ego is the uncertainty of the position of the ego vehicle caused by the ranging error of the sensor under the influence of foggy visibility, σ Tback is the uncertainty of the position of the following vehicle Tback in the target lane caused by the ranging error of the sensor under the influence of foggy visibility, and k is an empirical coefficient;

[0026] The collision probability P between the ego vehicle and the following vehicle Tback in the target lane collision is:

[0027]

[0028]

[0029] where d safe is the safe following distance under the current visibility condition; Φ(·) is the cumulative distribution function of the standard normal distribution; τ is the lane change time; is the expected position after the ego vehicle completes the lane change; is the expected position after the following vehicle Tback in the target lane completes the lane change; is the component of the acceleration of the ego vehicle in the direction perpendicular to the lane; is the velocity component of the ego vehicle in the direction perpendicular to the lane before the lane change; d x is the distance between the current lane and the target lane;

[0030] The safety benefit between the ego vehicle and the following vehicle Tback in the target lane is:

[0031]

[0032] where is the safety benefit of the ego vehicle, is the safety benefit of the vehicle Tback following in the target lane;

[0033] (2) The ego vehicle does not change lanes: The safety benefits of the ego vehicle, the vehicle Cfront in the current lane, the vehicle Tfront in the target lane, and the vehicle Tback in the target lane are all 1.

[0034] Based on the above, the speed benefit model includes:

[0035] (1) The speed benefit of the ego vehicle:

[0036] The speed benefit of the ego vehicle at time step t is:

[0037]

[0038] The total speed benefit of the ego vehicle after n time steps is:

[0039]

[0040] where V Cfront is the speed of the vehicle Cfront in the current lane, V Tfront is the speed of the vehicle Tfront in the target lane, V ego is the speed of the ego vehicle, and γ is the time step discount factor;

[0041] (2) The speed benefit of the vehicle Tback in the target lane:

[0042] The speed benefit of the vehicle Tback in the target lane at time step t is:

[0043]

[0044] The total speed benefit of the vehicle Tback in the target lane after n time steps is:

[0045]

[0046] where, is the speed of the vehicle Tback reaching the minimum safety distance point when Tback chooses not to avoid, is the speed of the vehicle Tback reaching the minimum safety distance point when Tback chooses to avoid, V Tback is the speed of the vehicle Tback in the target lane, T Tback,n-avoid is the time for the vehicle Tback in the target lane to travel from t 0 to the minimum safety distance point under the non-avoidance strategy; T Tback,avoidThe time taken for the following vehicle Tback in the target lane to travel to the minimum safety distance point under the avoidance strategy selection; a 0 The acceleration of the following vehicle TB in the target lane at time t0; a Tback The maximum braking acceleration of the following vehicle Tback in the target lane. max Based on the above, the comfort benefit model includes:

[0047] (1) The comfort benefit of the ego vehicle:

[0048] ① When the ego vehicle changes lanes, the comfort benefit of the ego vehicle at time step t

[0049] is: The total comfort benefit of the ego vehicle after n time steps

[0050]

[0051] is: The total comfort benefit of the ego vehicle after n time steps

[0052]

[0053] where a ego is the maximum braking acceleration during the lane change of the ego vehicle, a max is the maximum braking acceleration of the vehicle under the current visibility conditions; γ is the time step discount factor;

[0054] ② When the ego vehicle does not change lanes, the total comfort benefit of the ego vehicle is 0;

[0055] (2) The comfort benefit of the following vehicle Tback in the target lane:

[0056] ① When the following vehicle Tback in the target lane avoids, the comfort benefit of the following vehicle Tback at time step t is:

[0057]

[0058] The total comfort benefit of the following vehicle Tback in the target lane after n time steps is:

[0059]

[0060] where a Tback is the avoidance braking acceleration of the following vehicle Tback in the target lane, a max is the maximum braking acceleration of the following vehicle Tback in the target lane; γ is the time step discount factor;

[0061] ②If the following vehicle Tback in the target lane does not give way, since a Tback is 0, then the benefit gain of the following vehicle Tback in the target lane at time step t is 0, and the total comfort gain is 0.

