Decision behavior multi-dimensional evaluation method and system under multiple driving modes, and medium

Abstract

The invention relates to a multi-dimensional evaluation method and system for decision behaviors in multiple driving modes and a medium, and the method comprises the steps: U2, inputting a full-quantity semantic-level decision behavior group into a rigid fusing evaluation layer of a behavior decision multi-dimensional evaluation model suitable for multiple driving modes, and carrying out the evaluation of a safety dimension and a compliance dimension, if the evaluation does not reach the standard, directly judging that the decision behavior is not feasible, and if the evaluation reaches the standard, judging that the decision behavior is feasible and entering the step U3; and U3, inputting the semantic-level decision behavior passing through the rigid fusing evaluation layer into a subdivision weighting evaluation layer of the behavior decision multi-dimensional evaluation model suitable for multiple driving modes to evaluate the efficiency dimension and the comfort dimension, and calculating the total cost of subdivision weighting. According to the method, the problem of non-uniform evaluation normal forms and evaluation dimensions of the decision behavior evaluation model in multiple scenes is solved, the complexity of decision behavior evaluation is greatly simplified, and the accuracy and efficiency of decision behavior evaluation are improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a multi-dimensional evaluation method, system, and medium for decision-making behavior in multiple driving modes. Background Art

[0002] The behavioral decision-making module is a core component of autonomous driving systems and plays a vital role in improving their safety and intelligence. Behavioral decision-making algorithms include rule-based and data-driven ones. These algorithms identify driving scenarios based on real-time environmental information (such as surrounding vehicles, obstacles, and semantic maps), sensor data, and traffic regulations, and then formulate appropriate lateral and longitudinal behaviors. However, determining whether the behavioral outcomes generated by these decision-making algorithms are safe, efficient, comfortable, and compliant has long been a challenge in the industry. Traditional behavioral evaluation models focus on a single driving scenario, using different evaluation models for different scenarios. This leads to high scenario-dependence, mixed metrics (such as safety and comfort) and the difficulty of configuring scenario-based evaluation models based on driving modes. To address these issues, it is necessary to design an intelligent, scalable, and multi-dimensional evaluation model that comprehensively considers safety, efficiency, comfort, compliance, and driving mode variations, enabling autonomous vehicles to adopt more appropriate behaviors.

[0003] In the prior art, a Chinese patent (Application Number: 202411422069.3, Publication Number: CN 118991829A) discloses a safety constraint behavior decision system that considers the social preferences of vehicles behind the target lane. This system utilizes hierarchical reinforcement learning to implement autonomous driving behavior decisions. The system includes a first social preference recognition model, a second social preference recognition model, an upper-layer module, and a lower-layer module. The first social preference recognition model is configured to identify the social preference type of the nearest vehicle behind the adjacent lane that has entered the autonomous vehicle's range, obtaining a first recognition result. The second social preference recognition model is configured to identify the social preference type of the nearest vehicle behind the target lane, obtaining a second recognition result. The upper-layer module is configured to make autonomous driving behavior decisions based on the autonomous vehicle's driving conditions, the adjacent vehicle's driving conditions, and the first recognition result. If the first recognition result is self-serving, a penalty is imposed if the autonomous driving behavior decision is to change lanes to the adjacent lane. The lower-layer module is configured to execute a lane change decision to return to the lane if the second recognition result is self-serving when the autonomous vehicle initiates a left or right lane change. This patent solution mainly introduces a safety constraint behavior decision system that takes into account the social preferences of the vehicles behind the target lane. It cannot cover the evaluation of the vehicle's decision-making behavior in scenarios where there is only a vehicle in front of the target lane or there are vehicles in both the front and rear directions.

