An Automatic Lane-Changing Decision-Making Method for the Port Terminal Surface Based on Game Theory

Abstract

The present invention discloses a lane-changing decision-making method for autonomous driving on a port terminal surface based on game theory, including: S1, perceiving the vehicle information around the lane-changing vehicle; S2, judging whether a lane-changing intention is generated according to the relative speed and distance between the lane-changing vehicle and the surrounding vehicles; S3, if a lane-changing intention is generated, proceed to step S4; otherwise, return to step S2; S4, for the vehicle intending to change lanes, turn on the turn signal and make a small lateral movement for a tentative lane change; S5, consider the game between the vehicle intending to change lanes and the vehicle behind in the target lane: construct a lane-changing game payoff matrix according to the strategies of the vehicle intending to change lanes as changing lanes or maintaining the lane and the strategies of the vehicle behind in the target lane as avoiding or not avoiding; S6, make a lane-changing game decision judgment according to the total payoffs of the vehicle intending to change lanes and the vehicle behind in the target lane. This method greatly improves the accurate control of the decision-making module over the lane-changing opportunity under the premise of ensuring safety, and improves the operation efficiency of autonomous vehicles on the terminal surface.

Description

Technical Field

[0001] The present invention relates to the technical field of driverless technology, and particularly relates to an automatic driving lane-changing decision-making method for a port terminal surface based on game theory. Background Art

[0002] In recent years, with the rapid development of artificial intelligence and the breakthrough of sensor technology, driverless technology has been applied in some specific scenarios, such as: parks, mines, and ports. For port terminals, due to the high costs of truck drivers and facility maintenance, which bring cost pressure to port operations, driverless trucks will become the main direction of port intelligent transformation. The loading and unloading process of port containers usually involves three operation links, and an important link is that goods are loaded and unloaded through quay cranes on the terminal surface. When loading and unloading containers on the terminal surface, multi-lane planning and lane-changing situations are often involved. To improve the efficiency and safety of transportation, it is necessary to improve the lane-changing decision-making level.

[0003] As a sub-module of the driverless decision-making and planning module, lane-changing decision-making has always attracted the attention and research of scholars due to the complexity of its process and the key nature of decision-making. The methods commonly used in the lane-changing decision-making module include the gap threshold acceptance model and machine learning models. For the method of the gap threshold acceptance model, lane-changing judgment is made by comparing the distance between the host vehicle and the vehicle in front, and the distance between the host vehicle and the vehicles in front and behind on the target lane. Since the lane-changing model is relatively simple, it is difficult to handle complex lane-changing scenarios. For machine learning methods, a large amount of data is required to train the model, and when the model fails, it is difficult to accurately locate the cause of the failure due to the unexplainable nature of machine learning itself.

[0004] During the lane-changing process, since the vehicle behind on the target lane may choose behaviors such as decelerating to give way or accelerating, the completion or abandonment of any lane-changing behavior is the result of one party's yielding or compromise, manifested as the game balance between the two vehicles. Otherwise, the full competition between the two vehicles will lead to traffic accidents.

[0005] Game theory provides solutions and analysis means for problems with competitive tendencies, and is therefore widely applied in various fields. The main components of each game include participants, the strategies adopted by each participant, and the payoff function brought by the participants adopting corresponding strategies. Therefore, game theory can be applied to study the dynamic interaction of multiple vehicles during the lane-changing process. Currently, there is no report on the automatic driving lane-changing decision-making method for a port terminal surface based on game theory. Summary of the Invention

[0006] To solve the problems existing in the prior art, the present invention provides a lane-changing decision-making method for automatic driving on a port terminal surface, which can achieve a more reasonable lane-changing decision without a large amount of data processing and calculation.

