A method for allocating driving rights for human-machine co-driving based on incomplete information game

Abstract

The present invention provides a method for allocating driving authority for human-machine co-driving based on incomplete information game, including: performing driving risk analysis according to the driver's state and operating ability, vehicle driving safety and driving risk field force; taking the driver's driving style as unknown information, constructing a benefit function under incomplete information according to the driving risk analysis and human-machine driving performance; considering the benefit distribution between the human and the machine, using game theory to describe the interaction relationship between the human and the machine, and solving the optimal driving authority under the game equilibrium condition to achieve human-machine collaborative control; the present invention constructs a human-machine benefit function under incomplete information conditions and designs a driving authority allocation strategy through game theory, so as to realize a realistic human-machine co-driving driving authority allocation model, which can meet the driving needs of different types of drivers, thereby improving the human-machine collaborative control performance.

Description

Technical Field

[0001] The implementation of the present invention relates to the field of vehicle control technology, and specifically to a method for allocating human-machine co-driving driving authority based on incomplete information game. Background Art

[0002] The development of vehicle autonomous driving can assist drivers in completing driving tasks in most scenarios, reduce the burden on drivers, and greatly reduce the accident rate and improve traffic efficiency. However, due to technical limitations, legal constraints, and liability determination, it is difficult to achieve fully autonomous driving in a short period of time. Therefore, human-machine co-driving has become a mainstream research direction for autonomous driving.

[0003] In the process of shared human-machine co-driving, the driver and the machine jointly control the vehicle and both have control over the vehicle. In the process of co-driving, human-machine conflicts are inevitable, which may increase the driver's driving burden and reduce the stability of the vehicle. Therefore, studying how to reasonably allocate driving rights has become one of the main issues in current human-machine co-driving. The current research directions of driving authority allocation mainly include the following aspects: driving authority allocation strategy based on game theory, driving authority allocation strategy based on control theory, driving authority allocation strategy based on intention recognition and other driving authority allocation strategies. Game theory is an effective method to describe the coupling interaction between two or more people and resolve conflicts between players. However, most of the existing research on driving authority allocation based on game theory is based on the assumption of complete information, which is inconsistent with the facts. In actual practice, the driver and the machine cannot obtain all the game information. For example, the driver does not fully know the machine's payoff function and may estimate it based on his own experience or other auxiliary measures; the machine cannot fully determine the driver's driving style in the early stage of collaborative control, and it is difficult to provide accurate driving assistance. In order to realistically model the allocation of driving authority for human-machine co-driving and meet the driving needs of different types of drivers, it is necessary to study the allocation of driving authority for human-machine co-driving based on incomplete information game. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a method for allocating human-machine co-driving driving authority based on incomplete information game, aiming to create a realistic human-machine co-driving driving authority allocation model that can meet the driving needs of different types of drivers, thereby improving the human-machine collaborative control performance.

[0005] In order to achieve the purpose of the present invention, the present invention will be implemented by adopting the technical solution described below.

[0006] A method for allocating driving authority for human-machine co-driving based on incomplete information game, comprising the following steps:

[0007] S1. Perform driving risk analysis based on the three factors of "people-vehicle-road", wherein the driving risk analysis includes driver factors, vehicle factors and road factors, wherein the driver factors include the driver's cognitive state and the driver's control ability; the vehicle factors include the lateral and longitudinal control effects of the vehicle and the stability of the vehicle; and the road factors include the lateral and longitudinal driving risk field forces;

[0008] S2. Based on the unknown information of the driver's driving style, a benefit function under incomplete information conditions is constructed according to the driving risk analysis and the driving performance of the human and the machine. The benefit function includes a driver benefit function and a machine benefit function. The driver benefit function includes a driver driving risk benefit function and a driver driving performance benefit function; the machine benefit function includes a machine driving risk benefit function and a machine driving performance benefit function, wherein:

[0009] The driver's driving risk-benefit function includes the driver's cognitive state, the driver's control ability, the vehicle's lateral and longitudinal control effects, the vehicle's stability, and the driving risk field force;

[0010] The machine driving risk-benefit function is constructed by taking the driver's driving style as unknown information and introducing a driving style weight coefficient. The machine driving risk-benefit function includes the driver's cognitive state, the driver's control ability, the vehicle's lateral and longitudinal control effects, the vehicle's stability, the driving risk field force and the driving style weight coefficient;

[0011] S3. Based on the distribution of benefits between humans and machines, game theory is used to describe the interaction between humans and machines, and the driving authority function under the equilibrium condition of human-machine game is constructed through Nash product;

[0012] S4. Solve the driving authority function to obtain driving authority to achieve human-machine collaborative control.

[0013] As a preferred solution of the present invention, the evaluation index E of the driver factor is d is the evaluation index E of the driver's cognitive state dc and the evaluation index E of the driver's control ability dhp Evaluated, among which,

[0014] The evaluation index E of the driver's cognitive state dc The calculation formula is:

[0015]

[0016] Among them, E dc is the evaluation index of the driver's cognitive state; σ1, σ2 and σ3 are all tuning parameters; δ dis the front wheel turning angle of the driver; At is the driver's attention level, which is calculated by collecting driver-related information from external equipment;

[0017] The evaluation index E of the driver's control ability dhp The calculation formula is:

[0018]

[0019] Among them, E dhp It is the evaluation index of the driver's control ability; d and δ f are the driver's front wheel steering angle and the vehicle's actual front wheel steering angle respectively; y is the actual lateral offset of the vehicle; e ymax is the maximum lateral displacement allowed, the specific value can be determined according to actual requirements; v x is the actual longitudinal velocity of the vehicle; v xdes is the expected longitudinal speed, which is obtained from the speed planning part;

[0020] Then, the evaluation index E of the driver factor is d It is expressed as:

[0021] E d =ω dc E dc +ω dhp E dhp ;

[0022] Among them, E d Represents the driver factor in driving risk analysis; ω dc and ω dhp is the weight coefficient; E dc is the evaluation index of the driver's cognitive state; E dhp It is an evaluation index of the driver's control ability.

[0023] As a preferred solution of the present invention, the evaluation index E of the vehicle factor is v It is the evaluation index E of the vehicle's lateral and longitudinal control effect. ve and vehicle stability evaluation index E vs Evaluated, among which,

[0024] The evaluation index E of the horizontal and vertical control effect is ve The calculation formula is:

[0025]

[0026] Among them, E ve It is the evaluation index of the vehicle's lateral and longitudinal control effect; y is the actual lateral offset of the vehicle; eymax is the maximum lateral offset allowed, and the specific value is determined according to actual requirements; v x is the actual longitudinal velocity of the vehicle; v xdes is the expected longitudinal speed, which is obtained from the speed planning part;

[0027] The evaluation index E of the vehicle stability vs The calculation expression is:

[0028]

[0029] Among them, E vs is the evaluation index of vehicle stability; a y is the actual lateral acceleration of the vehicle; a ymax is the maximum allowed lateral acceleration; v y is the actual lateral velocity of the vehicle; v ydes is the expected lateral velocity; σ(β) is the standard deviation of the center of mass sideslip angle; β is the center of mass sideslip angle;

[0030] Then, the evaluation index E of the vehicle factor is v It is expressed as:

[0031] E v =ω ve E ve +ω vs E vs ;

[0032] Among them, E v Represents the vehicle factor in driving risk analysis; ω ve and ω vs is the weight coefficient; E ve It is the evaluation index of the vehicle's lateral and longitudinal control effect; E vs It is an evaluation index of vehicle stability.