[0004] In the prior art, a Chinese patent (application number: 202411993344.7, publication number: CN 119773807A) discloses an autonomous driving behavior decision-making method, a driving decision-making device, and a readable storage medium. The method calls a driving behavior policy network and a driving cost evaluation network to perform a Monte Carlo tree search based on the actual driving environment state. When the search termination condition is met, the optimal child node corresponding to the root node in the Monte Carlo tree that minimizes the driving behavior cost is determined. Each node in the Monte Carlo tree represents a driving environment state, and the root node represents the actual driving environment state. The initial driving behavior cost of each node is obtained by the driving cost evaluation network based on the driving behavior prediction result of the node and the driving environment state evaluation. The driving behavior prediction result of each node is obtained by the driving behavior policy network based on the driving environment state estimation of the node. The driving behavior policy network and the driving cost evaluation network are trained based on an adaptive dynamic programming structure. This patent solution mainly introduces an autonomous driving behavior decision-making method, a driving decision-making device and a readable storage medium. The cost evaluation network of this method is iteratively optimized based on the initial network weights and driving network training samples. There is a problem that the weight coefficient is fixed, which makes it impossible to dynamically adjust the network weights according to different driving modes. At the same time, there is a problem of insufficient generalization ability for long-tail scenarios due to strong dependence on training samples. Summary of the Invention

[0005] In view of the above shortcomings of the existing technology, the present invention provides a multi-dimensional evaluation method, system and medium for decision-making behavior in multiple driving modes. It not only solves the problem of inconsistent evaluation paradigms and evaluation dimensions of decision-making behavior evaluation models in multiple scenarios, but also greatly simplifies the complexity of decision-making behavior evaluation and improves the accuracy and efficiency of decision-making behavior evaluation.

[0006] In order to achieve the above-mentioned and other related purposes, the present invention provides the following technical solutions:

[0007] A multi-dimensional evaluation method for decision-making behavior in multiple driving modes, the method comprising:

[0008] U1. The autonomous driving system's behavior decision module generates a comprehensive semantic-level decision behavior set based on temporal, lateral, and longitudinal behavior.

[0009] U2. Input the full set of semantic-level decision-making behaviors into the rigid fuse evaluation layer of the multi-dimensional evaluation model for behavioral decision-making in multiple driving modes for safety and compliance evaluation. If the evaluation fails, the decision-making behavior is directly judged as infeasible. If the evaluation meets the standards, the decision-making behavior is judged as feasible and proceeds to step U3.

[0010] U3. Input the semantic-level decision-making behavior from the rigid fuse evaluation layer into the segmented weighted evaluation layer of the multi-dimensional evaluation model for behavioral decision-making applicable to multiple driving modes. This layer evaluates the efficiency and comfort dimensions and calculates the segmented weighted total cost.

[0011] U4. Traverse the cost of the decision behaviors in the decision behavior group and find the decision behavior execution with the lowest weighted total cost.

[0012] Furthermore, in step U2, the rigid fuse evaluation layer evaluates the decision-making behavior to construct a responsibility-sensitive strong safety model. The responsibility-sensitive strong safety model includes the longitudinal safety distance maintenance principle, the lateral safety distance maintenance principle and the minimum safety distance unchanged lane principle. The longitudinal minimum safety distance calculation function d of the longitudinal safety distance maintenance principle is lon,min for,

[0013] ,

[0014] Among them, v f is the longitudinal velocity of the preceding vehicle, v r is the longitudinal speed of the following vehicle, t lon is the longitudinal minimum reaction time, β lon,min is the minimum longitudinal braking acceleration, β lon,max is the maximum longitudinal braking acceleration, a lon,max is the maximum longitudinal acceleration.

[0015] Furthermore, the lateral minimum safety distance calculation function d of the lateral safety distance maintenance principle is lat,min for,

[0016] ,

[0017] Among them, v ego,lat is the lateral safety distance based on the lateral speed of the vehicle, v other,lat is the lateral speed of the vehicle, t lat is the minimum lateral reaction time, β lat,min is the minimum lateral vehicle deceleration.

[0018] Furthermore, the principle of the minimum safety distance being unchanged is a further safety constraint on the lateral behavior of the lane change decision-making behavior, and a preset longitudinal safety margin γ is set. lat and lateral safety margin γ lon , when the calculated minimum longitudinal safety distance d lon,min ≤γ lat When the lane change decision behavior is evaluated as infeasible; when the lateral minimum safety distance d is calculated lon,min ≤γ lon , the lane-changing decision behavior is evaluated as infeasible.