[0007] For this purpose, the present invention adopts the following technical solutions:

[0008] A lane-changing method for automatic driving on a port terminal surface based on game theory, comprising the following steps:

[0009] S1, sensing the information of surrounding vehicles through the body sensors of the lane-changing vehicle;

[0010] S2, judging whether a lane-changing intention is generated according to the relative speed and distance between the lane-changing vehicle and the surrounding vehicles;

[0011] S3, if a lane-changing intention is generated, then enter step S4; otherwise, return to step S2;

[0012] S4, for the vehicle intending to change lanes, start turning on the turn signal and make a small lateral movement to test the lane change;

[0013] S5, considering the game between the vehicle intending to change lanes and the vehicle behind in the target lane: constructing a lane-changing game payoff matrix according to the strategies of the vehicle intending to change lanes as changing lanes or keeping in the lane and the strategies of the vehicle behind in the target lane as avoiding or not avoiding;

[0014] S6, making a lane-changing game decision judgment according to the total payoff of the vehicle intending to change lanes and the vehicle behind in the target lane, where:

[0015] The calculation formula for the total payoff M is:

[0016] M = W1M safety + W2M dist + W3M velocity ,

[0017] In the formula:

[0018] M safety is the safety payoff of the interacting vehicle in the lane-changing game; M dist is the distance payoff of the interacting vehicle in the lane-changing game; M velocity is the speed payoff of the interacting vehicle in the lane-changing game; W1, W2, and W3 are the weights of M safety , M dist , M velocity respectively, and W1 + W2 + W3 = 1;

[0019] After normalizing the safety benefit, distance benefit, and speed benefit, they are introduced into the decision-making model to find the optimal strategy. Let CV be the vehicle intending to change lanes, FV be the vehicle ahead, LFV be the vehicle ahead in the target lane, and LRV be the vehicle behind in the target lane. During the lane-changing process, CV and LRV follow the Stackelberg game of two participants, and the objective function is as follows:

[0020]

[0021] The constraint conditions are defined as follows:

[0022] V i ≥0, i = CV, LRV;

[0023] a min ≤a i ≤a max , i = CV, LRV;

[0024] S CV >D min ;

[0025] In the formula: M i is the total benefit of vehicle i; a i is the possible acceleration of vehicle i, with the unit of m / s 2 ; c CV is whether CV will change lanes; is the optimal acceleration of CV, with the unit of m / s 2 ; is whether changing lanes is beneficial to CV; β 2 is the optimal decision of LRV; λ LRV is the decision of LRV under the known decision of CV; a min is the minimum acceleration of the vehicle, with the unit of m / s 2 ; a max is the maximum acceleration of the vehicle, with the unit of m / s 2 . Among them, the calculation formula of M safety is as follows:

[0026]

[0027] In the formula, SP t is the safety factor at the t-th moment of lane change; SP0 is the safety factor at the start of lane change, and the calculation formula of SP t is as follows:

[0028]

[0029] In the formula, T h is the time headway at the t-th moment of lane change, with the unit of s; T b is the desired time headway, with the unit of s.

[0030] Among them, the M dist is represented by the change in the distance between vehicles during the lane-changing process. Defining RP as the relative distance factor between two interacting vehicles, then:

[0031]

[0032] In the formula, RP t is the relative distance factor at the t-th moment of lane-changing; RP0 is the relative distance factor at the start of lane-changing.

[0033] When the CV and the LRV are on different lanes, the relative distance factor of interaction at the t-th moment is defined as follows:

[0034]

[0035] In the formula: RP LRV-CV (t) represents the relative distance factor of the LRV at the t-th moment, LRV-CV represents the interaction between the LRV and the CV; _LRV represents that the target lane is occupied by the LRV;

[0036] Among them, T LRV-CV represents the time factor between the LRV and the CV, and the expression is as follows:

[0037]

[0038] In the formula: D LRV is the longitudinal distance of the LRV in the road coordinate system (along the road direction); D CV is the longitudinal distance of the CV in the road coordinate system; V LRV is the longitudinal speed of the LRV; V CV is the longitudinal speed of the CV.