[0033] As a preferred embodiment of the present invention, the road factor E rf The formation process includes the following steps:

[0034] S41. The transverse and longitudinal risk field forces are calculated based on the distance from the vehicle position to the road boundary, the road speed limit and the collision time. The expression is:

[0035]

[0036]

[0037] F l =ω lv F lv +ω lttc F lttc ;

[0038] Among them, F c represents the lateral risk field force; k1 is a constant coefficient; k2 is the lateral slope adjustment coefficient; m represents the vehicle mass; v represents the vehicle speed; y d is the distance from the center of gravity of the vehicle to the road boundary; r0 represents the driver's risk focus; r1 represents the impact range of the road boundary; D represents the lane width; F lv represents the longitudinal risk field force considering the road speed limit; v l,min and v l,nax Respectively represent the minimum and maximum speeds allowed on the road; k x and k y Respectively represent the longitudinal and lateral gradient adjustment coefficients of vehicle speed; x p and p Respectively represent the longitudinal and lateral distances between the vehicle and other traffic participants; F lttc represents the longitudinal risk field force considering the collision time ttc; L represents the relative distance; ttc represents the collision time (Time To Collision); F l Represents the vertical risk field force; ω lv and ω lttc All are weight coefficients;

[0039] S42, evaluating the horizontal and vertical risk field forces to obtain a vertical and horizontal integrated risk assessment model;

[0040] S43, the horizontal risk field force F c and the longitudinal risk field force F l Normalization is performed to obtain the standardized driving risk field force F q :

[0041]

[0042] Among them, F q represents the standardized driving risk field force; E rf Represents the road factors in driving risk analysis.

[0043] As a preferred embodiment of the present invention, when E is satisfied d ≥ε d 、E v ≥ε v and E rf ≥ε rf When any one or both of the following conditions are met, the machine will remind the driver to pay attention to safety, but the driver will be kept driving alone; d , ε v and ε rf They are the thresholds for driver factors, vehicle factors and road factors in driving risk analysis respectively.

[0044] As a preferred embodiment of the present invention, when E d ≥ε d And E v ≥ε v And E rf ≥ε rf , indicating that the driving risk exceeds the risk threshold, the machine will take over the driving rights of the vehicle, and the vehicle will be controlled collaboratively by man and machine; when the driving risk is reduced, the driving rights of the vehicle will be returned to the driver.

[0045] As a preferred embodiment of the present invention, the process of establishing the driver benefit function and the machine benefit function under the incomplete information condition comprises the following steps:

[0046] S71, assuming the driver's driving style to be aggressive, normal, and weak, and assuming the machine's driving style to be normal;

[0047] S72. Based on the assumptions in S71, construct a driver driving performance benefit function and a machine driving performance benefit function, wherein:

[0048] The expression of the driver's driving performance benefit function is as follows:

[0049] E dpi = sin{η1×arctan[DS i ×η2q i -η3(DS i ×η2q i -arctan(DS i ×η2q i ))]};

[0050] Among them, E dpi is the driver's driving performance benefit function, where i=1 represents aggressive driving, i=2 represents normal driving, and i=3 represents weak driving; DS i is a coefficient related to the driver's driving style, where i = 1, 2, 3; q i is the driving authority of drivers with different driving styles, where i = 1, 2, 3; η1, η2 and η3 are all tuning parameters; the machine driving performance benefit function expression is as follows:

[0051] E mp = sin{η1×arctan[η2(1-q i )-η3(η2(1-q i )-arctan(η2(1-q i )))]};

[0052] Among them, E mp is the machine driving performance benefit function; qi is the driving authority of different types of drivers, where i = 1, 2, 3; η1, η2 and η3 are all tuning parameters;

[0053] S73, constructing a driver driving risk-benefit function and a machine driving risk-benefit function based on the assumptions in S71 and the lateral and longitudinal driving risk field forces in the driving risk analysis, wherein:

[0054] The expression of the driver's driving risk-benefit function is as follows:

[0055] E Fqdi =(1+exp((β1+β2(S i -β3))F q -β4)) -1 i=1,2,3;

[0056] Among them, E Fqdi is the driver's driving risk-return function, where i = 1 represents the aggressive type, i = 2 represents the normal type, and i = 3 represents the weak type; β1, β2, β3, and β4 are all tuning parameters; S i is the risk-field return parameter associated with driving style, where i = 1, 2, 3; F q To standardize the driving risk field force;

[0057] The expression of the machine driving risk-return function is:

[0058] E Fqm =1-(1+exp((β1+β2(S i -β3))F q -β4)) -1 i=1,2,3;

[0059] Among them, E Fqm is the risk-return function of machine driving; β1, β2, β3 and β4 are all tuning parameters; S i is the driving risk field benefit parameter related to driving style, where i = 1, 2, 3; F q To standardize the driving risk field force;

[0060] S74: construct a driver benefit function and a machine benefit function according to the driving risk analysis and the driver driving performance benefit function and the machine driving performance benefit function, wherein:

[0061] The specific expression of the driver benefit function is as follows:

[0062] Aggressive: J d1 =ω d1 E d +ω v1 E v+ω Fqd1 E Fqd1 +ω dp1 E dp1 ;

[0063] Normal type: J d2 =ω d2 E d +ω v2 E v +ω Fqd2 E Fqd2 +ω dp2 E dp2 ;

[0064] Weak type: J d3 =ω d3 E d +ω v3 E v +ω Fqd3 E Fqd3 +ω dp3 E dp3 ;

[0065] Among them, J d1 is the aggressive driver benefit function; ω d1 ,ω v1 ,ω Fqd1 and ω dp1 is the weight coefficient of the aggressive driver benefit function; E d is the driver factor in driving risk analysis; E v is the vehicle factor in driving risk analysis; E Fqd1 is the risk-benefit function of an aggressive driver; E dp1 is the driving performance benefit function of the aggressive driver; J d2 is the normal driver benefit function; ω d2 ,ω v2 ,ω Fqd2 and ω dp2 is the weight coefficient of the normal driver benefit function; E Fqd2 is the risk-benefit function of a normal driver; E dp2 is the driving performance benefit function of a normal driver; J d3 is the weak driver benefit function; ω d3 ,ω v3 ,ω Fqd3 and ω dp3 is the weight coefficient of the weak driver benefit function; E Fqd3 is the risk-benefit function of the weak driver; E dp3 The driving performance benefit function for weak drivers;

