Fuzzy traffic rule optimization method based on sequential logic

Abstract

The invention relates to a fuzzy traffic rule optimization method based on sequential logic, and the method comprises the following steps: sorting and classifying natural language traffic rules related to a target scene, and carrying out the classification and induction of the rules according to the state, behavior and sequential relation of an interactive vehicle; adopting sequential logic to digitalize the natural language traffic rule, and identifying a fuzzy proposition; key parameters of the fuzzy proposition are supplemented from three dimensions of behavior, time and space in combination with empirical parameters and existing research; building a simulation test platform; the parameter range of the fuzzy proposition is determined in combination with statistical analysis and machine learning methods, optimization is carried out through a genetic algorithm according to the simulation result of the simulation test platform, the optimal parameter of the fuzzy proposition is obtained, and traffic rule optimization is achieved. Compared with the prior art, the method has the advantages that the adaptability, interpretability and execution efficiency of the fuzzy traffic rule in an automatic driving scene are remarkably improved, and the scientificity of rule optimization and the safety of an automatic driving system are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and in particular, to a method for optimizing fuzzy traffic rules based on temporal logic. Background Art

[0002] With the rapid development of autonomous driving technology, the traffic operation environment is becoming increasingly complex. Since autonomous driving is a new technology and is still in the stage of immature technical system, fully driverless vehicles are difficult to achieve in the short term. In a quite long period in the future, the traffic environment where human-driven vehicles coexist with autonomous driving vehicles (i.e., "mixed human-machine driving") will become the norm. The complexity of this mixed traffic environment poses severe challenges to the popularization and implementation of autonomous driving technology.

[0003] In order to drive safely in a mixed traffic environment, autonomous driving vehicles not only need to perceive the surrounding environment and plan reasonable driving paths, but also need to have the ability to learn and comply with traffic rules. The driving behavior of human drivers depends to a large extent on the understanding of road traffic regulations, which are designed specifically for human drivers based on human cognitive abilities and subjective judgments. For example, common descriptions in regulations such as "shall not obstruct the normal driving of other vehicles", "maintain a safe distance", and "pay attention to avoiding pedestrians" rely on the driver's empirical judgment in specific traffic scenarios. However, autonomous driving vehicles cannot learn and understand the ambiguous descriptions in the rules like human drivers, which makes it difficult to directly apply existing traffic rules in autonomous driving scenarios.

[0004] Specifically, the ambiguous expressions in existing traffic rules have the following problems:

[0005] Rules are not quantifiable: Ambiguous descriptions in traffic regulations (such as "safe distance", "shall not obstruct") lack clear numerical definitions, and it is difficult for autonomous driving vehicles to generate clear decision-making bases accordingly. For example, the "safe distance" should have different definitions under different speed conditions, but the specific values are not clearly defined in the regulations.

[0006] Strong scene dependence: Fuzzy rules need to be flexibly operated in combination with scenes in the actual traffic environment.

[0007] Obvious driving differences: Different drivers may have different understandings of the same fuzzy rule, while autonomous driving vehicles need to execute the rules consistently, and this difference may lead to behavioral conflicts between autonomous driving vehicles and human drivers.

[0008] Due to these problems, it is often difficult for autonomous vehicles to effectively incorporate traffic rules into trajectory planning. In existing research, there is still a lack of digitalization and optimization methods for fuzzy rules. Therefore, there is an urgent need for an innovative method to accurately digitalize and optimize existing fuzzy traffic rules so that they can be directly understood and executed by autonomous driving systems. Summary of the Invention

[0009] The purpose of the present invention is to provide a method for optimizing fuzzy traffic rules based on temporal logic, which can not only solve the adaptability problem of fuzzy rules in the autonomous driving scenario, but also deeply integrate these rules into the trajectory planning algorithm, providing strong guarantees for the safety, efficiency, and comfort of autonomous vehicles in a mixed traffic environment.

[0010] The purpose of the present invention can be achieved through the following technical solutions:

[0011] A method for optimizing fuzzy traffic rules based on temporal logic, comprising the following steps:

[0012] Step 1) Sort and classify the natural language traffic rules involved in the target scenario, and classify and summarize the rules according to the states, behaviors, and temporal relationships of interacting vehicles;

[0013] Step 2) Digitalize the natural language traffic rules using temporal logic and identify fuzzy propositions;

[0014] Step 3) Supplement the key parameters of the fuzzy propositions from three dimensions of behavior, time, and space in combination with empirical parameters and existing research;

[0015] Step 4) Build a simulation test platform;

[0016] Step 5) Determine the parameter range of the fuzzy propositions by combining statistical analysis and machine learning methods, and optimize through a genetic algorithm according to the simulation results of the simulation test platform to obtain the optimal parameters of the fuzzy propositions and achieve traffic rule optimization.

