Automatic driving forced lane changing decision-making method based on dynamic continuous game

Abstract

The invention relates to an automatic driving forced lane changing decision-making method based on a dynamic continuous game, which belongs to the technical field of intelligent connected automobiles and comprises the following steps: acquiring motion state information of a self-vehicle and surrounding vehicles in a forced lane changing scene through a vehicle-mounted sensor and vehicle-to-vehicle communication; calculating a collision boundary; generating an initial decision based on non-cooperative non-zero and dynamic games; optimizing the initial strategy by adopting a game strategy based on information continuity; and outputting a final forced lane change decision under the condition that the strategy after game decision optimization based on information continuity is continuously kept for three periods. According to the automatic driving forced lane changing dynamic game decision-making method considering decision-making continuity, the decision-making model is constructed by adopting non-cooperative non-zero and dynamic games, so that the driving safety, comfort and efficiency are improved; through a dynamic game algorithm based on decision continuity, decision adaptability and predictability are enhanced, stability and consistency of decision output are ensured, and uncertainty and risk are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent connected vehicles, and in particular to an automatic driving forced lane change decision method based on dynamic continuous game. Background Art

[0002] Decision-making is particularly important for autonomous driving because it is responsible for combining perception information to develop safe and efficient driving strategies, including but not limited to lane keeping or lane changing operations. Lane changing accounts for a much higher proportion of serious accidents than speeding and rear-end collisions. More importantly, the proportion of fatal accidents related to lane changing is on the rise, which highlights the urgency and importance of optimizing lane changing decisions for autonomous driving.

[0003] Compared with free lane changes, the mandatory nature of forced lane changes exacerbates traffic congestion, increases the risk of accidents, and causes greater traffic disruption. Effective forced lane change decisions can not only significantly reduce traffic accidents, but also improve road traffic efficiency and alleviate congestion.

[0004] Game theory-based methods have shown significant advantages in multi-vehicle interaction, strategy optimization and equilibrium solution, especially in complex traffic scenarios. In contrast, although rule-based and learning-based methods are also very effective in many cases, they are not as flexible and accurate as game theory-based methods when dealing with dynamic interactions and seeking strategy equilibrium in some unknown scenarios. However, existing game theory-based methods only consider static or discrete game models to simulate the decision-making process between vehicles in forced lane change scenarios, while ignoring the continuous decision-making process under environmental changes, dynamic interactions and multiple factors. Therefore, the present invention proposes an automatic driving forced lane change decision-making method based on dynamic continuous game to cope with complex and changeable decision-making environments, thereby improving the effectiveness of the decision-making process. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an automatic driving forced lane changing decision-making method based on dynamic continuous game, which solves the traffic conflict problem in the forced lane changing scenario, and improves the dynamic optimization capability of decision-making while ensuring the safety, comfort and efficiency of the automatic driving car.

[0006] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0007] A method for automatic driving forced lane change decision-making based on dynamic continuous game, comprising the following steps:

[0008] Step 1: obtaining the motion state information of the vehicle and surrounding vehicles in the forced lane change scenario through on-board sensors and vehicle-to-vehicle communication, wherein the motion state information includes the position, speed and acceleration of the vehicle;

[0009] Step 2: Use the unequal diameter three-circle contour method to model the vehicle boundary and calculate the collision boundary between the vehicle and any surrounding vehicle;

[0010] Step 3: Generate initial decision based on non-cooperative non-zero-sum dynamic game, which includes the following steps:

[0011] Step 31: Construct a forced lane-changing decision model, which includes two participants, namely the ego vehicle and the vehicle behind in the target lane, and formulate game optional strategies for the two participants;

[0012] Step 32: Establish the driving safety benefit function, riding comfort benefit function, and traffic efficiency benefit function of the ego vehicle and the vehicle behind the target lane respectively, and obtain the total benefits corresponding to the ego vehicle and the vehicle behind the target lane by weighted summing the three benefit functions;

[0013] Step 33: The ego vehicle and the vehicle behind the target lane conduct a non-cooperative non-zero-sum dynamic game based on the total benefits of different strategy combinations to generate an initial decision;

[0014] Step 4: Optimize the initial strategy using a game strategy based on information continuity;

[0015] Step 5: Under the condition that the strategy optimized by the game decision based on information continuity is maintained for three consecutive cycles, the final forced lane change decision is output.