[0039] Among them, the calculation formula of the M velocity is as follows:

[0040]

[0041] In the formula: VP t is the speed coefficient at the end of lane-changing; VP0 is the speed coefficient at the start of lane-changing;

[0042] The calculation method of VP is as follows:

[0043]

[0044] The method for judging whether a lane-changing intention is generated in step S2 is as follows:

[0045] Let v CV be the actual vehicle speed of this vehicle, v FV be the actual vehicle speed of the vehicle in front in this lane, vLFV is the speed of the vehicle ahead in the target lane, v LRV is the speed of the vehicle behind in the target lane, c is the speed advantage factor, and there is:

[0046] For the vehicle ahead and the vehicle ahead on the left, the speed factor c satisfies:

[0047]

[0048] For the vehicle behind on the left, the speed satisfies:

[0049] v CV ≥v LRV (3)

[0050] When it is considered that the vehicle ahead is driving slowly, and then a lane-changing intention is generated. Among them, is the threshold value of the speed advantage coefficient, which is obtained through online calibration.

[0051] Among them, the fitting formula for the minimum following safety distance is as follows:

[0052] MFD = 0.0029(v CV ×3.6) 2 + 0.3049(v CV ×3.6) (4)

[0053] The maximum vehicle distance S at which the distance between the vehicle ahead in the current lane and the vehicle itself generates a lane-changing intention max :

[0054] S max =(v CV -v FV )×Δt + MFD + d safe , v CV >v FV (5)

[0055] The minimum vehicle distance S at which the distance between the vehicle ahead in the target lane and the vehicle itself generates a lane-changing intention min,LFV :

[0056] S min,LFV =(v cv -v LFV )×Δt + MFD + d safe , v cv >v LFV (6)

[0057] Or

[0058] S min,LFV = MFD + d safe , v CV ≤v LFV (7)

[0059] The minimum vehicle distance S that generates a lane - change intention based on the distance between the following vehicle in the target lane and the host vehicle min，LRV :

[0060] S min,LRV = 0.0029(v LRV ×3.6) 2 + 0.3049(v LRV ×3.6)+d safe , v CV ≥v LRV (8)

[0061] In the formula, Δt represents the safety - time coefficient of the dynamic process; d safe represents the safety - distance threshold.

[0062] Δt takes a value of 0.5 times the lane - change time; d safe is 2 - 5m.

[0063] The surrounding vehicle information includes the vehicle type, position, speed, and orientation.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] The lane - change decision - making method for autonomous driving on the port terminal surface based on game theory of the present invention comprehensively considers factors such as the minimum safety distance in the dynamic scenario and the benefits of dynamic interaction games among multiple vehicles. For the scenario where lane - changes are often required on the terminal surface, through the method of game theory, on the premise of ensuring safety, it greatly improves the accurate control of the lane - change timing by the decision - making module, and thus improves the operation efficiency of autonomous driving vehicles on the terminal surface. At the same time, the consideration of safety and the improvement of efficiency can better handle the problems of unsafe lane - change behaviors and the possible intervention of safety officers caused by the waste of queuing time due to the inability to change lanes for a long time. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a multi - lane lane - change schematic diagram;

[0067] Figure 2 It is a flowchart of the lane - change decision - making method for autonomous driving on the port terminal surface based on game theory of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0068] The following will describe in detail the lane - change decision - making method for autonomous driving on the port terminal surface based on game theory of the present invention with reference to the accompanying drawings.

[0069] If the traffic flow speed in the adjacent lane is generally fast and the space is large enough, the vehicle intending to change lanes starts to interact with surrounding vehicles by using the turn signal or making a small lateral movement; then, the vehicle intending to change lanes senses the reaction of the vehicle behind in the target lane through the vehicle body sensors, and determines whether to change lanes or maintain the lane according to whether the reaction of the vehicle interacting behind is to accelerate or decelerate.

[0070] The lane-changing process is generally divided into three stages: the generation of lane-changing intention, lane-changing decision-making, and lane-changing execution. Lane-changing decision-making and execution processes involve the dynamic interaction of multiple vehicles, which is a game process among multiple vehicles. Game theory has the advantageous feature of studying the interaction among decision-makers. Therefore, game theory can be applied to study the dynamic interaction of multiple vehicles during the lane-changing process.

