Evaluation method for interactive cognitive ability of unmanned vehicle

Abstract

The invention provides a method for evaluating the interactive cognitive ability of a driverless vehicle. The method comprises the following steps: constructing a hierarchical interactive scene library and collecting high-frequency data; constructing a cognitive entropy model; carrying out interactive cognitive calculation based on entropy reduction; and constructing a comprehensive scoring model, and carrying out comprehensive scoring and grading. According to the method for evaluating the interactive cognitive ability of the driverless vehicle, a clear optimization direction is provided for algorithm development, and an innovative implementation path is provided for evaluating the core ability of the driverless vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned driving, and particularly to a method for evaluating interactive cognitive ability of an unmanned vehicle. BACKGROUND

[0002] Unmanned driving and autonomous driving describe the external form, while autonomous driving and self-driving are two completely different technical routes. Autonomous driving is essentially teaching a robot to drive through a program by installing various sensors, computer systems and related software, starting a computer program, and letting the car drive automatically. It is a software-defined car. Self-driving necessarily has a subject of driving cognition. As an intelligent agent of the driver, self-driving comprehensively utilizes multiple sensors distributed in various parts of the vehicle body to perceive, plan, decide and control the behavior of the vehicle. Self-driving represented by Google Waymo uses multiple radars for near, medium and long distance detection, uses multiple cameras to perceive the surrounding environment, uses deep learning methods to identify traffic signals, and front vehicles or pedestrians and other moving obstacles, and focuses on intelligent perception and ultimately controls the behavior of the vehicle through planning and decision-making. In addition to solving the problems of vehicle perception intelligence, information fusion, planning, decision-making, control and other issues, self-driving also needs to interact with the environment and surrounding vehicles.

[0003] Currently, there is a lack of interactive cognition in unmanned driving. Unmanned vehicles need to autonomously respond to the various uncertainties that often occur during driving, which not only requires vehicle dynamics and various sensors, but also simulates the autonomous prediction and control of the driver's cognitive ability and interacts with the external environment. Therefore, unmanned vehicles need to have sufficient interactive cognitive ability to be accepted by the public.

[0004] Traditional ability evaluation of unmanned driving is mostly focused on the accuracy of perception, stability of control, safety and driving range, etc. Interactive cognitive ability is more inclined to soft intelligence, which describes the ability of vehicles to understand, predict, communicate and make decisions in complex and dynamic social traffic environments, like humans. Traditional performance representation methods of unmanned driving cannot effectively measure the cognitive and interactive ability of vehicles in complex, dynamic and human-machine integrated environments. Therefore, how to effectively represent the interactive cognitive ability of unmanned vehicles has become an important research problem, and there is currently very little research on this.

[0005] A method for evaluating the interactive cognitive ability of an unmanned vehicle is disclosed in Chinese patent application publication CN118966883A, which includes determining a set of dimensions of the interactive cognitive ability of the unmanned vehicle, and further includes the following steps: determining a set of evaluation objects of the interactive cognitive ability of the unmanned vehicle; and calculating a comprehensive evaluation score C of the interactive cognitive ability of the unmanned vehicle. The method has the disadvantage that the indicators involved in the evaluation still focus on hard indicators in the aspects of perception and control, and the process of interactive cognition is not clearly described, so the interactive cognition is not successfully modeled, and the evaluation effect of the interactive cognitive ability of the unmanned vehicle is weak. SUMMARY

[0006] To solve the above technical problems, the present application provides a method for evaluating the interactive cognitive ability of an unmanned vehicle, which converts the vague interactive ability into a specific model, so that the evaluation result is objective, comparable, and can be traced back to the specific link that is not handled well.

[0007] The present application provides a method for evaluating the interactive cognitive ability of an unmanned vehicle, which includes constructing a hierarchical interactive scene library and collecting data, and further includes the following steps:

[0008] Step 1: constructing a cognitive entropy model;

[0009] Step 2: performing interactive cognitive calculation based on entropy reduction;

[0010] Step 3: constructing a comprehensive scoring model and performing comprehensive scoring and grade division.