[0062] Based on the above, in step S400, the strategy combinations include:

[0063] S (C,A) = {ego lane change, Tback gives way}

[0064] S (C,NA) = {ego lane change, Tback does not give way}

[0065] S (NC,A) = {ego does not change lane, Tback gives way}

[0066] S (NC,NA) = {ego does not change lane, Tback does not give way}.

[0067] Based on the above, the total benefit function R total is:

[0068]

[0069] Among them, R safety is the safety benefit, R efficiency is the speed benefit, R comfort is the comfort benefit; α is the safety benefit influence factor; β is the speed benefit influence factor; γ is the comfort benefit influence factor;

[0070] The total benefit of each strategy combination is:

[0071] ①The total benefit of the strategy combination S (C,A) is:

[0072]

[0073] Among them, is the total benefit of the ego vehicle, is the safety benefit of the ego vehicle, is the speed benefit of the ego vehicle, is the comfort benefit of the ego vehicle; is the total benefit of the following vehicle Tback in the target lane, is the safety benefit of the following vehicle Tback in the target lane, is the speed benefit of the following vehicle Tback in the target lane, is the comfort benefit of the following vehicle Tback in the target lane; α is the safety benefit impact factor; β is the speed benefit impact factor; γ is the comfort benefit impact factor;

[0074] ② The total benefit of strategy combination S (C,NA) is:

[0075]

[0076] ③ The total benefit of strategy combination S (NC,A) is:

[0077]

[0078] ④ The total benefit of strategy combination S (NC,NA) is:

[0079]

[0080] Based on the above, the strategy combination with the maximum total benefit is taken as the optimal strategy combination.

[0081] The present invention has prominent substantive features and significant progress compared with the prior art. Specifically, the present invention constructs a lane-changing intention model through a probability model, simultaneously considers safety conditions and the cumulative dissatisfaction over a long period to trigger the lane-changing intention, constructs a lane-changing benefit model through a multi-step dynamic game framework, avoids short-sighted decisions by multi-step iterative simulation of future states, and uses a game benefit function to evaluate the best lane-changing method, solving the problems of safety, real-time performance, and efficiency of lane-changing decisions in a multi-vehicle interaction scenario under low visibility conditions, and realizing safe and efficient lane-changing decisions for autonomous vehicles under low visibility conditions. Description of the Drawings

[0082] Figure 1 is a flowchart showing the process of the present invention. Detailed Embodiments

[0083] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0084] As Figure 1As shown in the figure, a foggy-day intelligent vehicle lane-changing decision-making method based on dynamic game includes the following steps: S100. Construct a lane-changing intention model and a lane-changing benefit model. The lane-changing intention model includes a dissatisfaction model and a safety distance model, and the lane-changing benefit model includes a safety benefit model, a speed benefit model, and a comfort benefit model; S200. Obtain the dissatisfaction information of the vehicle through the dissatisfaction model, obtain the lane-changing safety distance information of the vehicle through the safety distance model, and obtain the lane-changing intention information according to the dissatisfaction information and the lane-changing safety distance information; S300. If a lane-changing intention is generated, obtain the safety benefit information of the vehicle and the vehicle behind in the target lane through the safety benefit model, obtain the speed benefit information of the vehicle through the speed benefit model, and obtain the comfort benefit information of the vehicle and the vehicle behind in the target lane through the comfort benefit model; S400. Introduce a multi-step time discount factor and a foggy-day visibility parameter, and perform non-linear normalization analysis according to the obtained speed benefit information, safety benefit information, and comfort benefit information to obtain the total benefit information under each strategy combination; S500. Obtain the optimal strategy combination according to the total benefit information, and use the optimal strategy combination as the lane-changing decision.

[0085] With the development of vehicle intelligence, the intelligent control unit of intelligent vehicles increasingly undertakes the responsibilities of drivers. Therefore, intelligent vehicles cannot ignore the understanding of the vehicle's highest decision-maker, that is, the driver. In this embodiment, the lane-changing intention of the driver is simulated by constructing a lane-changing intention model. In this model, we will consider two parts, namely the traffic environment and the vehicle state. The traffic environment refers to the traffic flow condition and the weather condition, which are the main influencing factors for drivers to change lanes. For example, when there is a vehicle driving at a very slow speed in front and the driving speed cannot reach the expected value, the driver will have a very high dissatisfaction and is likely to generate a lane-changing intention. The vehicle state refers to the vehicle speed, the relative position with other vehicles, and other data related to vehicle driving. When these data reach the threshold, dissatisfaction will be generated. In this embodiment, the dissatisfaction model M(t) is:

[0086] M(t) = M(t - 1) + ΔM(t)

[0087]

[0088] v d = min{v Tfront , v max × k}

[0089] where M(t) is the dissatisfaction at time t, ΔM(t) is the change in dissatisfaction at time t, ω v is the dissatisfaction weight of the vehicle speed of this vehicle, ω d is the dissatisfaction weight of the longitudinal distance between this vehicle and the vehicle in front in the target lane, v d is the expected vehicle speed, v Tfrontis the speed of the vehicle in front in the target lane, v max is the maximum speed limit of the current section. For example, under the conditions of a highway section, according to relevant regulations, when a motor vehicle is driving on a highway and encounters low visibility meteorological conditions such as fog, rain, snow, sand and dust, hail, etc., when the visibility is less than 200 meters, the vehicle speed shall not exceed 60 kilometers per hour, and keep a distance of more than 100 meters from the vehicle in front in the same lane, that is, v max = 60 km / h; when the visibility is less than 100 meters, the vehicle speed shall not exceed 40 kilometers per hour, and keep a distance of more than 50 meters from the vehicle in front in the same lane, that is, v max = 40 km / h; when the visibility is less than 50 meters, the vehicle speed shall not exceed 20 kilometers per hour, and should leave the highway from the nearest exit as soon as possible, that is, v max = 20 km / h, k is the speed limit coefficient, 0 ≤ k ≤ 1, which is positively correlated with the visibility l. In this embodiment, when the visibility is less than 200 meters, k ≈ 0.8; when the visibility is less than 100 meters, k ≈ 0.75; when the visibility is less than 50 meters, k ≈ 0.7. When there is no vehicle in front in the target lane, the expected vehicle speed v d then takes the reasonable speed v of the current section max × k. v c is the current vehicle speed of this vehicle, d c is the current distance from the vehicle in front in the target lane; d s is the safe distance between the vehicle and the vehicles in front and behind during normal driving of the vehicle.

[0090] The safety distance model d s is:

[0091] d s = d 0 ρ + t tap v c

[0092] ρ = μl

[0093] Among them, d 0 is the minimum safe distance under clear weather, t tap is the following time interval, ρ is the minimum safe distance parameter under a certain weather condition, μ is the road surface sliding friction coefficient, and l is the visibility. In reality, corresponding minimum safe distance tables are respectively configured for different road types such as cement roads, asphalt roads, dirt roads, etc. By obtaining the road image through the in-vehicle camera and through image recognition, the corresponding d 0 can be obtained. Similarly, after obtaining the road type and the dry-wet condition of the road surface, etc. through image recognition, the corresponding road surface sliding friction coefficient μ can be obtained by looking up the table, and the visibility l is recognized and obtained through the in-vehicle sensor. Then, the safety distance information can be obtained through calculation.

[0094] In practice, in step S200, a preset dissatisfaction threshold is set. After obtaining the actual dissatisfaction through the dissatisfaction model, it is compared with the dissatisfaction threshold. If the current dissatisfaction is greater than the dissatisfaction threshold and the distances from the vehicles in front and behind in the target lane are both greater than the safety distance d s then a lane-changing intention is generated; otherwise, no lane-changing intention is generated, that is, continue to follow the vehicle in this lane.

[0095] In this embodiment, the profit functions of speed profit and safety profit are used as the evaluation indicators of the strategy. Through the profit analysis from two perspectives, the total profit analysis is carried out, and the Nash equilibrium is carried out according to the profit matrix to obtain the optimal strategy combination of both sides of the game, and the optimal strategy combination is used as the final lane-changing decision.