[0017] The said Step 2) comprises the following steps:

[0018] Step 21) Use temporal logic to convert the natural language description of each rule into a propositional form, and decompose it into a subject, behavior, temporal constraint, and spatial constraint;

[0019] Step 22) Search for fuzzy propositions that are difficult to decompose and execute from the relationships of behavior, time, and space, as well as fuzzy propositions that can neither be refined into specific behaviors nor combined with key parameters of time or space.

[0020] The said Step 21) comprises the following steps:

[0021] Subject extraction: Identify the core participants in traffic rules;

[0022] Behavior recognition: Determine the key behaviors involved in the rules;

[0023] Temporal constraint analysis: Analyze the temporal relationships in the rules, including identifying constraints based on event order and constraints based on time variables;

[0024] Spatial constraint analysis: Determine the spatial variables involved in the rules, including identifying the relative distances between vehicles, lateral / longitudinal spacing constraints, and fixed spatial regions.

[0025] In step 3), for the behavior dimension, select one or a combination of the following parameters as the behavior quantification parameters of the fuzzy proposition: mean longitudinal speed, standard deviation of longitudinal speed, mean lateral speed, standard deviation of lateral speed, mean speed, standard deviation of speed, mean longitudinal acceleration, standard deviation of longitudinal acceleration, longitudinal acceleration change rate, mean lateral acceleration, standard deviation of lateral acceleration, lateral acceleration change rate, mean acceleration, standard deviation of acceleration, acceleration change rate.

[0026] In step 3), for the time dimension, select one or a combination of the following parameters as the time quantification parameters of the fuzzy proposition: collision time, time-integrated collision time, dwell time, reaction time, lane-changing duration, time headway.

[0027] In step 3), for the spatial dimension, select one or a combination of the following spatial parameters as the spatial quantification parameters of the fuzzy proposition: relative distance, lateral relative distance, longitudinal relative distance, safety distance with a fixed value, dynamic distance.

[0028] Step 4) includes the following steps:

[0029] Step 41) Accumulate safety-critical events using actual data, reconstruct the traffic scenarios involved in the simulation test platform, and keep the positions and states of the interacting objects unchanged;

[0030] Step 42) Replace the target vehicle with a car equipped with an autonomous driving algorithm to evaluate the adaptability and effectiveness of traffic rule optimization.

[0031] Step 5) includes the following steps:

[0032] Step 51) Determine the parameter range of the fuzzy proposition through statistical analysis;

[0033] Step 52) Construct a multi-objective optimization function based on maximizing safety, efficiency, and comfort;

[0034] Step 53) In the simulation test platform built in Step 4), based on the reconstructed scenario, with the multi-objective optimization function as the optimization goal, use the Non-dominated Sorting Genetic Algorithm II to calibrate the quantization parameters, determine the optimal parameter combination, and achieve traffic rule optimization.

[0035] Specifically, Step 51) is as follows: Use the maximum value, minimum value, 15th percentile, and 85th percentile to delimit the preliminary range of the parameters, and combine k-means++ clustering to supplement the feature values to obtain the value ranges of different parameters.

[0036] The multi-objective optimization function is defined as follows:

[0037]

[0038] where maxf Safety is the maximum safety objective, maxf efficiency is the maximum efficiency objective, maxf Comfort is the maximum comfort objective,

[0039]

[0040] where T is the simulation duration, N is the number of calibrated scenarios, TIT i is the TIT value in the calibrated scenario i, TTC * is the threshold of TTC, TTC i (t) is the TTC value at time t in the calibrated scenario i, and Δt is the simulation time interval;

[0041]

[0042] where Δv i is the speed difference between the average speed and the speed limit v limit in the calibrated scenario i, and v i (t) is the speed at time t;

[0043]

[0044] where Jerk i is the cumulative jerk value of the calibrated scenario i, and a i (t) is the acceleration value at time t in the calibrated scenario i.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] (1) Solved the ambiguity problem of traffic rules. The present invention digitizes natural language traffic rules through temporal logic, formalizes the ambiguous description in human driving experience into precise logical expressions, and combines quantization parameters to significantly improve the comprehensibility and executability of the rules.