[0011] Preferably, the hierarchical interactive scene library includes at least one of a basic interactive scene, a conflict interactive scene, a cooperative interactive scene, and a socialized interactive scene.

[0012] In any of the above solutions, preferably, the data includes at least one of self-vehicle data, environment data, environment semantic data, and awareness data of traffic participants,

[0013] The self-vehicle data includes position, speed, acceleration, heading angle, turn signal state, and planned path;

[0014] The environment data includes the position, speed, acceleration, heading angle, and orientation of all relevant traffic participants;

[0015] The environment semantic data includes communication signal light state, traffic signs, lane lines, and road markings;

[0016] The awareness data of the traffic participants includes the predicted trajectory, intention recognition probability, and decision confidence of each traffic participant.

[0017] In any of the above solutions, it is preferred that the cognitive entropy model comprises intention probability calculation, individual information entropy and global cognitive entropy.

[0018] In any of the above solutions, it is preferred that the step 1 comprises probabilistic modeling of the intention of a single traffic participant i within a system perception range for a future period of time, resulting in A i = {P i (a, t) | a is all possible behaviors}, where A i is intention probability calculation, t is the current time, P i (a, t) represents the probability of making a behavior a at time t, and Σp i (a, t) = 1.

[0019] In any of the above solutions, it is preferred that the individual information entropy H i (t) is the uncertainty of the intention of the single traffic participant i, and the formula is

[0020] .

[0021] In any of the above solutions, it is preferred that the global cognitive entropy H global (t) is a weighted sum of the individual information entropy of all related traffic participants, and the formula is

[0022]

[0023] w i (t) is the influence degree weight of the traffic participant i on the ego vehicle, and the formula is

[0024]

[0025] where d i (t) is the distance between the ego vehicle and the traffic participant i, TTC i (t) is the collision time between the ego vehicle and the traffic participant i, d0 and TTC0 are normalization constants, and α and β are weight coefficients and α + β = 1.

[0026] In any of the above solutions, it is preferred that the step 2 comprises the following sub-steps:

[0027] Step 21: Calculate the entropy reduction rate;

[0028] Step 22: Calculate the cumulative entropy reduction value.

[0029] In any of the above solutions, it is preferred that the formula for calculating the entropy reduction rate R(t, Δt) is

[0030]

[0031] Where Δt is the time interval between time t and the next sampling time.

[0032] In any of the above schemes, the preferred option is that the cumulative entropy reduction value I ΔH The calculation formula is

[0033]

[0034] Where tend is the end time and tstart is the start time.

[0035] In any of the above solutions, step 3 preferably includes the following sub-steps:

[0036] Step 31: Extract key metrics for the complete test scenario;

[0037] Step 32: Construct a comprehensive scoring model and calculate the comprehensive score S using weighted summation;

[0038] Step 33: Classify the capability level of the autonomous vehicle based on the S score.

[0039] In any of the above solutions, it is preferred that the key indicators include security indicators, efficiency indicators, and interactive cognition indicators.

[0040] The safety indicators include maximum deceleration, minimum TTC, and whether a collision occurs.

[0041] The efficiency metrics include task completion time and average speed;

[0042] The interactive cognitive indicators include average global cognitive entropy, maximum global cognitive entropy, cumulative entropy reduction, and average entropy reduction rate.

[0043] Preferably, in any of the above schemes, the average global cognitive entropy H avg The calculation formula is

[0044] .

[0045] In any of the above schemes, the preferred option is the average entropy reduction rate R. avg The calculation formula is

[0046] .

[0047] In any of the above schemes, the preferred method is that the formula for calculating the comprehensive score S is:

[0048] S=W s ·S s +W e ·S e +W i ·S i

[0049] Among them, S s S e S i The normalized scores for security, efficiency, and interaction cognition are respectively given by W. s W e W i Weights are assigned to three dimensions: security, efficiency, and interactive cognition, and W... s +W e +W i =1.