[0071] According to the expected payoffs of the two game players, the game can be divided into cooperative games and non-cooperative games: (1) In the case of cooperative games, the two game players form an alliance and share information on the basis of cooperation; (2) In non-cooperative games, the strategies of each participant are based on their own interests, and there is no communication between the two game players. Non-cooperative games can be divided into four types according to the visibility of the action sequence of the participants and the completeness of the information of the participants: complete information dynamic game, complete information static game, incomplete information dynamic game, and incomplete information static game.

[0072] See Figure 2 , the lane-changing decision-making method for automatic driving on the port terminal surface based on game theory of the present invention includes the following steps:

[0073] S1, sense the information of surrounding vehicles through the vehicle body sensors of the lane-changing vehicle, including: category, position, speed, orientation, etc.

[0074] S2, judge whether a lane-changing intention is generated according to the relative speed and distance between the lane-changing vehicle and surrounding vehicles.

[0075] See Figure 1 , where: CV is the vehicle intending to change lanes, FV is the vehicle in front, LFV is the vehicle in front in the target lane (left lane), and LRV is the vehicle behind in the target lane. The decision-making process for the vehicle CV intending to change lanes from the current lane to the target lane is a process of interactive game with other vehicles. The vehicle CV intending to change lanes generates a lane-changing intention in order to avoid some potential risks or seek a larger speed space, following distance, etc. Other vehicles need to consider their own environment to decide whether to yield to the vehicle CV intending to change lanes. Different strategies adopted by surrounding vehicles will affect the final decision result of CV. Therefore, the lane-changing decision-making process of vehicles can be regarded as a game process, which contains all the basic elements of the game and can also reach the game equilibrium.

[0076] The vehicle in front in this lane is moving slowly. The reason that prompts the driver to have the intention of changing lanes is that the target lane has a speed advantage over this lane. Because the slow-moving vehicle in front causes the speed of this vehicle in this lane to gradually decrease, and in order to obtain a faster speed, the vehicle then generates the intention of changing lanes. Therefore, the judgment basis for the speed advantage of the target lane mainly considers factors such as the speed of the vehicle in front in this lane, the desired speed of this vehicle, and the maximum speed limit on the target lane. Let v CV be the actual speed of this vehicle, v FV be the actual speed of the vehicle in front in this lane, v LFV be the speed of the vehicle in front in the target lane, v LRV be the speed of the vehicle behind in the target lane. Furthermore, the present invention defines a method for judging that the vehicle in front is moving slowly.

[0077] The speed advantage factor c reflects the speed advantage of the target lane relative to this lane. Furthermore, formulas (1), (2), and (3) can be obtained. When making a speed advantage judgment, when , it is considered that the vehicle in front is moving slowly, and then the intention of changing lanes is generated. Among them, is the threshold of the speed advantage coefficient, which is obtained through online calibration.

[0078] For the vehicle in front and the vehicle in front on the left, the speed factor c satisfies:

[0079]

[0080] For the vehicle behind on the left, the speed satisfies:

[0081] v CV ≥v LRV (3)

[0082] Regarding the minimum following safety distance of the intelligent vehicle, Japanese scholar T. Kuroda obtained its fitting formula through experiments (Kuroda T, Wakita Y, Shimizu H, et al. Simulation of generation and development of traffic jam in sag zone[J]. Transaction of the Japan Society for Computational Methods in Engineering (in Japanese), 2008, 8(11 - 081128).). The fitting formula for the minimum following safety distance (MFD, Minimum Following Distance) is as follows:

[0083] MFD = 0.0029(v CV ×3.6) 2 +0.3049(vCV ×3.6) (4)

[0084] a. The maximum vehicle distance S at which the distance between the vehicle in front in the current lane and the own vehicle generates a lane-changing intention max :