[0066] The expression of the machine profit function is as follows:

[0067] Radical, J m1 =ω dm1 E d +ω vm1 E v +ω Fqm1 E Fqm +ω mp1 E mp ;

[0068] Normal type, J m2 =ω dm2 E d +ω vm2 E v +ω Fqm2 E Fqm +ω mp2 E mp ;

[0069] Weak type, J m3 =ω dm3 E d +ω vm3 E v +ω Fqm3 E Fqm +ω mp3 E mp ;

[0070] Among them, J m1 is the machine profit function when the driver's driving style is aggressive; ω dm1 ,ω vm1 ,ω Fqm1 and ω mp1 is the weight coefficient; E Fqm is the risk-return function of machine driving; E mp is the machine driving performance benefit function; J m2 is the revenue function of the machine when the driver’s driving style is normal; ω dm2 ,ω vm2 ,ω Fqm2 and ω mp2 is the weight coefficient; J m3 is the revenue function of the machine when the driver’s driving style is weak; ω dm3 ,ω vm3 ,ω Fqm3 and ω mp3 is the weight coefficient;

[0071] S75. Construct a profit function of the machine under incomplete information conditions according to the driver profit function and the machine profit function, and the expression thereof is as follows:

[0072] J m =pi＝1 J m1 +p i＝2 J m2 +p i＝3 J m3 ;

[0073] Among them, J m is the profit function of the machine under incomplete information conditions; p i＝1 、p i＝2 and p i＝3 Both are driving style weight coefficients.

[0074] As a preferred solution of the present invention, the weight coefficients p of different driving styles are determined by double-layer fuzzy logic control. i＝1 、p i＝2 and p i＝3 , the specific steps are as follows:

[0075] S81, determining the weight coefficient p of the driver's driving style as an aggressive type i＝1

[0076] Determine the aggressive weight coefficient p from both horizontal and vertical perspectives i＝1 , where the lateral factor is the degree of human-machine conflict related to the front wheel turning angle δ con , the longitudinal factor is the impact degree A related to the longitudinal acceleration jerk ;

[0077] Using the first layer of fuzzy logic rules to establish p i＝1 , δ con and A jerk The connection between the three is set con and A jerk is the input of the first layer of fuzzy logic, p i＝1 is the output of the first layer of fuzzy logic; where p i＝1 , δ con and A jerk The value range of the three is [0,1], and the relevant fuzzy subsets are divided into {S, MS, M, MB, B} from small to large;

[0078] S82, determining the weight coefficient p of the driver's driving style as normal i＝2

[0079] According to the vehicle stability index E vs and the first layer fuzzy logic output radical weight coefficient p i＝1 To determine the weight coefficient p of the normal driving style i＝2 , and use the second layer of fuzzy logic rules to establish E vs 、p i＝1 and p i＝2 The connection between the three is setvs and p i＝1 is the input of the second layer of fuzzy logic. The value range of the variables is [0,1]. i＝2 is the output of the second layer of fuzzy logic, and its value range is [0,1-p i＝1 ]; Among them, the fuzzy subsets are divided into {S, MS, M, MB, B} from small to large;

[0080] S83, determining the weight coefficient p of the driver's driving style as a weak type i＝3

[0081] According to the determined p i＝1 and p i＝2 The value of p i＝1 +p i＝2 +p i＝3 =1 to determine the weight coefficient p of the weak driving style i＝3 .

[0082] As a preferred solution of the present invention, the first layer of fuzzy logic rules:

[0083]

[0084] The second layer of fuzzy logic rules:

[0085]

[0086]

[0087] As a preferred solution of the present invention, the process of constructing the driving authority function includes the following steps:

[0088] S101. According to the definition of Nash product, the driver's profit function and the machine's profit function are taken as multipliers of the Nash product, and the expression is as follows:

[0089] (J di -J dih )·(J m -J mh );

[0090] Among them, J di is the driver’s benefit function, where i=1 represents the aggressive type, i=2 represents the normal type, and i=3 represents the weak type; J dih is the driver's gain at the negotiation disagreement point; J m is the revenue function of the machine; J mh for the machine’s gains at the divergence point in the negotiation;

[0091] S102. When the human-machine game reaches a game equilibrium, the driver's driving authority q iThe product W of the driver's profit function and the machine's profit function satisfies the following expression:

[0092] W = argmax[J di (q i )-J dih ]·[J m (q i )-J mh ];

[0093] Among them, the product W is about driving authority q i Function of J di (q i ) is when the driving authority is q i The driver's benefit when i=1 is the aggressive type, i=2 is the normal type, and i=3 is the weak type; q i is the driver's driving authority, where i = 1, 2, 3; J m (q i ) is when the driving authority is q i The machine revenue at that time;

[0094] S103, convert the product W into an invisible function: W = F(q i );

[0095] Among them, F(q i )=argmax[J di (q i )-J dih ]·[J m (q i )-J mh ];q i is the driver’s driving authority, where i = 1, 2, 3;

[0096] Therefore, the driving authority q i Expressed as the inverse form of the implicit function: q i =F -1 (W).

[0097] Beneficial Effects

[0098] 1. Aiming at the problems that there are few studies on driving authority allocation based on game theory and that the game model is generally established based on the complete information assumption, the present invention proposes a human-machine co-driving driving authority allocation method based on incomplete information game, which can meet the driving needs of different types of drivers with a realistic human-machine co-driving driving authority allocation model, thereby improving the human-machine collaborative control performance;

[0099] 2. Comprehensively consider the three factors of "people-vehicle-road" to conduct driving risk analysis, including driver status and operating ability, vehicle control effect and stability, and driving risk field;

[0100] 3. Considering the distribution of benefits between humans and machines, game theory is used to describe the human-machine negotiation interaction process, and the optimal driving authority under the game equilibrium conditions is solved through Nash product. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] Figure 1 A simplified flowchart of a method for allocating human-machine co-driving rights based on incomplete information game according to the present invention;

[0102] Figure 2 A driving authority allocation framework diagram of a method for allocating driving authority for human-machine co-driving based on incomplete information game according to the present invention;

[0103] Figure 3 The membership function of the impact degree, the membership function of the human-machine conflict degree, the membership function diagram of the radical probability and the radical probability p in the first layer of fuzzy control of the human-machine co-driving authority allocation method based on incomplete information game described in the present invention are shown in FIG. i＝1 3D graph; where: (a) is the membership function diagram of the impact degree; (b) is the membership function diagram of the human-machine conflict degree; (c) is the membership function diagram of the radical probability; (d) is the membership function diagram of the radical probability p i＝1 3D graph of

[0104] Figure 4 The membership function of the radical probability, the membership function of the vehicle stability, the membership function diagram of the normal probability and the normal probability p in the second-layer fuzzy control of the human-machine co-driving driving authority allocation method based on incomplete information game described in the present invention are shown in FIG. i＝1 3D graph; where: (a) is the membership function diagram of the radical probability; (b) is the membership function diagram of the vehicle stability; (c) is the membership function diagram of the normal probability; (d) is the membership function diagram of the normal probability p i＝1 3D illustration. DETAILED DESCRIPTION

[0105] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings:

[0106] As a specific embodiment of the present invention, Figure 1 and Figure 2 As shown, a method for allocating human-machine co-driving driving rights based on incomplete information game includes the following steps:

[0107] S1. Analyze driving risks based on the three factors of “people-vehicle-road”:

[0108] Driving risk analysis is conducted by considering the three factors of "people-vehicle-road", including the driver's status and control ability, vehicle driving safety, and driving risk field force.