[0050] This invention proposes a method for evaluating the interactive cognitive ability of autonomous vehicles, thereby quantifying the ability of autonomous vehicles to proactively reduce environmental uncertainties. Attached Figure Description

[0051] Figure 1 This is a flowchart of a preferred embodiment of the evaluation method for the interactive cognitive ability of an autonomous vehicle according to the present invention. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0053] Example 1

[0054] like Figure 1 As shown, a method for evaluating the interactive cognitive ability of an autonomous vehicle includes step 100, which involves constructing a hierarchical interactive scenario library and collecting high-frequency data. The hierarchical interactive scenario library includes at least one scenario among basic interactive scenarios, conflict interactive scenarios, collaborative interactive scenarios, and socialized interactive scenarios.

[0055] The high-frequency data includes at least one of the following: vehicle data, environmental data, environmental semantic data, and traffic participant awareness data. The vehicle data includes location, speed, acceleration, heading angle, turn signal status, and planned path.

[0056] The environmental data includes the position, speed, acceleration, heading angle, and orientation of all relevant traffic participants;

[0057] The environmental semantic data includes traffic light status, traffic signs, lane lines, and road markings;

[0058] The traffic participant awareness data includes each traffic participant's predicted trajectory, intent recognition probability, and decision confidence level.

[0059] Perform step 110 to construct a cognitive entropy model, which includes intention probability calculation, individual information entropy, and global cognitive entropy.

[0060] For a single traffic participant i within the system's perception range, its intentions over a future period are probabilistically modeled, and the result is A. i ={P i (a,t)|a represents all possible actions}, where A i For calculating the probability of intent, t is the current time, P i (a,t) represents the probability of performing action a at time t, Σp i (a,t)=1.

[0061] The single-entity information entropy H i (t) represents the uncertainty of the intention of a single traffic participant i, and the formula is:

[0062] .

[0063] The global cognitive entropy H global (t) is the weighted sum of the individual information entropies of all relevant traffic participants, and the formula is:

[0064]

[0065] w i (t) represents the weight of the influence of traffic participant i on the vehicle, and the formula is as follows:

[0066]

[0067] Where, d i (t) represents the distance between the vehicle and traffic participant i, TTC i (t) represents the collision time between the vehicle and traffic participant i, d0 and TTC0 are normalization constants, and α and β are weighting coefficients with α+β=1.

[0068] Step 120 involves performing entropy reduction-based interactive cognitive computing, including the following sub-steps:

[0069] Perform step 121 to calculate the entropy reduction rate, wherein the entropy reduction rate R(t, The formula for calculating Δt is:

[0070]

[0071] Where Δt is the time interval between time t and the next sampling time.

[0072] Perform step 122 to calculate the cumulative entropy reduction value, wherein the cumulative entropy reduction value I ΔH The calculation formula is

[0073]

[0074] Where tend is the end time and tstart is the start time.

[0075] Execute step 130 to construct a comprehensive scoring model and perform comprehensive scoring and level classification, including the following sub-steps:

[0076] Perform step 131 to extract key metrics for the complete test scenario. These key metrics include security metrics, efficiency metrics, and interaction cognition metrics.

[0077] The safety indicators include maximum deceleration, minimum TTC, and whether a collision occurs.

[0078] The efficiency metrics include task completion time and average speed;

[0079] The interactive cognitive indicators include average global cognitive entropy, maximum global cognitive entropy, cumulative entropy reduction, and average entropy reduction rate.

[0080] The average global cognitive entropy H avg The calculation formula is

[0081] .

[0082] The average entropy reduction rate R avg The calculation formula is

[0083] .