[0085] S max =(v CV -v FV )×Δt + MFD + d safe ,v CV >v FV (5)

[0086] b. The minimum vehicle distance S at which the distance between the vehicle in front in the target lane and the own vehicle generates a lane-changing intention min,LFV :

[0087] S min,LFV =(v cv -v LFV )×Δt + MFD + d safe ,v cv >v LFV (6)

[0088] or

[0089] S min,LFV =MFD + d safe ,v CV ≤v LFV (7)

[0090] c. The minimum vehicle distance S at which the distance between the vehicle behind in the target lane and the own vehicle generates a lane-changing intention min，LRV :

[0091] S min,LRV =0.0029(v LRV ×3.6) 2 +0.3049(v LRV ×3.6)+d safe ,v CV ≥v LRV (8)

[0092] Where: v CV represents the speed of the own vehicle; v FV represents the speed of the vehicle in front; v LFV represents the speed of the vehicle in front on the left; v LRV represents the speed of the vehicle behind on the left; Δt represents the safety time coefficient in the dynamic process, generally taking a value of 0.5 times the lane-changing duration; d safe represents the safety distance threshold, generally taking 2 - 5m.

[0093] S3. If a lane-changing intention is generated, go to step S4; otherwise, return to step S2, specifically as follows:

[0094] Step 1: Determine whether the target lane has a speed advantage over the current lane. If there is no speed advantage, output that no lane change intention is generated; if there is a speed advantage, proceed to Step 2 for further judgment:

[0095] Step 2: Determine whether the current vehicle distance is less than or equal to the maximum vehicle distance S at which a lane change intention is generated max . If the condition holds, proceed to Step 3 for further judgment; if the condition does not hold, output that no lane change intention is generated;

[0096] Step 3: Determine whether the distance between the vehicle in front of the left vehicle in the target lane and the own vehicle is greater than the minimum vehicle distances S min，LFV and S min，LRV . If the condition holds, output that a lane change intention is generated; if the condition does not hold, output that no lane change intention is generated.

[0097] S4. For the vehicle intending to change lanes, start turning on the turn signal and make a small lateral movement for a trial lane change.

[0098] S5. Consider the game between the vehicle intending to change lanes and the vehicle behind in the target lane. Construct a lane change game payoff matrix according to the strategies of the vehicle intending to change lanes as changing lanes or maintaining the lane and the strategies of the vehicle behind in the target lane as giving way or not giving way.

[0099] The vehicle lane change process is in most cases a game process among multiple vehicles, that is, a process of interaction among multiple vehicles. Assume that the influence of the vehicle in front FV and the vehicle in front of the left LFV on the game process is small, that is, only consider the game between the vehicle intending to change lanes CV and the vehicle behind in the target lane LRV. So the decisions of CV are {changing lanes, maintaining the lane}, and the decisions of LRV are {giving way, not giving way}. Then the vehicle lane change game decisions and results are shown in Table 1.

[0100] During the game process between the two vehicles, they know information such as each other's coordinates, vehicle speeds, and lanes they are in, which is a complete information game; the two vehicles simultaneously adopt strategies such as changing lanes, accelerating, and decelerating, which is a static game; during the game process, the two vehicles are independent individuals, and their decisions are based on their own benefits, which is a non - cooperative game. To sum up, the game between CV and LRV is a non - cooperative static game under complete information, and the basic elements are as follows:

[0101] 1) Participants: The participants in this game are the vehicle changing lanes and the vehicles around it. This invention mainly studies the game phenomenon between CV and LRV, so the main participants are CV and LRV;

[0102] 2) Strategies: CV can choose to change lanes or continue to drive in the original lane; LRV can choose to decelerate and give way to provide a lane change space for CV, or can choose to accelerate to prevent CV's lane change behavior;

[0103] 3) Revenue function: Certain revenues will be brought about by the lane-changing and lane-keeping strategies adopted by the vehicle intending to change lanes; the same applies to the vehicle behind in the target lane.