[0047] (2) Multi-dimensional parameter optimization. Introduce three dimensions of behavioral parameters, time parameters, and space parameters, and combine statistical analysis, machine learning, and NSGA-II genetic algorithm for parameter calibration to ensure the rationality and optimality of parameter selection, taking into account safety, efficiency, and comfort.

[0048] (3) Improved the safety of the autonomous driving system. By quantifying and optimizing fuzzy propositions, reduce the ambiguity of traffic rules in the autonomous driving scenario, reduce the risk of behavioral conflicts of the autonomous driving system in complex interaction scenarios, and significantly improve driving safety. Description of the Drawings

[0049] Figure 1 is the flowchart of the method of the present invention;

[0050] Figure 2 is a schematic diagram of the safety comparison of different control methods in an embodiment;

[0051] Figure 3 is a schematic diagram of the stability comparison of different control methods in an embodiment. Detailed Embodiment

[0052] The present invention will be described in detail below with reference to the drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and gives detailed implementation manners and specific operation processes, but the protection scope of the present invention is not limited to the following embodiments.

[0053] This embodiment provides a method for optimizing fuzzy traffic rules based on temporal logic, as Figure 1 shown, including the following steps:

[0054] Step 1) Sort and classify the natural language traffic rules involved in the target scenario, and classify and summarize the rules according to the state, behavior, and temporal relationship of the interacting vehicles.

[0055] In this embodiment, taking the scenario of pedestrians crossing the road as an example, sort and classify the natural language traffic rules involved in the scenario of pedestrians crossing the road. According to the different states, actions, and timings of the interacting vehicles, they are divided into three categories: normal passing, decelerating to give way, and stopping to give way.

[0056] Step 2) Digitize the natural language traffic rules using temporal logic and identify fuzzy propositions.

[0057] Step 21) Use temporal logic to convert the natural language description of each rule into a propositional form, and decompose it into a subject, an action, temporal constraints, and spatial constraints.

[0058] 1. Subject extraction: Identify the core participants in traffic rules, such as left-turning vehicles, straight-going vehicles, pedestrians, traffic lights, etc.

[0059] 2. Action recognition: Determine the key actions involved in the rule, such as "yield", "stop", "enter the intersection", etc.

[0060] 3. Temporal constraint analysis: Analyze the time relationships in the rule, including:

[0061] Constraints based on the order of events: For example, "Left-turning vehicles should pass after straight-going vehicles" can be converted into the logical condition of "Straight-going vehicles pass through the intersection first".

[0062] Constraints based on time variables: For example, parameters such as "headway".

[0063] 4. Spatial constraint analysis: Determine the spatial variables involved in the rule, such as:

[0064] The relative distance between vehicles, lateral / longitudinal spacing constraints;

[0065] Fixed spatial areas such as intersections, lane boundaries, etc.

[0066] In this embodiment, taking the scenario of pedestrians crossing the road as an example, the metric temporal logic (MTL) is used to convert three types of rules, namely normal passing, slow down to yield, and stop to yield, into propositional forms. Taking the slow down to yield rule as an example, the natural language traffic rule is "When a motor vehicle passes through an intersection without traffic lights, traffic signs, traffic markings, or the command of a traffic police officer, it shall slow down and give way to pedestrians and vehicles with the right of way." Among them, the core traffic participants are motor vehicles and pedestrians, the actions are slow down and give way, the time constraint is sunny daytime, and the spatial constraint is an intersection without traffic lights, traffic signs, traffic markings, or the command of a traffic police officer. Then the MTL propositional form is as follows,

[0067] Passing:

[0068] Slowing down:

[0069] Stopping:

[0070] Among them, T pass 、T deceleration and Tstop respectively indicate whether normal passage, yielding while decelerating, and yielding while stopping operations are carried out; T indicates yes; F indicates no; R intersection / segment indicates whether it is located at an intersection / section; T uncontrolled indicates whether there is no signal control at the intersection / section; T maintenance indicates whether the intersection / section is under temporary control; Type(Pedestrian) indicates whether the host vehicle interacts with pedestrians; SafetyViolation indicates whether safety constraints are violated; Movement(cross) indicates whether the target executes a crossing behavior; Location(crosswalk) indicates whether it is located at a crosswalk; Weather(sun) indicates whether it is sunny; Light(day) indicates whether it is daytime; Road(dry) indicates whether the ground is dry.

[0071] Step 22) Find fuzzy propositions that are difficult to decompose and execute from the behavioral, temporal, and spatial relationships, as well as fuzzy propositions that can neither be refined into specific behaviors nor combined with key temporal or spatial parameters.