[0109] 1) Drivers:

[0110] In the actual driving process, changes in the driver's cognitive state will affect the driver's driving ability. When the cognitive load is too large or too small, the driver's driving ability will decline. Therefore, it is necessary to consider the driver's cognitive state when performing driving risk analysis. The present invention calculates the driver's cognitive state from the driver's front wheel angle and the driver's attention, and the calculation formula is:

[0111]

[0112] Among them, E dc is the evaluation index of driver cognitive state; σ1, σ1 and σ1 are all tuning parameters; δ d is the front wheel turning angle of the driver; At is the driver's attention level, which can be obtained by external equipment (such as an eye tracker) to obtain driver-related information and calculate it, which will not be discussed in detail here.

[0113] In addition, the driver's ability to control the vehicle directly affects the vehicle's driving safety. The ratio of the deviation between the front wheel steering angle imposed by the driver and the actual front wheel steering angle of the vehicle to the maximum deviation, the ratio of the actual lateral offset to the maximum allowed lateral offset, and the ratio of the actual longitudinal speed to the expected longitudinal speed are used as evaluation indicators of the driver's control ability. The calculation formula is:

[0114]

[0115] Among them, E dhp It is the evaluation index of the driver's control ability; d and δ f are the driver's front wheel steering angle and the vehicle's actual front wheel steering angle respectively; y is the actual lateral offset of the vehicle; e ymax is the maximum lateral displacement allowed, the specific value can be determined according to actual requirements; v x is the actual longitudinal velocity of the vehicle; v xdes is the expected longitudinal speed, which can be obtained from the speed planning part and will not be discussed in detail here.

[0116] Finally, the driver factor in driving risk analysis can be expressed as:

[0117] E d =ω dc E dc +ω dhp E dhp ;

[0118] Among them, E d Represents the driver factor in driving risk analysis; ω dc and ω dhp is the weight coefficient; E dc is the evaluation index of driver cognitive status; E dhp It is an evaluation index of the driver's control ability.

[0119] 2) Vehicles

[0120] The control and stability of the vehicle reflect whether the vehicle is traveling safely and stably along the expected path.

[0121] First, the ratio of the actual lateral offset to the maximum allowable lateral offset and the ratio of the actual longitudinal speed to the expected longitudinal speed are used as evaluation indicators of the vehicle's lateral and longitudinal control effects. The calculation formula is:

[0122]

[0123] Among them, E ve It is the evaluation index of the vehicle's lateral and longitudinal control effect; y is the actual lateral offset of the vehicle; e ymax is the maximum lateral displacement allowed, the specific value can be determined according to actual requirements; v x is the actual longitudinal velocity of the vehicle; v xdes is the expected longitudinal speed, which can be obtained from the speed planning part and will not be discussed in detail here. In addition, the vehicle stability is reflected by the ratio of the actual lateral acceleration of the vehicle to the maximum allowed lateral acceleration, the actual lateral speed to the expected lateral speed and the standard deviation of the center of mass sideslip angle. The calculation expression is:

[0124]

[0125] Among them, E vs is the evaluation index of vehicle stability; a y is the actual lateral acceleration of the vehicle; a ymax is the maximum lateral acceleration allowed; v y is the actual lateral velocity of the vehicle; v ydes is the expected lateral velocity; σ(β) is the standard deviation of the center-of-mass sideslip angle; β is the center-of-mass sideslip angle.

[0126] Finally, the vehicle factor in driving risk analysis can be expressed as:

[0127] E v =ω ve E ve +ω vs E vs ;

[0128] Among them, E v Represents the vehicle factor in driving risk analysis; ω ve and ω vs is the weight coefficient; E ve It is the evaluation index of the vehicle's lateral and longitudinal control effect; E vs It is an evaluation index of vehicle stability.

[0129] 3) Road

[0130] The driving risk field is used to characterize the impact of the driving environment (road aspect) on driving safety. The present invention analyzes from two perspectives, horizontal and vertical, and calculates the horizontal and vertical risk field forces by considering the distance from the vehicle position to the road boundary, the road speed limit and the collision time. The expression is:

[0131]

[0132] F l =ω lv F lv +ω lttc F lttc ;

[0133] Among them, F c represents the lateral risk field force; k1 is a constant coefficient; k2 is the lateral slope adjustment coefficient; m represents the vehicle mass; v represents the vehicle speed; y d is the distance from the center of gravity of the vehicle to the road boundary; r0 represents the driver's risk focus; r1 represents the impact range of the road boundary; D represents the lane width; F lv represents the longitudinal risk field force considering the road speed limit; v l,min and v l,nax Respectively represent the minimum and maximum speeds allowed on the road; k x and k y Respectively represent the longitudinal and lateral gradient adjustment coefficients of vehicle speed; x p and p Respectively represent the longitudinal and lateral distances between the vehicle and other traffic participants; F lttc represents the longitudinal risk field force considering the collision time ttc; L represents the relative distance; ttc represents the time to collision; F l Represents the vertical risk field force; ω lv and ω lttc are all weight coefficients.

[0134] Based on the evaluation of the horizontal and vertical risk field forces, we can obtain a vertical and horizontal integrated risk assessment model; the horizontal risk field force F c and the longitudinal risk field force F lAfter normalization, the standardized driving risk field force F can be obtained. q :

[0135]

[0136] Among them, F q represents the standardized driving risk field force; E rf Represents road factors in driving safety analysis.

[0137] In summary, when E is satisfied d ≥ε d 、E v ≥ε v and E rf ≥ε rf When any one or both of the following conditions are met, the machine will remind the driver to pay attention to safety, but the driver will be kept driving alone; d , ε v and ε rf They are respectively the thresholds for driver factors, vehicle factors and road factors in driving risk analysis. The specific values ​​can be designed according to actual requirements.