[0104] During the lane-changing process, the revenues of each participant in the game correspond to the driving revenues of each vehicle. And during the lane-changing process, the revenue function plays an important role in the decision-making process. To generate a reasonable decision-making logic, the present invention designs and considers the combination of three revenue functions. The first function M safety quantifies the safety revenue of the interacting vehicles in the lane-changing game; the second function M dist quantifies the distance revenue of the interacting vehicles in the lane-changing game; the third function M velocity quantifies the speed revenue of the interacting vehicles in the lane-changing game. Specifically as follows:

[0105] (1) Safety revenue:

[0106] The safety revenue is expressed by the vehicle operation safety factor during the lane-changing process. The safety revenue is as follows:

[0107]

[0108] In the formula, SP t is the safety factor at the t-th moment of lane-changing; SP0 is the safety factor at the start of lane-changing.

[0109] The safety factor of the vehicle at the t-th moment of lane-changing is represented by the time headway. The formula is as follows:

[0110]

[0111] In the formula, T h is the time headway (s) at the t-th moment of lane-changing; T b is the expected time headway (s).

[0112] (2) Distance revenue:

[0113] The distance revenue is expressed by the change in the vehicle distance during the lane-changing process. RP is defined as the relative distance factor between two interacting vehicles. The calculation formula for the distance revenue is as follows:

[0114]

[0115] In the formula, RP t is the relative distance factor at the t-th moment of lane-changing; RP0 is the relative distance factor at the start of lane-changing.

[0116] The RP value of the relative distance factor of a vehicle indicates its competitive advantage in the game, which can prevent the 3-second time headway for ensuring safety from being invaded by another vehicle interacting with it. Its value is (-1, 1). When the CV and the LRV are in different lanes, the relative distance factor of their interaction at the t-th moment is as follows:

[0117]

[0118] Where: RP LRV-CV (t) represents the relative distance factor of the LRV at time t, LRV-CV represents the interaction between the LRV and the CV; _LRV indicates that the target lane is occupied by the LRV;

[0119] Among them, T LRV-CV represents the time factor of the interaction between the LRV and the CV, and the expression is as follows:

[0120]

[0121] Where: D LRV is the longitudinal distance of the LRV in the road coordinate system (along the road direction); D CV is the longitudinal distance of the CV in the road coordinate system; V LRV is the longitudinal speed of the LRV; V CV is the longitudinal speed of the CV.

[0122] As can be seen from the above formulas, the spatial factors for vehicles far apart and traveling in parallel are -1 and 1 respectively. If the two vehicles are not far apart, the distance coefficient will gradually increase as the relative distance increases.

[0123] (3) Speed benefit:

[0124] The benefit function Mvelocity is represented by the change in the speed factor during the lane change process, and VP is defined as the relative speed function between two interacting vehicles. The benefit function formula is as follows:

[0125]

[0126] Where: VP t is the speed coefficient at the end of the lane change; VP0 is the speed coefficient at the start of the lane change.

[0127] The calculation method of VP is as follows:

[0128]

[0129] (4) Total benefit:

[0130] The total benefit function is a linear combination of the three benefits, as follows:

[0131] M = W1M safety + W2M dist + W3M velocity (17)

[0132] Where: W i is the weight of each benefit, and W1 + W2 + W3 = 1.

[0133] After determining the lane-changing revenue function of the multi-vehicle dynamic interaction game, the vehicles in the dynamic interaction game need to make decisions based on their own driving revenues. For the convenience of decision-making, the revenue matrices of vehicle CV and vehicle LRV are shown in Table 1:

[0134] Table 1 Game Revenue Matrix of CV and LRV

[0135]

[0136] M CV and M LRV are the revenues of CV and LRV respectively.

[0137] S6. Make a lane-changing game decision judgment according to the total revenue of the vehicle to be lane-changed and the vehicle behind in the target lane.