[0096] The safety profit model considering the influence of foggy visibility uncertainty includes:

[0097] (1) The ego vehicle changes lanes:

[0098] Under foggy conditions, due to the limitation of vehicle vision by visibility, the perceived positions and speeds of the ego vehicle and the vehicle Tback in the target lane have errors. The current position x ego of the ego vehicle and the current position x Tback of the vehicle Tback in the target lane follow a normal distribution, and the standard deviation of the positions of the two vehicles is inversely proportional to the visibility:

[0099]

[0100]

[0101] where is the measured value of the position of the ego vehicle under ideal conditions; is the measured value of the position of the vehicle Tback in the target lane under ideal conditions; σ ego is the uncertainty of the position of the ego vehicle caused by the ranging error of the sensor under the influence of foggy visibility, and σ Tback is the uncertainty of the position of the vehicle Tback in the target lane caused by the ranging error of the sensor under the influence of foggy visibility. k is an empirical coefficient, and in this embodiment, k = 0.3.

[0102] The collision probability P collision between the ego vehicle and the vehicle Tback in the target lane is:

[0103]

[0104] where d safe is the safe vehicle distance under the current visibility condition; Φ(·) is the cumulative distribution function of the standard normal distribution; τ is the lane-changing time; is the expected position after the ego lane change of the vehicle; is the expected position after the lane change of the following vehicle Tback in the target lane; is the component of the ego vehicle's acceleration in the direction perpendicular to the lane; is the velocity component of the ego vehicle in the direction perpendicular to the lane before the lane change; d x is the distance between the current lane and the target lane, that is, the distance between the vehicle and the leading vehicle in the target lane in the direction perpendicular to the lane, and this distance is obtained by detecting with the intelligent vehicle sensor.

[0105] The safety benefit between the ego vehicle and the following vehicle Tback in the target lane is:

[0106]

[0107] Among them, is the safety benefit of the ego vehicle, is the safety benefit of the following vehicle Tback in the target lane; P collision is the collision probability between the ego vehicle and the following vehicle Tback in the target lane.

[0108] (2) The ego vehicle does not change lanes: The safety benefits of the ego vehicle, the leading vehicle Cfront in the current lane, the leading vehicle Tfront in the target lane, and the following vehicle Tback in the target lane are all 1.

[0109] The speed benefit model introducing the time-step discount factor γ includes:

[0110] (1) The speed benefit of the ego vehicle:

[0111] The speed benefit of the ego vehicle at time step t is:

[0112]

[0113] The total speed benefit of the ego vehicle after n time steps is:

[0114]

[0115] Among them, V Cfront is the speed of the leading vehicle Cfront in the current lane, V Tfront is the speed of the leading vehicle Tfront in the target lane, V egois the speed of the vehicle ego. γ is the time step discount factor. The larger the value of γ, the more emphasis is placed on long-term benefits. The smaller the value, the more emphasis is placed on short-term benefits and immediate benefits, and decisions can be made faster. For short-term lane change decisions on highways or smooth roads (1 to 5 seconds), it is 0.8 to 0.95 in this embodiment. For short-term lane change decisions on congested or complex road conditions (1 to 5 seconds), it is 0.7 to 0.85 in this embodiment. Time step n is the total number of time steps when the benefit function reaches the critical threshold for lane change.

[0116] (2) Speed ​​gain of the vehicle behind the target lane Tback:

[0117] The speed gain of the vehicle behind the target lane Tback at time step t for:

[0118]

[0119]

[0120] The total speed gain of the vehicle behind the target lane after n time steps for:

[0121]

[0122] in, The speed at which the target lane rear vehicle Tback reaches the minimum safe distance point when Tback chooses not to yield. V is the speed at which the target lane rear vehicle Tback reaches the minimum safe distance point when Tback chooses to avoid. Tback is the speed of the vehicle behind the target lane, T Tback,n-avoid For the target lane vehicle Tback, choose the non-avoidance strategy from t 0 (t=0) The time from (t=0) to the minimum safe distance point; T Tback,avoid Select an avoidance strategy for the target lane vehicle Tback from t 0 The time to reach the minimum safe distance point; a Tback The target lane rear vehicle Tback is at t 0 The acceleration at the moment, a max is the maximum braking acceleration of the vehicle behind the target lane Tback (also the maximum braking acceleration of the vehicle under the current visibility condition). In this embodiment, a rapid acceleration is used, that is, if the acceleration of the vehicle exceeds 0.5g (about 5m / s 2 ) is regarded as a rapid acceleration.