[0016] The present invention provides a dynamic game decision method for forced lane change in automatic driving considering decision continuity. Compared with the prior art, the method has the following beneficial effects:

[0017] (1) The present invention introduces non-cooperative non-zero-sum dynamic game and constructs a forced lane change decision model that takes into account driving safety, riding comfort, and traffic efficiency. It makes decisions more intelligently while ensuring safety.

[0018] (2) The present invention uses a dynamic game algorithm based on decision continuity to capture dynamic changes in information flow to enhance the adaptability and predictability of decision-making; in addition, it can also maintain the stability and consistency of decision output to reduce uncertainty and risk in the decision-making process;

[0019] (3) The present invention optimizes the traditional equal-diameter three-circle contour method and takes into account the vehicle's driving speed and the driver's interaction mode during the contour expansion process, which can more appropriately reflect the dynamic and personalized driving status. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the automatic driving forced lane change decision method according to the present invention;

[0021] Figure 2This is a schematic diagram of a forced lane change scenario;

[0022] Figure 3 This is a schematic diagram of the unequal diameter three-circle contour method;

[0023] Figure 4 This is a schematic diagram of a non-cooperative non-zero-sum dynamic game;

[0024] Figure 5 This is a schematic diagram of the game strategy based on information continuity;

[0025] Figure 6 Schematic diagram of the game strategy based on the continuity of output results. DETAILED DESCRIPTION

[0026] The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0027] like Figure 1 As shown, an embodiment of the present invention proposes an automatic driving forced lane change decision method based on dynamic continuous game, which includes the following steps 1 to 5.

[0028] Step 1: Obtain the motion status information of the vehicle and surrounding vehicles in the forced lane change scenario through on-board sensors and vehicle-to-vehicle communication, and identify the forced lane change scenario.

[0029] like Figure 2 As shown in the figure, the vehicles involved in the forced lane change scenario are the ego vehicle (EV), the vehicle in front of the target lane (LV), the vehicle behind the target lane (FV), the vehicle in the third lane (TV), and the stationary obstacle vehicle (BV). When the EV recognizes an obstacle in front, it determines whether it is a stationary obstacle by obtaining the motion state information. When the stationary obstacle vehicle BV is identified, it is considered to have entered the lane change scenario and has a lane change intention. The EV needs to select a suitable gap and merge into the 2 lanes safely, comfortably and efficiently without colliding with the BV and other environmental vehicles (FV, TV, LV). During this process, the EV needs to obtain the position, speed, acceleration and other motion state information of other environmental vehicles.

[0030] Step 2: Calculate the forced lane change collision area and collision boundary.

[0031] In order to reduce the parameters describing the vehicle's geometric outline while reflecting the vehicle's geometric shape to the greatest extent, this embodiment uses the three-circle contour method to model the vehicle boundary. The three-circle contour connects the center of mass coordinates of the vehicle with the centers of the three circles, thereby calculating and determining the collision boundary between the vehicle and any surrounding vehicle, and obtaining the collision area analysis result. During vehicle driving, the size of the area of ​​interest in different directions is different, and the safety of the forward area is more important. Therefore, the optimized equal-diameter three-circle method is used to expand the circular outline in front of the vehicle, such as Figure 3 As shown, the formula for expanding the circular contour in front of the vehicle is as follows:

[0032]

[0033] Among them, R e is the radius of the circle in front of the vehicle; ρ is a fixed coefficient of 0.1; v is the vehicle speed; R safe is the safety factor, which is related to the driver's interaction mode; R is the radius of the circle behind the vehicle and the circle in the middle of the vehicle. The centers of the circle in front of the vehicle, the circle in the middle of the vehicle, and the circle behind the vehicle are (x f ,y f )、(x m , y m)、(x r ,y r ).