[0084] Execute step 132 to construct a comprehensive scoring model and calculate the comprehensive score S using weighted summation. The formula for calculating the comprehensive score S is as follows:

[0085] S=W s ·S s +W e ·S e +W i ·S i

[0086] Among them, S s S e S i The normalized scores for security, efficiency, and interaction cognition are respectively given by W. s W e W i Weights are assigned to three dimensions: security, efficiency, and interactive cognition, and W... s +W e +W i =1.

[0087] Perform step 133 to classify the capability level of the autonomous vehicle based on the score of S.

[0088] Example 2

[0089] This patent proposes an innovative evaluation method: an evaluation method for the interactive cognitive ability of autonomous vehicles based on cognitive entropy reduction. The core value of an excellent autonomous vehicle lies in its ability to continuously reduce the uncertainty of future traffic conditions through interaction with the environment, thereby making safer, more efficient, and compliant decisions. The interaction process is viewed as an information game, where the vehicle is not only a participant but also a proactive driver of uncertainty reduction.

[0090] A method for evaluating the interactive cognitive capabilities of autonomous vehicles includes:

[0091] 1. Scene Construction and Data Acquisition

[0092] This is the foundation of the evaluation method, which requires the construction of typical scenarios that can fully stimulate interactive cognitive abilities.

[0093] 1.1 Scene Library Design

[0094] Build a hierarchical library of interactive scenarios to cover interactive needs ranging from simple to complex.

[0095] L1. Basic interaction scenarios: following other vehicles, changing lanes, merging into the main road. These scenarios have clear rules, but the randomness of other vehicles needs to be handled.

[0096] L2. Conflict and Interaction Scenarios: Crossing intersections without traffic lights, merging into roundabouts, and mixed traffic of pedestrians and vehicles. These scenarios involve ambiguous right-of-way and require negotiation and bargaining.

[0097] L3. Collaborative Interaction Scenarios: Emergency vehicle avoidance, multi-vehicle collaborative merging, and "zipper-style" passage through congested areas. These scenarios require vehicles to understand social collaboration rules and actively cooperate.

[0098] L4. Social Interaction Scenarios: Handling vehicles cutting in line, understanding pedestrians' hesitation, and anticipating non-motorized vehicle violations. These scenarios require a high level of social understanding and predictive ability.

[0099] 1.2 Data Acquisition

[0100] In a simulated environment or closed test track, have the autonomous vehicle under test run the above scenario. Collect the following data.

[0101] Vehicle data: position, speed, acceleration, heading angle, turn signal status, and planned path.

[0102] Environmental data: Position, speed, acceleration, heading angle, and orientation of all relevant traffic participants (vehicles, pedestrians, non-motorized vehicles).

[0103] Environmental semantic data: traffic light status, traffic signs, lane lines, and road markings.

[0104] Traffic participant intent data: predicted trajectory, intent recognition probability (lane change probability, braking probability, etc.), and decision confidence for each traffic participant.

[0105] 2. Construction of Cognitive Entropy Model

[0106] Cognitive entropy is defined as a measure of the uncertainty that an autonomous vehicle has about the future state of its surrounding traffic environment.

[0107] 2.1 Intent Probability Calculation

[0108] For a single traffic participant i within the system's perception range, the system needs to probabilistically model its intentions over a future period, resulting in A. i ={P i (a,t)|a represents all possible actions}, t is the current time, P i (a,t) represents the probability of performing action a at time t, Σp i (a,t)=1.

[0109] 2.2 Individual Information Entropy

[0110] For a single traffic participant i, the uncertainty of its intention can be measured by information entropy, which is called the individual information entropy H. i (t), defined as follows:

[0111]

[0112] H i The larger the value of (t), the more uncertain the system is about the future behavior of the participant, and the higher the risk of interaction.