[0123] The comfort benefit model that introduces the time step discount factor γ includes:

[0124] (1) The comfort benefit of the vehicle ego:

[0125] ① When the ego vehicle of this vehicle changes lanes, due to the vehicle speed being too fast or too slow, which affects the comfort of the driver and passengers, the comfort benefit of this vehicle depends on the acceleration deviation value. Therefore, the comfort benefit of the ego vehicle of this vehicle at time step t is:

[0126]

[0127] The total comfort benefit of the ego vehicle of this vehicle after n time steps is:

[0128]

[0129] where a ego is the maximum braking acceleration during the lane change process of the ego vehicle of this vehicle, and a max is the maximum braking acceleration of the vehicle under the current visibility conditions; γ is the time step discount factor;

[0130] ② When the ego vehicle of this vehicle does not change lanes, since a geo is 0, then the comfort benefit of the ego vehicle of this vehicle at time step t is 0, and the total comfort benefit of the ego vehicle of this vehicle is 0.

[0131] (2) Comfort benefit of the vehicle Tback in the target lane:

[0132] ① If the vehicle Tback in the target lane takes an avoidance strategy, the comfort benefit of the vehicle Tback in the target lane at time step t is:

[0133]

[0134] The total comfort benefit of the vehicle Tback in the target lane after n time steps is:

[0135]

[0136] where a Tback is the avoidance braking acceleration of the vehicle Tback in the target lane, and a max is the maximum braking acceleration of the vehicle Tback in the target lane; γ is the time step discount factor;

[0137] ② If the vehicle Tback in the target lane takes a non-avoidance strategy, since a Tback is 0, then the benefit of the vehicle Tback in the target lane at time step t is 0, and the total comfort benefit is 0.

[0138] According to the lane-changing situation of the vehicle ego and the avoidance situation of the vehicle Tback in the target lane behind, four strategy combinations are obtained, and the four strategy combinations include:

[0139] S (C,A) = {ego changes lanes, Tback avoids}

[0140] S (C,NA) = {ego changes lanes, Tback does not avoid}

[0141] S (NC,A) = {ego does not change lanes, Tback avoids}

[0142] S (NC,NA) = {ego does not change lanes, Tback does not avoid}.

[0143] Total revenue function R total Considering that safety accidents are extremely likely to occur under foggy conditions, safety factors should be placed in the absolute maximum position. At the same time, to avoid the comfort revenue under the traditional weighted average calculation method from overly influencing the decision-making, a non-linear normalization calculation method is adopted. The total revenue function R tota is:

[0144]

[0145] where R safety is the safety revenue, R efficiency is the speed revenue, R comfort is the comfort revenue; α is the safety revenue influence factor (α, β, and γ are all exponents in the formula). In this embodiment, α≈2. When the safety is low (for example, R safety ≤0.2), the safety revenue drops sharply, and the influence of the safety revenue on the decision-making is more obvious than that of the speed revenue and the comfort revenue; β is the speed revenue influence factor, usually 0≤β≤1.5. In this embodiment, β≈0.9 to avoid the speed-optimal strategy from dominating the lane-changing decision; γ is the comfort revenue influence factor, usually 0≤γ≤1. In this embodiment, γ is set to be approximately 0.5 to ensure that the comfort will not overly influence the decision-making.

[0146] The total revenue of each strategy combination is:

[0147] ① The total revenue of the strategy combination S (C,A) is:

[0148]

[0149] where is the total revenue of the vehicle ego, is the safety revenue of the vehicle ego, is the speed revenue of the vehicle ego, The comfort benefit of the ego vehicle of this vehicle is the total benefit of the following vehicle Tback in the target lane, is the safety benefit of the following vehicle Tback in the target lane, is the speed benefit of the following vehicle Tback in the target lane, is the comfort benefit of the following vehicle Tback in the target lane; α is the safety benefit influence factor; β is the speed benefit influence factor; γ is the comfort benefit influence factor;

[0150] ② The strategy combination S (C,NA) The total benefit of is:

[0151]

[0152] ③ The strategy combination S (NC,A) The total benefit of is:

[0153]

[0154] ④ The strategy combination S (NC,NA) The total benefit of is:

[0155]

[0156] After obtaining the total benefits of the four groups of strategy combinations, a comparative analysis is carried out, and the strategy combination with the largest total benefit is used as the optimal strategy combination, and the content of this strategy combination is used as the final lane-changing decision.