[0034] To ensure safety, the outlines of the ego vehicle and other vehicles cannot overlap, and the minimum Euclidean distance between any pair of circles between the ego vehicle and any surrounding vehicle must meet the following conditions:

[0035]

[0036] Where, d i (n,n i ) is the minimum Euclidean distance between the vehicle and any pair of circles around the i-th vehicle, n and n i are the numbers of the circles of the self-car and the surrounding i-th car, n=1,2,3,n i =1,2,3, ε is the safety threshold, g(n,n i ) is the relative distance between the vehicle and the i-th vehicle in the circle, and its calculation formula is as follows:

[0037]

[0038] In the formula, x n and n are the horizontal and vertical coordinates of the nth circle of the vehicle, and The nth car around the i-th car i The horizontal and vertical coordinates of the circle, R e,n is the radius of the nth circle of the vehicle, The nth car around the i-th car i The radius of a circle.

[0039] The greater the distance between the two vehicles is than the safety threshold, the safer it is. When the safety threshold is zero, the two vehicles collide. Therefore, in order to represent extreme situations, the collision boundary of the vehicles needs to be calculated.

[0040] Step 3: Generate initial decisions based on non-cooperative non-zero-sum dynamic games.

[0041] In step three, it is necessary to generate an initial decision for the vehicle based on the motion state information of the vehicle and surrounding vehicles and the collision area analysis results calculated in step two, which specifically includes the following steps three-one to three-three.

[0042] Step 31: Construct a forced lane-changing decision model and formulate optional game strategies.

[0043] The forced lane change decision model includes two participants, namely the ego vehicle (EV) and the target lane following vehicle (FV). The game optional strategies that meet their respective driving conditions are defined for the two participants, as shown in Table 1. EV (1,1) represents the total benefit of the ego vehicle when the ego vehicle chooses the lane change strategy and the vehicle behind the target lane chooses the avoidance strategy. U FV (1,1) represents the total benefit of the target lane rear vehicle when the ego vehicle chooses the lane change strategy and the target lane rear vehicle chooses the avoidance strategy.

[0044] Table 1 Strategy set and benefits of the ego vehicle and the vehicle behind the target lane

[0045]

[0046] Step 32: Establish a game profit function and calculate the total profit of the ego vehicle and the vehicle behind in the target lane under different strategies.

[0047] During the lane changing process, the collision boundary 1 between EV and FV is the left front of EV and the right rear of FV, and the collision boundary 2 is the left rear of the ego vehicle and the right front of the vehicle behind in the target lane. The driving safety benefit is measured based on the time difference between EV and FV reaching the two collision boundaries.

[0048] Specifically, when the ego vehicle chooses the lane change strategy, the driving safety benefit function of the ego vehicle and the vehicle behind in the target lane is:

[0049]

[0050] In the formula, The driving safety benefit of EV when choosing to change lanes for EV, is the driving safety benefit of FV when EV chooses to change lanes, j represents the optional game strategy of FV, j = 1, 2, 3, j = 1 represents the strategy of avoiding the self-vehicle, j = 2 represents the strategy of not avoiding the self-vehicle, j = 3 represents the strategy of keeping unchanged, t E1 and t F2 are the time when EV reaches collision boundary 1 and collision boundary 2, respectively, t F1 and t F2are the time when FV reaches collision boundary 1 and collision boundary 2 respectively, and ΔT is the safety critical time, for example, 3s. The present invention designs the speed of FV with reference to the IDM model, and calculates t F1 and t F2 , and finally the driving safety benefits of EV and FV under different strategy combinations can be obtained.