[0072] For example: The safety proposition is difficult to be further refined into actions, and at the same time, no suitable key parameters can be found from the temporal and spatial perspectives. Therefore, the SafetyViolation under yielding while decelerating T deceleration is a fuzzy proposition and requires quantification of key parameters. That is, it is necessary to clarify under what safety conditions the action of yielding while decelerating needs to be executed.

[0073] Step 3) Combine empirical parameters and existing research to supplement the key parameters of fuzzy propositions from the three dimensions of behavior, time, and space.

[0074] Step 31) For the behavioral dimension, select one or a combination of the following parameters as the behavioral quantification parameters of the fuzzy proposition: mean longitudinal speed, standard deviation of longitudinal speed, mean lateral speed, standard deviation of lateral speed, mean speed, standard deviation of speed, mean longitudinal acceleration, standard deviation of longitudinal acceleration, longitudinal acceleration change rate, mean lateral acceleration, standard deviation of lateral acceleration, lateral acceleration change rate, mean acceleration, standard deviation of acceleration, acceleration change rate.

[0075] Step 32) For the temporal dimension, select one or a combination of the following parameters as the temporal quantification parameters of the fuzzy proposition: time to collision, time integral of time to collision, dwell time, reaction time, lane change duration, time headway.

[0076] Step 33) For the spatial dimension, select one or a combination of the following spatial parameters as the spatial quantization parameter of the fuzzy proposition: relative distance, lateral relative distance, longitudinal relative distance, safety distance with a fixed value, dynamic distance. Among them, the dynamic clustering is the distance based on the responsibility-sensitive model or the distance traveled by the vehicle within 3 seconds. The determination of the distance based on the responsibility-sensitive model can be referred to the literature:

[0077] 1) Xu, X., Wang, X., Wu, X., Hassanin, O., & Chai, C. (2021). Calibration and evaluation of the Responsibility-Sensitive Safety model of autonomous car-following maneuvers using naturalistic driving study data. Transportation Research Part C: Emerging Technologies, 123, 102988. doi:10.1016 / j.trc.2021.102988;

[0078] 2) Wang, X., Ye, C., Quddus, M., & Morris, A. (2023). Pedestrian safety in an automated driving environment: calibrating and evaluating the responsibility-sensitive safety model. Accident Analysis & Prevention, 192, 107265. doi:10.1016 / j.aap.2023.107265;

[0079] To avoid obscuring the purpose of this application, it will not be elaborated further in this embodiment.

[0080] Specifically, for propositions that cannot be directly converted into specific numerical values, it is necessary to further analyze the source of their fuzziness and supplement it. For example:

[0081] 1. Actions that are difficult to decompose (words that are not easily directly mapped to specific physical variables), such as:

[0082] For "smooth driving", it needs to be measured by the standard deviation of acceleration or the rate of change of acceleration.

[0083] For "slowly passing through", it needs to use speed (such as ≤5 km / h), acceleration threshold (≤3 m / s2 ) statement

[0084] For "yielding the right of way", safety indicators are required and it should be measured in combination with the time - space range. For example, under what time - space range does it meet what safety standard and then yielding the right of way is required.

[0085] 2. Vague expressions in time relationships, such as:

[0086] For "brief stay", indicators such as stay time are required for quantification.

[0087] For "decelerating in advance", it is necessary to define how many seconds in advance (such as the 3s rule).

[0088] For "accelerating through as soon as possible", it needs to be measured by reaction time or acceleration threshold.

[0089] For "evading in time", it needs to be measured by reaction time or TTC.

[0090] 3. Vague expressions in space relationships, such as:

[0091] For "maintaining a reasonable vehicle distance", it can be quantified as relative distance or time - headway.

[0092] For "appropriate lane change", it needs to be combined with lateral acceleration or lane - change duration.

[0093] In this embodiment, taking the determination of key parameters from the time perspective as an example: Select the Time - To - Collision (TTC) as the safety indicator, decompose the deceleration of the vehicle yielding the right of way into a deceleration process in three stages, and trigger different deceleration strategies through TTC calculation.

[0094] Step 4) Build a simulation test platform.

[0095] Step 41) Use actual data to accumulate safety - critical events, reconstruct the involved traffic scenarios in the simulation test platform, and keep the positions and states of the interacting objects unchanged;

[0096] Step 42) Replace the target vehicle with a car equipped with an autonomous driving algorithm to evaluate the adaptability and effectiveness of traffic rule optimization.