[0113] 2.3 Global Cognitive Entropy

[0114] The global cognitive entropy H of the entire driving environment global (t) is the weighted sum of the individual information entropies of all relevant traffic participants. H global (t) comprehensively reflects the degree of uncertainty of the entire interactive environment faced by the autonomous vehicle at time t. Global cognitive entropy H global The definition of (t) is as follows:

[0115]

[0116] weight w i (t) reflects the degree of influence this participant has on the autonomous vehicle. Weight w i (t) should take into account spatiotemporal factors such as distance, relative velocity, and collision time, and is defined as follows:

[0117]

[0118] d i(t): The distance between the autonomous vehicle and participant i.

[0119] TTC i (t): The collision time between the autonomous vehicle and participant i.

[0120] d0 and TTC0: Normalization constants used to adjust the sensitivity of distance and TTC.

[0121] α, β: weighting coefficients, α+β=1, used to balance the importance of distance and TTC.

[0122] weight w i (t) needs to be normalized so that Σw i (t)=1.

[0123] 3. Interactive cognitive computing based on entropy reduction

[0124] The evaluation system needs to dynamically calculate the changes in entropy during the interaction process. The goal of autonomous vehicles is to proactively take measures during the interaction to improve H... global (t) is developing in the direction of decreasing.

[0125] 3.1 Calculation of Entropy Reduction Rate

[0126] Entropy reduction rate measures the ability of an autonomous vehicle to reduce its global cognitive entropy per unit time through its own decision-making. The formula for calculating the entropy reduction rate is as follows:

[0127]

[0128] R(t, Δt)>0: This indicates that within the time interval [t, t+Δt], the system's behavior successfully reduced the uncertainty of the environment.

[0129] R(t, Δt)<0: This indicates that the system's behavior increases the uncertainty of the environment.

[0130] R(t, The larger the value of Δt, the stronger the system's ability to understand the environment and guide interaction.

[0131] 3.2 Calculation of Cumulative Entropy Reduction

[0132] Cumulative entropy reduction measures the total entropy reduction benefit created by the system throughout the entire interaction scenario. The formula for calculating cumulative entropy reduction is as follows:

[0133]

[0134] tstart and tend are the start and end times, respectively, and I is the net gain of the entire interaction process. A larger I indicates a stronger interactive cognitive ability and a more efficient task completion process. It can be used to compare the performance of different algorithms in completing the same task.

[0135] 4. Overall Scoring and Grading

[0136] By combining the above quantitative indicators with traditional safety and efficiency indicators, a comprehensive evaluation system can be formed.

[0137] 4.1 Key Indicator Extraction

[0138] For a complete test scenario, calculate the following core metrics:

[0139] (1) Safety indicators: maximum deceleration, minimum TTC, whether a collision occurs.

[0140] (2) Efficiency indicators: task completion time and average speed.

[0141] (3) Interactive cognitive indicators:

[0142] Average global cognitive entropy: .

[0143] Maximum global cognitive entropy: H max It reflects the most chaotic moment.

[0144] Cumulative entropy reduction: I reflects the core cognitive ability of interaction.

[0145] Average entropy reduction rate: It reflects the efficiency of entropy reduction.

[0146] 4.2 The comprehensive scoring model uses a weighted summation method to calculate a final comprehensive score S. The comprehensive score is calculated as follows:

[0147] S=W s ·S s +W e ·S e +W i ·S i

[0148] S s S e S i These are normalized scores (0-100 points) for three dimensions: safety, efficiency, and interaction cognition.

[0149] W s W e W i These are the weights for the corresponding dimensions, and W s +W e +Wi =1.

[0150] 4.3 Classification of Interactive Cognitive Ability Levels

[0151] As shown in Table 1, the capability levels of autonomous vehicles are classified according to the S score.