[0051] When the ego vehicle chooses not to change lanes, both the ego vehicle and the vehicle behind the target lane will drive in the original lane, and no collision will occur. The safety situation of both parties will basically not change, and both are in safe driving. Therefore, the driving safety benefit function of the ego vehicle and the vehicle behind the target lane is:

[0052]

[0053] In the formula, The driving safety benefit of EV when EV chooses not to change lanes. Driving safety benefit of FV when EV chooses not to change lanes.

[0054] Sudden acceleration or deceleration of the vehicle will directly affect the comfort of passengers. Both EV and FV are expected to reduce the inconvenience caused by frequent speed changes. for:

[0055]

[0056] In the formula, k1 and k2 are weight coefficients, and are the longitudinal acceleration and lateral acceleration of EV, t n is the current time, t n+m is the lane change completion time, i represents the optional game strategy of EV, i=1,2, i=1 represents the lane change strategy, and i=2 represents the lane keeping strategy.

[0057] FV's ride comfort benefits for:

[0058]

[0059] In the formula, is the acceleration of the vehicle behind in the target lane, which is determined based on the IDM model to follow the EV or LV.

[0060] When an EV changes lanes, the traffic efficiency gain depends on the longitudinal speed after the lane change. When an EV does not change lanes, its following target is BV, so the traffic efficiency gain of the EV is for:

[0061]

[0062] In the formula, v LV is the speed of LV, v EV is the speed of the EV, v desired is the expected speed of the EV, is the correlation coefficient.

[0063] FV's traffic efficiency benefits for:

[0064]

[0065] In the formula, v FV is the speed of FV, is the correlation coefficient.

[0066] After establishing the above profit functions, the three profit functions of the ego vehicle and the vehicle behind the target lane, namely, the driving safety profit function, the riding comfort profit function, and the traffic efficiency profit function, are weighted and summed, and finally the total profit corresponding to the ego vehicle and the vehicle behind the target lane is obtained. The formula is as follows:

[0067]

[0068] Where U EV (i,j),U FV (i, j) are the total benefits of the ego vehicle and the vehicle behind in the target lane, are the weighted coefficients corresponding to each part of the income.

[0069] Step 33: Conduct non-cooperative non-zero-sum dynamic game solution analysis.

[0070] By establishing the total benefit function of the ego vehicle and the vehicle behind the target lane, the ego vehicle and the vehicle behind the target lane conduct a non-cooperative non-zero-sum dynamic game based on the total benefits of different strategy combinations in the final decision to generate an initial decision.

[0071] When performing game solving analysis, the non-cooperative non-zero-sum dynamic Stackelberg game solving process is transformed into a two-level optimization problem. This process demonstrates the core characteristics of the Stackelberg game, that is, the leader (the ego car) acts according to its own priority decisions, while the follower (the car behind the target lane) reacts based on the leader's behavior. The non-cooperative non-zero-sum dynamic game process is as follows Figure 4 As shown, the two-level optimization problem is shown in the following formula:

[0072]

[0073] In the formula, s EV * is the optimal strategy of the vehicle, S FV * (s EV) is the optimal strategy set of the vehicle behind the target lane when the vehicle strategy is given, S EV is the optional strategy of the ego car, S FV is the optional strategy of the vehicle behind the target lane, s FV * is the optimal strategy for the vehicle behind the target lane, U EV (S EV ,S FV ) Select S for the vehicle EV strategy and the target lane rear vehicle selects S FV The benefit of the car when using the strategy, U FV (s EV ,s FV ) is the car selection s EV strategy and the target lane following vehicle selects s FV The benefit of the vehicle behind the target lane when using the strategy, U FV (s EV ,s FV * ) is the car selection s EV strategy and the target lane following vehicle selects s FV * The payoff of the vehicle behind in the target lane under the optimal strategy.

[0074] Step 4: Use a game strategy based on information continuity to optimize the initial decision.