[0157] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. A dynamic game-based intelligent vehicle lane-changing decision method in foggy weather, characterized in that: Includes steps: S100, constructing a lane-changing intention model and a lane-changing benefit model, wherein the lane-changing intention model includes a dissatisfaction model and a safety distance model, and the lane-changing benefit model includes a safety benefit model, a speed benefit model, and a comfort benefit model; S200, obtaining dissatisfaction information of the vehicle through a dissatisfaction model, obtaining lane-changing safety distance information of the vehicle through a safety distance model, and obtaining lane-changing intention information according to the dissatisfaction information and the lane-changing safety distance information; S300: If a lane change intention is generated, safety benefit information of the vehicle and the vehicle behind the target lane is obtained through a safety benefit model, speed benefit information of the vehicle is obtained through a speed benefit model, and comfort benefit information of the vehicle and the vehicle behind the target lane is obtained through a comfort benefit model; S400, introducing a multi-step time discount factor and a foggy visibility parameter, and performing a nonlinear normalization analysis based on the acquired speed benefit information, safety benefit information, and comfort benefit information to obtain the total benefit information under each strategy combination; S500: Obtain an optimal strategy combination according to the total benefit information, and use the optimal strategy combination as a lane-changing decision.

2. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 1 is characterized by: The dissatisfaction model M(t) is: M(t)=M(t-1)+ΔM(t) v d =min{v Tfront ,v max ×k} Where M(t) is the dissatisfaction at time t, ΔM(t) is the change in dissatisfaction at time t, ω v is the speed dissatisfaction weight, ω d is the dissatisfaction weight of the longitudinal distance between the vehicle and the vehicle in front of the target lane, v d is the expected vehicle speed, v Tfront is the speed of the vehicle ahead in the target lane, v max is the maximum speed limit of the current road section, k is the speed limit coefficient; v c is the current speed of the vehicle, d c is the current distance to the vehicle in front of the target lane; d s It is the safe distance between the vehicle and the vehicles in front and behind when the vehicle is driving normally.

3. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 1 is characterized by: Safety distance model d s for: d s =d0ρ+t tap v c ρ=μl Among them, d0 is the minimum safe distance between vehicles under clear sky, t tap is the following vehicle time interval, ρ is the minimum safe vehicle distance parameter under certain weather conditions, μ is the road sliding friction coefficient, and l is visibility.

4. The method for intelligent vehicle lane changing decision-making in foggy weather based on dynamic game according to claim 1 is characterized by: In step S200, a dissatisfaction threshold is preset. If the current dissatisfaction is greater than the dissatisfaction threshold and the distance to the front and rear vehicles in the target lane is greater than the safety distance d s When , the lane change intention is generated; Otherwise, no lane change intention is generated.

5. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 1 is characterized in that: The security benefit model includes: (1) The vehicle's ego lane change: In foggy weather, the current position x of the vehicle ego ego and the current position x of the vehicle behind the target lane Tback Tback It follows a normal distribution, and the standard deviation of the two car positions is inversely proportional to visibility: in, is the measured value of the vehicle ego position under ideal conditions; is the measured value of the position of the vehicle behind the target lane Tback under ideal conditions; σ ego is the uncertainty of the vehicle ego position caused by the sensor ranging error under the influence of foggy visibility, σ Tback is the uncertainty of the position of the vehicle behind Tback in the target lane caused by the sensor ranging error under the influence of foggy visibility, and k is the empirical coefficient; The collision probability P of the vehicle ego and the vehicle behind in the target lane Tback collision for: Among them, d safe is the safe vehicle distance under the current visibility conditions; Φ(·) is the cumulative distribution function of the standard normal distribution; τ is the lane change time; is the desired position of the vehicle ego after the lane change is completed; is the expected position of the vehicle Tback in the target lane after the lane change is completed; is the component of the acceleration of the vehicle ego in the direction perpendicular to the lane; is the speed component of the vehicle ego in the vertical lane direction before changing lanes; d x is the distance between the current lane and the target lane; The safety benefit of the vehicle ego and the vehicle behind in the target lane Tback is: in, For the safety benefit of the vehicle ego, is the safety benefit of the vehicle Tback in the target lane; (2) The vehicle ego does not change lanes: The safety benefits of the vehicle ego, the vehicle in front of the current lane Cfront, the vehicle in front of the target lane Tfront, and the vehicle in the back of the target lane Tback are all 1.

6. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 1 is characterized in that: The velocity benefit model includes: (1) The speed gain of the vehicle ego: The speed gain of the vehicle ego at time step t for: The total speed gain of the vehicle ego after n time steps for: Among them, V Cfront is the speed of the front vehicle Cfront in the current lane, V Tfront is the speed of the front vehicle Tfront in the target lane, V ego is the speed of the vehicle ego, γ is the time step discount factor; (2) Speed ​​gain of the vehicle behind the target lane Tback: The speed gain of the vehicle behind the target lane Tback at time step t for: The total speed gain of the vehicle behind the target lane after n time steps for: in, The speed at which the target lane rear vehicle Tback reaches the minimum safe distance point when Tback chooses not to yield. V is the speed at which the target lane rear vehicle Tback reaches the minimum safe distance point when Tback chooses to avoid. Tback is the speed of the vehicle behind the target lane, T Tback,n-avoid T is the time taken by the target lane vehicle Tback to travel from t0 to the minimum safe distance point when the vehicle behind chooses no avoidance strategy; Tback,avoid The time from t0 to the minimum safe distance point when the target lane rear vehicle Tback selects the avoidance strategy; a Tback is the acceleration of the vehicle TB behind the target lane at time t0; a max is the maximum braking acceleration of the vehicle behind Tback in the target lane.

7. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 1 is characterized in that: The comfort benefit model includes: (1) The comfort benefit of the vehicle ego: ① The comfort benefit of the vehicle ego at time step t when the vehicle ego changes lanes for: The total comfort benefit of the vehicle ego after n time steps for: Among them, a geo is the maximum braking acceleration of the vehicle ego during lane change, a max is the maximum braking acceleration of the vehicle under the current visibility conditions; γ is the time step discount factor; ② The total comfort benefit of the vehicle ego is that the vehicle ego does not change lanes. is 0; (2) Comfort benefit of the vehicle behind the target lane Tback: ① The target lane rear vehicle Tback avoids, and the comfort benefit of the target lane rear vehicle Tback at time step t for: The total comfort benefit of the vehicle Tback in the target lane after n time steps for: Among them, a Tback is the evasive braking acceleration of the vehicle behind Tback in the target lane, a max is the maximum braking acceleration of the vehicle behind Tback in the target lane; γ is the time step discount factor; ② The vehicle behind the target lane, Tback, does not yield due to a Tback is 0, then the benefit of the vehicle behind the target lane Tback at time step t is 0, the total comfort benefit is 0.

8. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 1 is characterized in that: In step S400, the strategy combination includes: S (C,A) ={ego changes lanes, Tback avoids} S (C,NA) ={ego changes lanes, Tback does not yield} S (NC,A) ={ego does not change lanes, Tback avoids} S (NC,NA) ={ego does not change lanes, Tback does not yield}.

9. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 8, characterized in that: Total Revenue Function R total for: Among them, R safety For safety benefit, R efficiency is the speed gain, R comfort is the comfort benefit; α is the safety benefit influencing factor; β is the speed benefit influencing factor; γ is the comfort benefit influencing factor; The total return of each strategy combination is: ①Strategy combination S (C,A) The total revenue is: in, is the sum of the vehicle ego’s earnings, For the safety benefit of the vehicle ego, is the speed gain of the vehicle ego, is the comfort benefit of the vehicle ego; is the sum of the benefits of the vehicle behind Tback in the target lane, is the safety benefit of the vehicle Tback in the target lane, is the speed gain of the vehicle Tback in the target lane, is the comfort benefit of the vehicle Tback behind the target lane; α is the safety benefit influencing factor; β is the speed benefit influencing factor; γ is the comfort benefit influencing factor; ②Strategy combination S (C,NA) The total revenue is: ③Strategy combination S (NC,A) The total revenue is: ④Strategy combination S (NC,NA) The total revenue is:

10. The dynamic game-based intelligent vehicle lane-changing decision method in foggy weather according to claim 9, characterized in that: The strategy combination with the largest total profit is regarded as the optimal strategy combination.

Images