[0138] When E d ≥ε d And E v ≥ε v And E rf ≥ε rf , it indicates that the driving risk exceeds the risk threshold at this time. The machine will take over the driving rights of the vehicle, and the vehicle will be controlled by human-machine collaboration. When the driving risk is reduced, the driving rights of the vehicle will be returned to the driver.

[0139] S2. Based on the unknown information of the driver's driving style, the profit function under the condition of incomplete information is constructed according to the driving risk analysis and the driving performance of the human and the machine:

[0140] In the process of human-machine collaborative control, the machine is not sure of the driver's driving style at the beginning of driving, and does not know his benefit function, so it is difficult to provide appropriate driving assistance. Therefore, based on the driver's driving style as unknown information, a benefit function under incomplete information conditions is constructed according to driving risk analysis and the driving performance of both humans and machines.

[0141] Assume that the driver's driving style is divided into aggressive, normal and weak types, and the machine's driving style is normal; the driver knows his own driving style and the machine's driving style, but the machine is not sure of the driver's driving style.

[0142] First, a driving performance benefit function is constructed. Based on the above assumptions about human-machine driving styles, a driving performance benefit function is constructed.

[0143] Drivers with different driving styles have different driving performances under different degrees of driving authority. For aggressive drivers, they pursue their own control over the vehicle and hope to have higher driving authority, so aggressive drivers will get higher driving performance benefits under medium and high driving authority. For normal drivers, their attitude towards machine assistance is moderate. Although they hope that the machine can provide driving assistance, they do not want the machine to intervene too much. Therefore, normal drivers will get higher driving performance benefits under medium driving authority. For weak drivers, they lack driving experience, do not trust their own driving skills, and hope that the machine can provide more driving assistance. Therefore, weak drivers will get higher driving performance benefits under low driving authority.

[0144] According to the above strategy, the driver's driving performance benefit function expression is as follows:

[0145] E dpi = sin{η1×arctan[DS i ×η2q i -η3(DS i ×η2q i -arctan(DS i ×η2q i ))]};

[0146] Among them, E dpi is the driver’s driving performance benefit function, where i=1 represents aggressive type, i=2 represents normal type, and i=3 represents weak type. i is a coefficient related to the driver's driving style, where i = 1, 2, 3. i is the driving authority of drivers with different driving styles, where i = 1, 2, 3. η1, η2 and η3 are all tuning parameters. The previous article assumes that the machine's driving style is normal, similar to a normal driver. Although it wants to provide driving assistance to the driver, it does not want excessive intervention. Therefore, the normal machine has a higher driving performance benefit under medium driving authority.

[0147] According to the above strategy, the driving performance benefit function of the machine is expressed as follows:

[0148] E mp = sin{η1×arctan[η2(1-q i )-η3(η2(1-q i )-arctan(η2(1-q i )))]};

[0149] Among them, E mp is the driving performance benefit function of the machine. iis the driving authority of different types of drivers, where i=1 represents aggressive type, i=2 represents normal type, and i=3 represents weak type. η1, η2 and η3 are all tuning parameters.

[0150] Then, construct the risk field benefit function. Based on the driving risk field force in the driving risk analysis in step 1, the driver's driving style is considered to determine the benefit function of the driver and the machine. As the risk field force increases, the driving risk will gradually increase; it is generally difficult for drivers to handle emergencies in high-risk scenarios, and at this time the driver hopes that the machine can provide more driving assistance. Therefore, as the risk field force increases, the driver's benefit will gradually decrease, while the machine's benefit will gradually increase.

[0151] For drivers with different driving styles, as the risk field force increases, their risk field benefits decrease at different speeds: aggressive drivers face increased risk field force and dare to drive the vehicle to pursue psychological stimulation, and their benefits decrease at the slowest speed; weak drivers face increased risk field force and their inner panic and fear increase, and their benefits decrease at the fastest speed; the normal driver's benefit decrease speed is between the first two; the specific expression is as follows:

[0152]

[0153] in, is the risk-field benefit function of the driver, where i = 1 represents the aggressive type, i = 2 represents the normal type, and i = 3 represents the weak type; β1, β2, β3, and β4 are all tuning parameters; S i is the risk-field return parameter associated with driving style, where i = 1, 2, 3; F q It is a standardized driving risk field force.

[0154] For a machine, its risk field return is not only related to the risk field force, but also to the driver’s driving style. The specific function expression is:

[0155] E Fqm =1-(1+exp((β1+β2(S i -β3))F q -β4)) -1 i=1,2,3;

[0156] Among them, E Fqm is the risk field return function of the machine; β1, β2, β3 and β4 are all tuning parameters; S i is the risk-field return parameter associated with driving style, where i = 1, 2, 3; F q is the standardized driving risk field force. Finally, based on the unknown information of the driver's driving style, the benefit function under incomplete information is constructed according to the driving risk analysis described in step 1 and the driving performance benefit function described in step 1.

[0157] The driver benefit function includes driving risk benefits (driver status and control ability, vehicle safety, and driving risk field force) and the driver's driving performance benefits. The driver benefit function expressions for different driving styles are as follows:

[0158] Aggressive: J d1 =ω d1 E d +ω v1 E v +ω Fqd1 E Fqd1 +ω dp1 E dp1 ;

[0159] Normal type: J d2 =ω d2 E d +ω v2 E v +ω Fqd2 E Fqd2 +ω dp2 E dp2 ;

[0160] Weak type: J d3 =ω d3 E d +ω v3 E v +ω Fqd3 E Fqd3 +ω dp3 E dp3 ;

[0161] Among them, J d1 is the profit function of the aggressive driver; ω d1 ,ω v1 ,ω Fqd1 and ω dp1 is the weight coefficient of the aggressive driver's benefit function; E d E is the driver factor in driving risk analysis; v E is the vehicle factor in driving risk analysis; Fqd1 is the risk field reward function of the aggressive driver; E dp1 is the driving performance benefit function of aggressive drivers; J d2 is the profit function of a normal driver; ω d2 ,ω v2 ,ω Fqd2 and ω dp2 is the weight coefficient of the normal driver's benefit function; E Fqd2 is the risk field benefit function of a normal driver; E dp2 is the driving performance benefit function of a normal driver; Jd3 is the profit function of the weak driver; ω d3 ,ω v3 ,ω Fqd3 and ω dp3 is the weight coefficient of the weak driver's benefit function; E Fqd3 is the risk field reward function of the weak driver; E dp3 is the driving performance benefit function of a weak driver.