[0097] In this embodiment, reconstruct 150 pedestrian - crossing scenarios extracted from Shanghai natural driving data, and keep the positions and states of the interacting objects unchanged. Replace the vehicle in each pedestrian - crossing scenario with a car equipped with an autonomous emergency braking (AEB) system with a three - stage deceleration strategy, and keep the position and state of the pedestrian unchanged.

[0098] Step 5) Combine statistical analysis and machine learning methods to determine the parameter range of fuzzy propositions, and based on the simulation results of the simulation test platform, optimize through genetic algorithms to obtain the optimal parameters of fuzzy propositions, thus realizing traffic rule optimization.

[0099] Step 51) Determine the parameter range of fuzzy propositions through statistical analysis: Use the maximum value, minimum value, 15th percentile, and 85th percentile to delimit the preliminary range of parameters, and supplement eigenvalue through k-means++ clustering to obtain the value ranges of different parameters.

[0100] Among them, the range of TTC parameters is that in the first deceleration stage, TTC is in (4 seconds, 3.5 seconds), in the second deceleration stage, TTC is in (3 seconds, 2.5 seconds), and in the third deceleration stage, TTC is in (2 seconds, 1.5 seconds).

[0101] Step 52) Construct a multi-objective optimization function based on the maximum safety, maximum efficiency, and maximum comfort.

[0102] Multi-objective optimization function It is defined as follows:

[0103]

[0104] Among them, maxf Safety is the maximum safety objective, maxf efficiency is the maximum efficiency objective, maxf Comfort is the maximum comfort objective,

[0105]

[0106] Among them, T is the simulation duration, N is the calibrated number of scenarios, TITi i is the TIT (Time-to-Interaction termination) value in the calibrated scenario i, TTC * is the threshold of TTC (Time-to-Collision), TTC i (t) is the TTC value at time t in the calibrated scenario i, and Δt is the simulation time interval;

[0107]

[0108] Among them, Δv i is the speed difference between the average speed and the speed limit v limit in the calibrated scenario i, and v i (t) is the speed at time t;

[0109]

[0110] Among them, Jerk i is the cumulative jerk value of the calibration scenario i, and a i (t) is the acceleration value at time t in the calibration scenario i.

[0111] Step 53) In the simulation test platform built in step 4), based on the reconstructed scenario, with the multi-objective optimization function as the optimization goal, use the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to calibrate the quantization parameters, determine the optimal parameter combination, and achieve traffic rule optimization.

[0112] In the simulation test platform built in step 4), based on the extracted 150 pedestrian crossing scenarios, use NSGA-II to calibrate TTC, and the optimal parameter combinations for the three stages of TTC are 4s, 3s, and 1.5s. That is, when the TTC between the vehicle and the pedestrian is less than 4s, it indicates a violation of the safety constraint SafetyViolation, and a deceleration yielding action needs to be executed. The specific deceleration yielding strategy is that when the TTC between the vehicle and the pedestrian is less than 4s but greater than 3s, the first-stage deceleration is adopted; when the TTC between the vehicle and the pedestrian is less than 3s but greater than 1.5s, the second-stage deceleration is adopted; when the TTC between the vehicle and the pedestrian is less than 1.5s, the third-stage deceleration is adopted.

[0113] After comparing human drivers, autonomous driving control based on optimized traffic rules, and the Responsibility-Sensitive Safety (RSS model), the results show that the solution based on optimized traffic rules performs best in terms of safety and stability. As Figure 2 and Figure 3 shown, compared with human drivers and the RSS model, the Time Integrated Time-to-collision (TIT) of the solution based on optimized traffic rules has increased by 70.77% and 32.6% respectively, which proves its better safety. And the Time-Exposed TTC (TET) has increased by 389.76% and 3.38% respectively, further indicating that its stability has been significantly improved.

[0114] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field of the present invention based on the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art should be within the protection scope determined by the claims.

Claims

1. A fuzzy traffic rule optimization method based on temporal logic, characterized in that: The following steps are involved: Step 1) sorting and classifying the natural language traffic rules involved in the target scenario, and classifying and summarizing the rules according to the status, behavior and time sequence relationship of the interacting vehicles; Step 2) using temporal logic to digitize natural language traffic rules and identify fuzzy propositions; Step 3) Combine empirical parameters with existing research to supplement the key parameters of fuzzy propositions from three dimensions: behavior, time, and space; Step 4) Building a simulation test platform; Step 5) Combine statistical analysis with machine learning methods to determine the parameter range of the fuzzy proposition, and optimize it through genetic algorithms based on the simulation results of the simulation test platform to obtain the optimal parameters of the fuzzy proposition and achieve traffic rule optimization.