[0152] Table 1. Levels of Interactive Cognitive Ability

[0153] Rank S score range Feature description L4 90 - 100 Can actively guide and optimize traffic flow through its own behavior, significantly reducing cognitive entropy, showing human-like social collaboration ability. L3 75 - 89 Can clearly express intentions in complex conflicts, efficiently negotiate, and achieve significant entropy reduction, making the interaction process smooth and natural. L2 60 - 74 Can correctly respond to environmental changes and avoid entropy increase, but lacks the ability to actively guide and optimize the environment, appearing slightly conservative. L1 40 - 59 Severe dependence on preset rules, poor performance in rule-free scenarios, and easy to increase environmental entropy, making the interaction rigid. L0 0 - 39 Cannot correctly understand the environment, intentions are confused, decisions are contradictory, and are troublemakers in interactions.

[0154] The method of this invention focuses on higher-order intelligence, transforming fuzzy interaction capabilities into concrete models (entropy, entropy reduction rate), which are quantifiable and interpretable, making the evaluation results objective and comparable.

[0155] It focuses not only on the results but also on the process. It is predictive and process-oriented. A system with drastic entropy fluctuations, even if it ultimately becomes safe, indicates that its cognitive processes are unstable and pose potential risks.

[0156] Maximizing cumulative entropy reduction provides a clear optimization objective for algorithm development, promoting algorithm evolution. The evaluation method of this invention will encourage developers to research more advanced intent prediction, game theory decision-making, and human-computer interaction technologies.

[0157] The same evaluation methods can be used to assess human driver data, thereby setting a cognitive entropy reduction benchmark for autonomous vehicles that surpasses human levels, rather than simply mimicking humans.

[0158] The evaluation method of this invention provides an innovative and highly promising implementation path for assessing the core capabilities of autonomous vehicles.

[0159] Example 3

[0160] This embodiment explains the calculation of intent probability.

[0161] For a single traffic participant i within the system's perception range, the system needs to probabilistically model its intentions over a future period, resulting in A. i ={P i (a,t)|a represents all possible actions}, t is the current time, P i (a,t) represents the probability of performing action a at time t, Σp i (a,t)=1.

[0162] For example, for a vehicle i at the intersection ahead, the modeling result of its intent probability is the intent probability calculation A. i ,

[0163] A i ={P i(a,t)|a can take the values ​​of straight, left turn, right turn, and stop}={p i (straight ahead, t), p i (Turn left, t), p i (Turn right, t), p i (parking, t)}

[0164] Where t is the current time, Σp i (a,t)=1.

[0165] Example 4

[0166] This embodiment explains the information entropy of a single entity.

[0167] For a single traffic participant i, the uncertainty of its intention can be measured by information entropy, which is called the individual information entropy H. i (t), defined as follows:

[0168]

[0169] H i The larger the value of (t), the more uncertain the system is about the future behavior of the participant, and the higher the risk of interaction.

[0170] When p i When one probability in (a,t) is 1 and the others are 0 (completely certain intention), H i (t)=0, the entropy is the lowest.

[0171] When all p i When (a,t) are equal (with completely uncertain intentions), H i When (t) reaches its maximum value, the entropy is at its highest.

[0172] H i The larger the value of (t), the more uncertain the system is about the future behavior of the participant, and the higher the risk of interaction.

[0173] Example 5

[0174] This embodiment explains global cognitive entropy.

[0175] The weight w of global cognitive entropy i (t) should take into account spatiotemporal factors such as distance, relative velocity, and collision time, and is defined as follows:

[0176]

[0177] α, β: Weighting coefficients, α+β=1, used to balance the importance of distance and TTC. For example, setting α=0.4 and β=0.6 emphasizes time urgency.

[0178] Example 6

[0179] This example explains the calculation of the comprehensive score.

[0180] The overall score is calculated using a weighted summation method:

[0181] S = Ws·Ss + We·Se + Wi·Si

[0182] S s S e S i These are normalized scores (0-100 points) for three dimensions: safety, efficiency, and interaction cognition.

[0183] W s W e W i These are the weights for the corresponding dimensions, and W s +W e +W i =1.

[0184] Recommended weight: W s =0.5 (highest safety weight), W e =0.2, W i =0.3 (reflecting the emphasis on higher-order intelligence).