[0075] The game strategy based on information continuity in this step introduces the concept of cumulative benefits of participants in repeated games, that is, the sum of historical game benefits, and also considers the benefits of multiple factors. This means that previous game results will be considered in the repeated game model, and cumulative benefits will be used to prevent fluctuations due to the influence of instantaneous decision factors, and capture dynamic changes in information flow to enhance the adaptability and predictability of decisions.

[0076] like Figure 5 As shown, suppose the game is repeated from the initial time ts1 to the current time ts n =The gains within T are recorded, the same interactive lane-changing decision game is repeated regularly, and the value of ts1 is adjusted in real time to achieve limited repeated game and optimize the decision model.

[0077]

[0078] Simplify the above formula to:

[0079]

[0080] In the formula, the player is the vehicle or the vehicle behind the target lane. is the payoff of the repeated game at time t when the player adopts the (i, j) strategy. is the game payoff of the game participant player adopting the (i, j) strategy at time T, δ is the time preference factor, Δt is the interval between discrete game decisions, and k is the number of historical repeated games.

[0081] Step 5: Obtain the final decision based on the game strategy of output result continuity.

[0082] The game decision based on the continuity of the output results adopted in this step is based on the condition that the game decision based on the continuity of information is maintained for three consecutive cycles to output the final forced lane change decision. Figure 6 As shown in the figure, the decision state of changing lanes only lasts for one period Δt, so the cruising driving state in the current lane is maintained. If the interactive game decision is directly output, it may cause the vehicle to adjust back to the original lane trajectory immediately after just starting the lane change action. Such frequent adjustments will not only have a negative impact on the continuity of the decision, but also interfere with the stable driving state of the vehicle.

[0083] When the interactive game decision is emergency braking, the ego vehicle system will immediately respond and perform emergency braking operations to prioritize vehicle safety. When the interactive game decision is to change lanes, it will switch to changing lanes after confirmation of the continuous state information sequence to avoid decision jumps.

[0084] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for forced lane change decision-making in automatic driving based on dynamic continuous game, characterized in that: The following steps are involved: Step 1: obtaining the motion state information of the vehicle and surrounding vehicles in the forced lane change scenario through on-board sensors and vehicle-to-vehicle communication, wherein the motion state information includes the position, speed and acceleration of the vehicle; Step 2: Use the unequal diameter three-circle contour method to model the vehicle boundary and calculate the collision boundary between the vehicle and any surrounding vehicle; Step 3: Generate initial decision based on non-cooperative non-zero-sum dynamic game, which includes the following steps: Step 31: Construct a forced lane-changing decision model, which includes two participants, namely the ego vehicle and the vehicle behind in the target lane, and formulate game optional strategies for the two participants; Step 32: Establish the driving safety benefit function, riding comfort benefit function, and traffic efficiency benefit function of the ego vehicle and the vehicle behind the target lane respectively, and obtain the total benefits corresponding to the ego vehicle and the vehicle behind the target lane by weighted summing the three benefit functions; Step 33: The ego vehicle and the vehicle behind the target lane conduct a non-cooperative non-zero-sum dynamic game based on the total benefits of different strategy combinations to generate an initial decision; Step 4: Optimize the initial strategy using a game strategy based on information continuity; Step 5: Under the condition that the strategy optimized by the game decision based on information continuity is maintained for three consecutive cycles, the final forced lane change decision is output.

2. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1, characterized in that: In the unequal diameter three-circle contour method, the contour of the circle in front of the vehicle is expanded according to the following formula: Among them, R e is the radius of the circle in front of the vehicle, ρ is a fixed coefficient of 0.1, v is the vehicle speed, R safe is the safety factor, and R is the radius of the circle behind the vehicle and the circle in the middle of the vehicle.

3. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 2, characterized in that: The collision boundary is determined according to that the minimum Euclidean distance of any pair of circles between the vehicle and any surrounding vehicle is greater than or equal to a safety threshold.

4. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1 or 2, characterized in that: When the ego vehicle chooses the lane change strategy, the driving safety benefit function of the ego vehicle and the vehicle behind it in the target lane is: In the formula, For the driving safety benefit of the vehicle, is the driving safety benefit of the vehicle behind the target lane, j = 1, 2, 3, t E1 and t E2 are the time when the ego vehicle reaches the collision boundary 1 and collision boundary 2 with the vehicle behind in the target lane, respectively, and t F1 and t F2 are the time when the rear vehicle in the target lane reaches the collision boundary 1 and collision boundary 2 between the ego vehicle and the rear vehicle in the target lane, respectively, and ΔT is the safety critical time; When the ego vehicle chooses not to change lanes, the driving safety benefit function of the ego vehicle and the vehicle behind it in the target lane is: In the formula, For the driving safety benefit of the vehicle, is the driving safety benefit of the vehicle behind in the target lane.

5. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 4, characterized in that: The value of the safety critical time ΔT is 3s.

6. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1 or 2, characterized in that: The ride comfort benefit function of the vehicle is: In the formula, is the ride comfort benefit of the vehicle, k1 and k2 are weight coefficients, and are the longitudinal acceleration and lateral acceleration of the vehicle, t n is the current time, t n+m is the lane change completion time, i=1,2; The riding comfort benefit function of the vehicle behind the target lane is: In the formula, is the riding comfort benefit of the vehicle behind the target lane, is the acceleration of the vehicle behind the target lane, which is determined by following the ego vehicle or the vehicle in front of the target lane according to the IDM model.

7. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1 or 2, characterized in that: The traffic efficiency benefit function of the vehicle is: In the formula, is the traffic efficiency benefit of the vehicle, v LV is the speed of the vehicle in front of the target lane, v EV is the speed of the vehicle, v desired is the desired speed of the vehicle, is the correlation coefficient; The traffic efficiency benefit function of the vehicle behind the target lane is: In the formula, is the traffic efficiency gain, v FV is the speed of the vehicle behind in the target lane, is the correlation coefficient.

8. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1 or 2, characterized in that: In step 33, the non-cooperative non-zero-sum dynamic game solution process is transformed into a two-level optimization problem, which is shown in the following formula: In the formula, s EV * is the optimal strategy of the vehicle, S FV * (s EV ) is the optimal strategy set of the vehicle behind the target lane when the vehicle strategy is given, S EV is the optional strategy of the ego car, S FV is the optional strategy of the vehicle behind the target lane, s FV * is the optimal strategy for the vehicle behind the target lane, U EV (s EV ,s FV ) is the car selection s EV strategy and the target lane following vehicle selects s FV The benefit of the car when using the strategy, U FV (s EV ,s FV ) is the car selection s EV strategy and the target lane following vehicle selects s FV The benefit of the vehicle behind the target lane when using the strategy, U FV (s EV ,s FV * ) is the car selection s EV strategy and the target lane following vehicle selects s FV * The payoff of the vehicle behind in the target lane under the optimal strategy.

9. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1 or 2, characterized in that: In step 4, the game strategy based on information continuity introduces the concept of cumulative benefits of participants in repeated games. Suppose that the repeated game from the initial time ts1 to the current time ts n =The benefits within T are recorded, the same interactive lane-changing decision game is repeated regularly, and the value of ts1 is adjusted in real time to achieve limited repeated games and optimize the decision model; Simplify the above formula to: In the formula, the player is the vehicle or the vehicle behind the target lane. is the payoff of the repeated game at time t when the player adopts the (i, j) strategy. is the game payoff of the game participant player adopting the (i, j) strategy at time T, δ is the time preference factor, Δt is the interval between discrete game decisions, and k is the number of historical repeated games.

10. The automatic driving forced lane change decision method based on dynamic continuous game according to claim 1 or 2, characterized in that: When the initial decision is emergency braking, the vehicle immediately responds and performs an emergency braking operation.

Images