2. The fuzzy traffic rule optimization method based on temporal logic according to claim 1 is characterized in that: The step 2) comprises the following steps: Step 21) using temporal logic to convert the natural language description of each rule into propositional form, decomposing it into subject, action, temporal constraint and spatial constraint; Step 22) Find fuzzy propositions that are difficult to decompose and execute from the behavior, time and space relationships, as well as fuzzy propositions that can neither be refined into specific behaviors nor combined with key time or space parameters.

3. The fuzzy traffic rule optimization method based on temporal logic according to claim 2 is characterized in that: The step 21) comprises the following steps: Subject extraction: identifying the key players in traffic rules; Behavior identification: determine the key behaviors involved in the rules; Timing constraint analysis: parsing the temporal relationships in rules, including identifying constraints based on event order and constraints based on time variables; Spatial constraint analysis: Determine the spatial variables involved in the rule, including identifying the relative distance between vehicles, lateral / longitudinal spacing constraints, and fixed spatial areas.

4. The fuzzy traffic rule optimization method based on temporal logic according to claim 1 is characterized in that: In the step 3), for the behavior dimension, one of the following parameters or a combination thereof is selected as the behavior quantification parameter of the fuzzy proposition: longitudinal speed mean, longitudinal speed standard deviation, lateral speed mean, lateral speed standard deviation, speed mean, speed standard deviation, longitudinal acceleration mean, longitudinal acceleration standard deviation, longitudinal acceleration change rate, lateral acceleration mean, lateral acceleration standard deviation, lateral acceleration change rate, acceleration mean, acceleration standard deviation, acceleration change rate.

5. The fuzzy traffic rule optimization method based on temporal logic according to claim 1 is characterized in that: In the step 3), for the time dimension, one of the following parameters or a combination thereof is selected as the time quantification parameter of the fuzzy proposition: collision time, time-integrated collision time, dwell time, reaction time, lane change duration, and headway time.

6. The fuzzy traffic rule optimization method based on temporal logic according to claim 1 is characterized in that: In the step 3), for the spatial dimension, one of the following spatial parameters or a combination thereof is selected as the spatial quantification parameter of the fuzzy proposition: relative distance, horizontal relative distance, vertical relative distance, fixed value safety distance, and dynamic distance.

7. The fuzzy traffic rule optimization method based on temporal logic according to claim 1 is characterized in that: The step 4) comprises the following steps: Step 41) Utilize the actual data to accumulate safety-critical events, reconstruct the involved traffic scenarios in the simulation test platform, and keep the positions and states of the interactive objects unchanged; Step 42) Replace the target vehicle with a car equipped with an autonomous driving algorithm to evaluate the adaptability and effectiveness of traffic rule optimization.

8. The fuzzy traffic rule optimization method based on temporal logic according to claim 1 is characterized in that: The step 5) comprises the following steps: Step 51) determining the parameter range of the fuzzy proposition through statistical analysis; Step 52) constructing a multi-objective optimization function based on maximum safety, maximum efficiency, and maximum comfort; Step 53) In the simulation test platform built in step 4), based on the reconstructed scene, with the multi-objective optimization function as the optimization target, the non-dominated sorting genetic algorithm II is used to calibrate the quantitative parameters, determine the optimal parameter combination, and realize traffic rule optimization.

9. The fuzzy traffic rule optimization method based on temporal logic according to claim 8 is characterized in that: The step 51) is specifically as follows: using the maximum value, minimum value, 15th percentile and 85th percentile to define the initial range of the parameter, and combining k-means++ clustering to supplement the eigenvalues ​​to obtain the value range of different parameters.

10. The fuzzy traffic rule optimization method based on temporal logic according to claim 8, characterized in that: The multi-objective optimization function The definition is as follows: Among them, maxf Safety is the maximum safety target, maxf efficiency is the maximum efficiency target, maxf Comfort For the ultimate comfort goal, Where T is the simulation time, N is the number of calibrated scenes, TIT i is the TIT value in the calibration scene i, TTC * is the threshold of TTC, TTC i (t) is the TTC value at time t in the calibration scene i, and Δt is the simulation time interval; Where Δv i is the average speed and speed limit v in the calibration scene i limit The speed difference between i (t) is the speed at time t; Among them, Jerk i is the cumulative jerk value of calibration scene i, a i (t) is the acceleration value at time t in the calibration scene i.

Images