[0162] For the machine, it does not know the driver's driving style, so for drivers with different driving styles, the machine has a corresponding benefit function. The machine benefit function includes driving risk benefit (driver status and control ability, vehicle safety and driving risk field force), the machine's driving performance benefit and the driving style weight coefficient p. The specific expression is as follows:

[0163] ① When the driver's driving style is aggressive, the machine profit function J m1 :

[0164] J m1 =ω dm1 E d +ω vm1 E v +ω Fqm1 E Fqm +ω mp1 E mp ;

[0165] Among them, J m1 is the profit function of the machine when the driver's driving style is aggressive; ω dm1 ,ω vm1 ,ω Fqm1 and ω mp1 is the weight coefficient; E d E is the driver factor in driving risk analysis; v E is the vehicle factor in driving risk analysis; Fqm is the risk field return function of the machine; E mp is the driving performance benefit function of the machine.

[0166] ② When the driver's driving style is normal, the machine profit function J m2 :

[0167] J m2 =ω dm2 E d +ω vm2 E v +ω Fqm2 E Fqm +ω mp2 E mp ;

[0168] Among them, J m2 is the revenue function of the machine when the driver’s driving style is normal; ω dm2 ,ω vm2 ,ω Fqm2 and ω mp2 is the weight coefficient.

[0169] ③ When the driver's driving style is weak, the machine's profit function J m3 :

[0170] J m3 =ω dm3 E d +ω vm3 E v +ω Fqm3 E Fqm +ω mp3 E mp ;

[0171] Among them, J m3 is the revenue function of the machine when the driver’s driving style is weak; ω dm3 ,ω vm3 ,ω Fqm3 and ω mp3 is the weight coefficient.

[0172] In the process of human-machine collaborative control, the machine is not sure of the driver's driving style in the early stage of driving, and its benefit function is unknown, so it is difficult to provide appropriate driving assistance. Therefore, this paper introduces the driving style weight coefficient p to determine the machine's benefit function under incomplete information conditions, and its expression is as follows:

[0173] J m =p i＝1 J m1 +p i＝2 J m2 +p i＝3 J m3 ;

[0174] Among them, J m is the profit function of the machine under incomplete information conditions; p i＝1 、p i＝2 and p i＝3 are driving style weight coefficients respectively.

[0175] The present invention determines the weight coefficients p of different driving styles through double-layer fuzzy logic control. i＝1 、p i＝2 and p i＝3 , the specific steps are as follows:

[0176] First, determine the weight coefficient p of the driver's aggressive driving style i＝1This paper determines the radical weight coefficient p from two perspectives: horizontal and vertical i＝1 , where the lateral factor is the degree of human-machine conflict related to the front wheel turning angle δ con , the longitudinal factor is the impact degree A related to the longitudinal acceleration jerk , but p i＝1 , δ con and A jerk The relationship between the three is difficult to express through mathematical expressions. Therefore, this paper uses fuzzy logic to establish the relationship between the three. con and A jerk is the fuzzy logic input, p i＝1 is the fuzzy logic output; the value range of these variables is [0,1], and the relevant fuzzy subsets are divided into {S, MS, M, MB, B} from small to large, and the specific fuzzy rules are shown in Table 1; in addition, the relevant fuzzy membership functions are as follows Figure 3 (a)-(c) show that the radical weight coefficient p i＝1 The three-dimensional image of Figure 3 (d) as shown.

[0177] Table 1 The first layer fuzzy output p i＝1 Fuzzy rules

[0178]

[0179] Then determine the weight coefficient p of the driver's normal driving style i＝2 This paper uses the vehicle stability index E vs and the first layer fuzzy logic output radical weight coefficient p i＝1 To determine the weight coefficient p of the normal driving style i＝2 , and use fuzzy logic to establish the relationship between the three. Set E vs and p i＝1 For fuzzy logic input, the value range of the variable is [0,1], p i＝2 is the fuzzy logic output, and its value range is [0,1-p i＝1 ]; and the relevant fuzzy subsets are divided into {S, MS, M, MB, B,} from small to large, and the specific fuzzy rules are shown in Table 2; in addition, the relevant fuzzy membership functions are as follows Figure 4 (a)-(c) show; the normal weight coefficient p i＝2 The three-dimensional image of Figure 4 (d) as shown.

[0180] Table 2 The second layer fuzzy output p i＝2 Fuzzy rules

[0181]

[0182] Finally, the weight coefficient p of the driver's driving style is determined to be weak i＝3 According to the determined p i＝1 and p i＝2 The value of p i＝1 +p i＝2 +p i＝3 =1 to determine the weight coefficient p of the weak driving style i＝3 .

[0183] S3. Based on the distribution of benefits between humans and machines, game theory is used to describe the interaction between humans and machines, and the driving authority function under the equilibrium condition of human-machine game is constructed through Nash product:

[0184] During driving, both the driver and the machine seek to maximize their own profit functions, which inevitably leads to conflicts of interest between humans and machines. Considering the distribution of profits between humans and machines, the present invention adopts game theory to describe the interaction between humans and machines; the driver and the machine are regarded as two game players, and both parties continue to bargain according to their own profit functions until they reach an agreement to maximize their own profits.

[0185] The process of constructing the driving authority function includes the following steps:

[0186] (1) According to the definition of Nash product, the driver's profit function and the machine's profit function are taken as the multipliers of the Nash product, and the expression is as follows:

[0187] (J di -J dih )●(J m -J mh );

[0188] Among them, J di is the driver’s benefit function, where i=1 represents the aggressive type, i=2 represents the normal type, and i=3 represents the weak type; J dih is the driver's gain at the negotiation disagreement point; J m is the revenue function of the machine; J mh for the machine’s gains at the divergence point in the negotiation;

[0189] (2) When the human-machine game reaches a game equilibrium, the driver’s driving authority q i The product W of the driver's profit function and the machine's profit function satisfies the following expression:

[0190] W = argmax[J di (q i )-J dih ]·[J m (q i )-J mh ];

[0191] Among them, the product W is about driving authority q i Function of J di (q i ) is when the driving authority is q i The driver's benefit when i=1 is the aggressive type, i=2 is the normal type, and i=3 is the weak type; q i is the driver's driving authority, where i = 1, 2, 3; J m (q i ) is when the driving authority is q i The machine revenue at that time;

[0192] (3) Convert the product W into: W = F(q i );

[0193] Among them, F(q i )=argmax[J di (q i )-J dih ]●[J m (q i )-J mh ];q i is the driver’s driving authority, where i=1,2,3.

[0194] Therefore, the driving authority q can be obtained i =F -1 (W);

[0195] S4. Solve the driving authority function to obtain driving authority to achieve human-machine collaborative control:

[0196] The driving authority value range is 0-1, which is theoretically continuous and infinite. In order to simplify the calculation, the present invention divides the driving authority into 1000 parts, that is, the step size is 0.001; this converts the infinite strategy solution problem into a finite strategy solution problem. When calculating, each decision only needs to calculate 1000 groups. In addition, there may be multiple solution results, and the present invention selects the first solution as the final result.