[0185] With interactive cognition S i For example, the score is calculated as follows:

[0186] I and H can be avg Combined,

[0187] S i =Sigmoid(k1·I)+(1−Sigmoid(k2·H avg ))

[0188] The Sigmoid function is used to map values ​​of different dimensions to the (0,1) interval; k1 and k2 are adjustment coefficients.

[0189] This formula indicates that S i It is positively correlated with the cumulative entropy reduction rate I and with the average cognitive entropy H. avg Negative correlation.

[0190] To better understand this invention, specific embodiments have been described in detail above, but these are not intended to limit the invention. Any simple modifications made to the above embodiments based on the technical essence of this invention still fall within the scope of this invention. Each embodiment in this specification focuses on its differences from other embodiments; similar or identical parts between embodiments can be referred to mutually. For system embodiments, since they basically correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

Claims

1. A method for evaluating the interactive cognitive ability of an autonomous vehicle, comprising constructing a hierarchical interactive scenario library and collecting high-frequency data, characterized in that, It also includes the following steps: Step 1: Construct a cognitive entropy model; Step 2: Perform interactive cognitive computing based on entropy reduction; Step 3: Construct a comprehensive scoring model and perform comprehensive scoring and level classification.

2. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 1, characterized in that, Step 1 involves probabilistically modeling the intentions of a single traffic participant i within the system's perception range over a future period, resulting in A. i ={P i (a,t)|a represents all possible actions}, where A i For calculating the probability of intent, t is the current time, P i (a,t) represents the probability of performing action a at time t, Σp i (a,t)=1.

3. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 2, characterized in that, The single-entity information entropy H i (t) represents the uncertainty of the intention of a single traffic participant i, and the formula is: 。 4. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 3, characterized in that, The global cognitive entropy H global (t) is the weighted sum of the individual information entropies of all relevant traffic participants, and the formula is: ； w i (t) represents the weight of the influence of traffic participant i on the vehicle, and the formula is as follows: ； Where, d i (t) represents the distance between the vehicle and traffic participant i, TTC i (t) represents the collision time between the vehicle and traffic participant i, d0 and TTC0 are normalization constants, and α and β are weighting coefficients with α+β=1.

5. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 4, characterized in that, Step 2 includes the following sub-steps: Step 21: Calculate the entropy reduction rate; Step 22: Calculate the cumulative entropy reduction.

6. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 5, characterized in that, The formula for calculating the entropy reduction rate R(t,Δt) is as follows: ； Where Δt is the time interval between time t and the next sampling time.

7. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 6, characterized in that, The formula for calculating the cumulative entropy reduction I is as follows: ； Where tend is the end time and tstart is the start time.

8. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 7, characterized in that, Step 3 includes the following sub-steps: Step 31: Extract key metrics for the complete test scenario; Step 32: Construct a comprehensive scoring model and calculate the comprehensive score S using weighted summation; Step 33: Classify the capability level of the autonomous vehicle based on the S score.

9. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 8, characterized in that, The key indicators include security indicators, efficiency indicators, and interactive cognition indicators. The safety indicators include maximum deceleration, minimum TTC, and whether a collision occurs. The efficiency metrics include task completion time and average speed; The interactive cognitive indicators include average global cognitive entropy, maximum global cognitive entropy, cumulative entropy reduction, and average entropy reduction rate; The average global cognitive entropy H avg The calculation formula is: ; The average entropy reduction rate R avg The calculation formula is: .

10. The method for evaluating the interactive cognitive ability of an autonomous vehicle as described in claim 9, characterized in that, The formula for calculating the comprehensive score S is as follows: S=W s ·S s +W e ·S e +W i ·S i Among them, S s S e S i The normalized scores for security, efficiency, and interaction cognition are respectively given by W. s W e W i Weights are assigned to three dimensions: security, efficiency, and interactive cognition, and W... s +W e +W i =1.

Images