[0138] After normalizing the safety revenue, distance revenue, and speed revenue, introduce them into the decision-making model to find the optimal strategy. Solution to the multi-vehicle dynamic interaction game in the lane-changing process: From the perspective of CV, while considering the reaction strategy of LRV, maximize the lower limit of the driving revenue. In other words, this solution is a set of strategies that maximize the revenues of both parties in the interaction game in the worst-case scenario under the satisfaction of the constraint conditions. During the lane-changing process, CV and LRV follow the Stackelberg game of 2 players. It is a two-layer optimization problem, and the objective function is as follows:

[0139]

[0140] Since it is impossible for the game players to predict and respond to the operating state of the other party very accurately and timely, therefore, the minimum safety distance D between vehicles should be added to the constraint conditions min , so the constraint conditions are defined as follows:

[0141] V i ≥0, i = CV, LRV (20)

[0142] a min ≤a i ≤a max , i = CV, LRV (21)

[0143] S CV >D min (22)

[0144] In the formula: M i is the total revenue of vehicle i; a i is the possible acceleration of vehicle i (m / s 2 ); c CV is whether CV will change lanes; The optimal acceleration of the CV (m / s 2 ); Whether lane changing is beneficial to the CV; β 2 The optimal decision of the LRV; λ LRV The decision of the LRV under the known decision of the CV; a min The minimum acceleration of the vehicle (m / s 2 ); a max The maximum acceleration of the vehicle (m / s 2 ).

Claims

1. A lane-changing decision-making method for autonomous driving on a port terminal surface based on game theory, comprising the following steps: S1. Perceive the information of surrounding vehicles through the body sensors of the lane-changing vehicle; S2. Judge whether a lane-changing intention is generated according to the relative speed and distance between the lane-changing vehicle and the surrounding vehicles; S3. If a lane-changing intention is generated, go to step S4; otherwise, return to step S2; S4. For the vehicle intending to change lanes, start turning on the turn signal and make a small lateral movement to test lane-changing; S5. Consider the game between the vehicle intending to change lanes and the vehicle behind in the target lane: construct a lane-changing game payoff matrix according to the strategies of the vehicle intending to change lanes as changing lanes or maintaining the lane and the strategies of the vehicle behind in the target lane as avoiding or not avoiding; S6. Make a lane-changing game decision judgment according to the total payoffs of the vehicle intending to change lanes and the vehicle behind in the target lane, where: The calculation formula for the total payoff M is: M = W1M safety + W2M dist + W3M velocity , In the formula: M safety is the safety benefit of the interacting vehicle in the lane-changing game; M dist is the distance benefit of the interacting vehicle in the lane-changing game; M velocity is the speed benefit of the interacting vehicle in the lane-changing game; W1, W2, and W3 are the weights of M safety , M dist , M velocity respectively, and W1 + W2 + W3 = 1; After normalizing the safety payoff, distance payoff and speed payoff, introduce them into the decision-making model to find the optimal strategy. Let CV be the vehicle intending to change lanes, FV be the vehicle in front, LFV be the vehicle in front in the target lane, and LRV be the vehicle behind in the target lane. During the lane-changing process, CV and LRV follow the Stackelberg game of two participants, and the objective function formula is: The constraint conditions are defined as follows: V i ≥ 0, i = CV, LRV; a min ≤a i ≤a max , i = CV, LRV; S CV >D min ; Where: M i is the total revenue of vehicle i; a i is the possible acceleration of vehicle i, with the unit of m / s 2 ; c CV indicates whether the CV will change lanes; is the optimal acceleration of the CV, with the unit of m / s 2 ; indicates whether changing lanes is beneficial to the CV; β 2 is the optimal decision of the LRV; λ LRV is the decision of the LRV under the known decision of the CV; a min is the minimum acceleration of the vehicle, with the unit of m / s 2 ; a max is the maximum acceleration of the vehicle, with the unit of m / s 2 .