Claims

1. A method for allocating driving rights for human-machine co-driving based on incomplete information game, characterized in that: The steps include: S1. Carry out driving risk analysis based on the three factors of "people-vehicle-road", wherein the driving risk analysis includes driver factors, vehicle factors and road factors, wherein the driver factors include the driver's cognitive state and the driver's control ability; the vehicle factors include the lateral and longitudinal control effects of the vehicle and the stability of the vehicle; and the road factors include the lateral and longitudinal driving risk field forces; S2. Based on the unknown information of the driver's driving style, a benefit function under incomplete information conditions is constructed according to the driving risk analysis and the driving performance of the human and the machine. The benefit function includes a driver benefit function and a machine benefit function. The driver benefit function includes a driver driving risk benefit function and a driver driving performance benefit function; the machine benefit function includes a machine driving risk benefit function and a machine driving performance benefit function, wherein: The driver's driving risk-benefit function includes the driver's cognitive state, the driver's control ability, the vehicle's lateral and longitudinal control effects, the vehicle's stability, and the driving risk field force; The machine driving risk-benefit function is constructed by taking the driver's driving style as unknown information and introducing a driving style weight coefficient. The machine driving risk-benefit function includes the driver's cognitive state, the driver's control ability, the vehicle's lateral and longitudinal control effects, the vehicle's stability, the driving risk field force and the driving style weight coefficient; S3. Based on the distribution of benefits between humans and machines, game theory is used to describe the interaction between humans and machines, and the driving authority function under the equilibrium condition of human-machine game is constructed through Nash product; S4. Solve the driving authority function to obtain driving authority to achieve human-machine collaborative control.

2. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 1 is characterized in that: The evaluation index E of the driver factor d is the evaluation index E of the driver's cognitive state dc and the evaluation index E of the driver's control ability dhp Evaluated, among which, The evaluation index E of the driver's cognitive state dc The calculation formula is: Among them, E dc is the evaluation index of the driver's cognitive state; σ1, σ2 and σ3 are all tuning parameters; δ d is the front wheel turning angle of the driver; At is the driver's attention level, which is calculated by collecting driver-related information from external equipment; The evaluation index E of the driver's control ability dhp The calculation formula is: Among them, E dhp It is the evaluation index of the driver's control ability; d and δ f are the driver's front wheel steering angle and the vehicle's actual front wheel steering angle respectively; y is the actual lateral offset of the vehicle; e ymax is the maximum lateral displacement allowed, the specific value can be determined according to actual requirements; v x is the actual longitudinal velocity of the vehicle; v xdes is the expected longitudinal speed, which is obtained from the speed planning part; Then, the evaluation index E of the driver factor is d It is expressed as: E d =ω dc E dc +oh dhp E dhp ; Among them, E d Represents the driver factor in driving risk analysis; ω dc and ω dhp is the weight coefficient; E dc is the evaluation index of the driver's cognitive state; E dhp It is an evaluation index of the driver's control ability.

3. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 2 is characterized in that: The evaluation index E of the vehicle factor v It is the evaluation index E of the vehicle's lateral and longitudinal control effect. ve and vehicle stability evaluation index E vs Evaluated, among which, The evaluation index E of the horizontal and vertical control effect is ve The calculation formula is: Among them, E ve It is the evaluation index of the vehicle's lateral and longitudinal control effect; y is the actual lateral offset of the vehicle; e ymax is the maximum lateral offset allowed, and the specific value is determined according to actual requirements; v x is the actual longitudinal velocity of the vehicle; v xdes is the expected longitudinal speed, which is obtained from the speed planning part; The evaluation index E of the vehicle stability vs The calculation expression is: Among them, E vs is the evaluation index of vehicle stability; a y is the actual lateral acceleration of the vehicle; a ymax is the maximum lateral acceleration allowed; v y is the actual lateral velocity of the vehicle; v ydes is the expected lateral velocity; σ(β) is the standard deviation of the center of mass sideslip angle; β is the center of mass sideslip angle; Then, the evaluation index E of the vehicle factor is v It is expressed as: E v =ω ve E ve +oh vs E vs ; Among them, E v Represents the vehicle factor in driving risk analysis; ω ve and ω vs is the weight coefficient; E ve It is the evaluation index of the vehicle's lateral and longitudinal control effect; E vs It is an evaluation index of vehicle stability.

4. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 3 is characterized in that: The road factor E rf The formation process includes the following steps: S41. The transverse and longitudinal risk field forces are calculated based on the distance from the vehicle position to the road boundary, the road speed limit and the collision time. The expression is: F l =ω lv F lv +oh lttc F lttc ; Among them, F c represents the lateral risk field force; k1 is a constant coefficient; k2 is the lateral slope adjustment coefficient; m represents the vehicle mass; v represents the vehicle speed; y d is the distance from the center of gravity of the vehicle to the road boundary; r0 represents the driver's risk focus; r1 represents the impact range of the road boundary; D represents the lane width; F lv represents the longitudinal risk field force considering the road speed limit; v l,min and v l,nax Respectively represent the minimum and maximum speeds allowed on the road; k x and k y Respectively represent the longitudinal and lateral gradient adjustment coefficients of vehicle speed; x p and p Respectively represent the longitudinal and lateral distances between the vehicle and other traffic participants; F lttc represents the longitudinal risk field force considering the collision time ttc; L represents the relative distance; ttc represents the collision time (Time To Collision); F l Represents the vertical risk field force; ω lv and ω lttc All are weight coefficients; S42, evaluating the horizontal and vertical risk field forces to obtain a vertical and horizontal integrated risk assessment model; S43, the horizontal risk field force F c and the longitudinal risk field force F l After normalization, the standardized driving risk field force F is obtained. q : Among them, F q represents the standardized driving risk field force; E rf Represents the road factors in driving risk analysis.

5. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 4 is characterized by: When E is satisfied d ≥ε d 、E v ≥ε v and E rf ≥ε rf When any one or both of the following conditions are met, the machine will remind the driver to pay attention to safety, but the driver will be kept driving alone; d , ε v and ε rf They are the thresholds of driver factors, vehicle factors and road factors in driving risk analysis respectively.

6. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 5 is characterized by: When E d ≥ε d And E v ≥ε v And E rf ≥ε rf , indicating that the driving risk exceeds the risk threshold, the machine will take over the driving rights of the vehicle, and the vehicle will be controlled collaboratively by man and machine; when the driving risk is reduced, the driving rights of the vehicle will be returned to the driver.

7. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 4 is characterized in that: The process of establishing the driver benefit function and the machine benefit function under the incomplete information condition includes the following steps: S71, assuming the driver's driving style to be aggressive, normal, and weak, and assuming the machine's driving style to be normal; S72. Based on the assumptions in S71, construct a driver driving performance benefit function and a machine driving performance benefit function, wherein: The expression of the driver's driving performance benefit function is as follows: E dpi =sin{η1×arctan[DS i ×η2q i -η3(DS i ×η2q i -arctan(DS i ×η2q i ))]}; Among them, E dpi is the driver's driving performance benefit function, where i=1 represents aggressive driving, i=2 represents normal driving, and i=3 represents weak driving; DS i is a coefficient related to the driver's driving style, where i = 1, 2, 3; q i is the driving authority of drivers with different driving styles, where i = 1, 2, 3; η1, η2 and η3 are all tuning parameters; The machine driving performance benefit function expression is as follows: E mp =sin{η1×arctan[η2(1-q i )-η3(η2(1-q i )-arctan(η2(1-q i )))]}; Among them, E mp is the machine driving performance benefit function; q i is the driving authority of different types of drivers, where i = 1, 2, 3; η1, η2 and η3 are all tuning parameters; S73, constructing a driver driving risk-benefit function and a machine driving risk-benefit function based on the assumptions in S71 and the lateral and longitudinal driving risk field forces in the driving risk analysis, wherein: The expression of the driver's driving risk-benefit function is as follows: in, is the driver's driving risk-return function, where i = 1 represents the aggressive type, i = 2 represents the normal type, and i = 3 represents the weak type; β1, β2, β3, and β4 are all tuning parameters; S i is the risk-field return parameter associated with driving style, where i = 1, 2, 3; F q To standardize the driving risk field force; The expression of the machine driving risk-return function is: in, is the risk-return function of machine driving; β1, β2, β3 and β4 are all tuning parameters; S i is the driving risk field benefit parameter related to driving style, where i = 1, 2, 3; F q To standardize the driving risk field force; S74: construct a driver benefit function and a machine benefit function according to the driving risk analysis and the driver driving performance benefit function and the machine driving performance benefit function, wherein: The specific expression of the driver benefit function is as follows: Aggressive: Normal type: Weak type: Among them, J d1 is the aggressive driver benefit function; ω d1 ,ω v1 , and ω dp1 is the weight coefficient of the aggressive driver benefit function; E d is the driver factor in driving risk analysis; E v Vehicle factors in driving risk analysis; is the risk-benefit function of an aggressive driver; E dp1 is the driving performance benefit function of the aggressive driver; J d2 is the normal driver benefit function; ω d2 ,ω v2 , and ω dp2 is the weight coefficient of the normal driver benefit function; is the risk-benefit function of a normal driver; E dp2 is the driving performance benefit function of a normal driver; J d3 is the weak driver benefit function; ω d3 ,ω v3 , and ω dp3 is the weight coefficient of the weak driver benefit function; is the risk-benefit function of the weak driver; E dp3 The driving performance benefit function for weak drivers; The expression of the machine profit function is as follows: Radical Normal type, Weak type, Among them, J m1 is the machine profit function when the driver's driving style is aggressive; ω dm1 ,ω vm1 , and ω mp1 is the weight coefficient; is the risk-return function of the machine; E mp is the machine driving performance benefit function; J m2 is the revenue function of the machine when the driver’s driving style is normal; ω dm2 ,ω vm2 , and ω mp2 is the weight coefficient; J m3 is the revenue function of the machine when the driver’s driving style is weak; ω dm3 ,ω vm3 , and ω mp3 is the weight coefficient; S75. Construct a profit function of the machine under incomplete information conditions according to the driver profit function and the machine profit function, and the expression thereof is as follows: J m =p i＝1 J m1 +p i＝2 J m2 +p i＝3 J m3 ; Among them, J m is the profit function of the machine under incomplete information conditions; p i＝1 、p i＝2 and p i＝3 Both are driving style weight coefficients.

8. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 7 is characterized in that: Determine the weight coefficient p of different driving styles by double-layer fuzzy logic control i =1, p i＝2 and p i＝3 , the specific steps are as follows: S81, determining the weight coefficient p of the driver's driving style as an aggressive type i＝1 Determine the aggressive weight coefficient p from both horizontal and vertical perspectives i＝1 , where the lateral factor is the degree of human-machine conflict related to the front wheel turning angle δ con , the longitudinal factor is the impact degree A related to the longitudinal acceleration jerk ; Using the first layer of fuzzy logic rules to establish p i＝1 ,δ con and A jerk The connection between the three is set con and A jerk is the input of the first layer of fuzzy logic, p i＝1 is the output of the first layer of fuzzy logic; where p i＝1 ,δ con and A jerk The value range of the three is [0,1], and the relevant fuzzy subsets are divided into {S, MS, M, MB, B} from small to large; S82, determining the weight coefficient p of the driver's driving style as normal i＝2 According to the evaluation index E of vehicle stability vs and the first layer fuzzy logic output radical weight coefficient p i＝1 To determine the weight coefficient p of the normal driving style i＝2 , and use the second layer of fuzzy logic rules to establish E vs 、p i＝1 and p i＝2 The connection between the three is set vs and p i＝1 is the input of the second layer of fuzzy logic. The value range of the variables is [0,1]. i＝2 is the output of the second layer of fuzzy logic, and its value range is [0,1-p i＝1 ]; Among them, the fuzzy subsets are divided into {S, MS, M, MB, B} from small to large; S83, determining the weight coefficient p of the driver's driving style as a weak type i＝3 According to the determined p i＝1 and p i＝2 The value of p i＝1 +p i＝2 +p i＝3 =1 to determine the weight coefficient p of the weak driving style i＝3 .

9. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 8 is characterized in that: The first layer of fuzzy logic rules: The second layer of fuzzy logic rules:

10. The method for allocating human-machine co-driving rights based on incomplete information game according to claim 7, characterized in that: The construction process of the driving authority function includes the following steps: S101. According to the definition of Nash product, the driver's profit function and the machine's profit function are taken as multipliers of the Nash product, and the expression is as follows: (J di -J dih )·(J m -J mh ); Among them, J di is the driver’s benefit function, where i=1 represents the aggressive type, i=2 represents the normal type, and i=3 represents the weak type; J dih is the driver's gain at the negotiation disagreement point; J m is the revenue function of the machine; J mh for the machine’s gains at the divergence point in the negotiation; S102. When the human-machine game reaches a game equilibrium, the driver's driving authority q i The product W of the driver's profit function and the machine's profit function satisfies the following expression: W=argmax[J di (q i )-J dih ]·[J m (q i )-J mh ]; Among them, the product W is about driving authority q i Function of J di (q i ) is when the driving authority is q i The driver's benefit when i=1 is the aggressive type, i=2 is the normal type, and i=3 is the weak type; q i is the driver's driving authority, where i = 1, 2, 3; J m (q i ) is when the driving authority is q i The machine revenue at that time; S103, convert the product W into an invisible function: W = F(q i ); Among them, F(q i )=argmax[J di (q i )-J dih ]·[J m (q i )-J mh ];q i is the driver’s driving authority, where i = 1, 2, 3; Therefore, the driving authority q i Expressed as the inverse form of the implicit function: q i =F -1 (W).

Images