2. The method for automatic driving lane-changing decision of a port terminal surface based on game theory according to claim 1, wherein M safety The calculation formula is as follows: where, SP t is the safety factor at the t-th moment of lane change; SP0 is the safety factor at the start of lane change, and SP t is calculated as follows: where T h is the time headway at the t-th moment of lane change, with the unit of s; T b is the desired time headway, with the unit of s.

3. The method for automatic driving lane-changing decision-making on a port terminal surface based on game theory according to claim 1, wherein The said M dist Represented by the change in the distance between vehicles during the lane-changing process, define RP as the relative distance factor between two interacting vehicles, then the calculation formula for the distance benefit is as follows: where, RP t is the relative distance factor at the t-th moment of lane change; RP0 is the relative distance factor at the start of lane change.

4. The method for automatic driving lane-changing decision-making on a port terminal surface based on game theory according to claim 3, wherein When CV and LRV are on different lanes, the relative distance factor of their interaction at time t is defined as follows: Where: RP LRV-CV (t) represents the relative distance factor of the LRV at time t, LRV-CV represents the interaction between the LRV and the CV; _LRV indicates that the target lane is occupied by the LRV; Among them, T LRV-CV represents the time factor of LRV and CV, and the expression is as follows: Where: D LRV is the longitudinal distance of the LRV in the road coordinate system; D CV is the longitudinal distance of the CV in the road coordinate system; V LRV is the longitudinal speed of the LRV; V CV is the longitudinal speed of the CV.

5. The lane-changing decision-making method for automatic driving on the port terminal surface based on game theory according to claim 1, characterized in that, M velocity The calculation formula is as follows: Where: VP t is the speed coefficient at the end of lane change; VP0 is the speed coefficient at the start of lane change; The calculation method of VP is as follows:

6. The method for automatic driving lane-changing decision of a port terminal surface based on game theory according to claim 1, characterized in that, The method for judging whether a lane-changing intention is generated in step S2 is as follows: Let v CV be the actual vehicle speed of this vehicle, v FV be the actual vehicle speed of the vehicle ahead in this lane, v LFV be the vehicle speed of the vehicle ahead in the target lane, v LRV be the vehicle speed of the vehicle behind in the target lane, c is the speed advantage factor, and there is: For the vehicle in front and the vehicle in front on the left, the speed factor c satisfies: For the vehicle behind on the left, the speed satisfies: v CV ≥ v LRV (3) When it is considered that the vehicle in front is moving slowly, and then a lane-changing intention is generated, where is the threshold value of the speed advantage coefficient, which is obtained through online calibration.

7. The method for automatic driving lane-changing decision of a port terminal surface based on game theory according to claim 6, characterized in that, The fitting formula for the minimum following safety distance is as follows: MFD = 0.0029(v CV × 3.6) 2 + 0.3049(v CV × 3.6)(4) The maximum vehicle distance S at which the distance between the vehicle in front in the current lane and the host vehicle generates a lane change intention max : S max = (v CV - v FV ) × Δt + MFD + d safe , v CV > v FV (5) The minimum vehicle distance S that generates a lane-changing intention based on the distance between the vehicle in front in the target lane and the host vehicle min,LFV : S min,LFV = (v cv - v LFV ) × Δt + MFD + d safe , v cv > v LFV (6) Or S min,LFV = MFD + d safe , v CV ≤ v LFV (7) The minimum vehicle distance S that generates a lane change intention based on the distance between the vehicle behind in the target lane and the host vehicle min，LRV : S min,LRV = 0.0029(v LRV × 3.6) 2 + 0.3049(v LRV × 3.6) + d safe ,v CV ≥ v LRV (8) where Δt represents the safety time factor of the dynamic process; d safe represents the safety distance threshold.

8. The method for making an automatic driving lane-changing decision on a port terminal surface based on game theory according to claim 7, characterized in that: Δt takes a value of 0.5 times the lane-changing duration; d safe is 2 to 5 m.

9. The lane-changing decision-making method for automatic driving on the port terminal surface based on game theory according to any one of claims 1-8, characterized in that: The surrounding vehicle information includes the category, position, speed and orientation of the vehicle.

Images