[0019] Furthermore, in the rigid fuse evaluation layer, the compliance dimension considers all traffic regulations to evaluate the decision-making behavior, including the compliance of lateral lane change behavior, longitudinal acceleration behavior and longitudinal uniform speed behavior. If the lane line within a certain longitudinal range of the lane where the vehicle is located is a solid line, the left lane line of the target lane for left lane change is a solid line, or the right lane edge line of the target lane for right lane change is a solid line, the corresponding lane change decision behavior is evaluated as infeasible. If the longitudinal behavior is acceleration behavior or uniform speed behavior, but the vehicle speed has reached the maximum speed limit of the vehicle lane or the target lane, the corresponding decision behavior is evaluated as infeasible.

[0020] Furthermore, in step U3, the total efficiency cost function F of the efficiency dimension efficiency for,

[0021] ,

[0022] Among them, F e,v is the cost of the vehicle to reach the desired speed, F e,n is the cost of the ego vehicle reaching the desired driving lane, Δv1 is the speed difference between the ego vehicle exceeding the speed limit and the desired speed, Δv2 is the speed difference between the ego vehicle failing to reach the desired speed and the desired speed, Δv3 is the speed difference between the low speed ahead and the ego vehicle, Δv3 is the speed difference between the high speed ahead and the desired speed of the ego vehicle, y1 is the ego vehicle overspeeding discount factor, y2 is the ego vehicle failing to reach the desired speed discount factor, y3 is the front vehicle blocking the ego vehicle at low speed, y4 is the front vehicle blocking the ego vehicle from reaching the desired speed discount factor, v5 is the lane change speed, y5 is the lane change penalty discount factor, and y6 is the navigation recommended lane change reward discount factor; the total comfort cost function F of the comfort dimension is: comfort for,

[0023] F comfort =F c,lon +F c,lat , where F c,lon is the longitudinal comfort cost of the coarse trajectory in the time domain, F c,lat is the lateral comfort cost of the coarse trajectory in the time domain.

[0024] Furthermore, the longitudinal comfort cost F of the time domain coarse trajectory c,lon for,

[0025] ,

[0026] The lateral comfort cost F of the time domain rough trajectory c,lat for,

[0027] ,

[0028] Among them, n lat is the number of trajectory points exceeding the index threshold in the horizontal direction, nlon is the number of longitudinal trajectory points exceeding the index threshold, c lat is the lateral comfort threshold, c lon is the longitudinal comfort threshold, and N is the total number of trajectory points.

[0029] Furthermore, the function F of the weighted total cost of the subdivision cost for,

[0030] F cost =δ efficiency F efficiency +δ comfort F comfort ,

[0031] Among them, δ efficiency is the efficiency weight, δ comfort To adapt to different driving needs of users, a variety of driving modes are designed for comfort weight, including conservative mode, aggressive mode and balanced mode. Based on the driving mode selected by the user, the efficiency weight δ is dynamically adjusted. efficiency and comfort weight δ comfort .

[0032] In order to achieve the above-mentioned and other related purposes, the present invention also provides a multi-dimensional evaluation system for decision-making behavior in multiple driving modes, including a computer device that is programmed or configured to execute the steps of any one of the multi-dimensional evaluation methods for decision-making behavior in multiple driving modes.

[0033] In order to achieve the above-mentioned and other related purposes, the present invention also provides a computer-readable storage medium, which stores a computer program programmed or configured to execute any one of the multi-dimensional evaluation methods for decision-making behaviors in multiple driving modes.

[0034] The present invention has the following positive effects:

[0035] 1. This invention uses a rigid fuse evaluation layer that inputs the full semantic-level decision-making behavior group into a multi-dimensional evaluation model for behavioral decision-making applicable to multiple driving modes to evaluate safety and compliance dimensions. This not only solves the problem of inconsistent evaluation paradigms and evaluation dimensions in the decision-making behavior evaluation model across multiple scenarios, but also greatly simplifies the complexity of decision-making behavior evaluation and improves its accuracy and efficiency.

[0036] 2. The present invention evaluates the efficiency and comfort dimensions of the behavior decision-making multidimensional evaluation model applicable to multiple driving modes by inputting the semantic-level decision-making behavior through the rigid fuse evaluation layer into the segmented weighted evaluation layer, and traverses the cost of the decision-making behavior of the decision-making behavior group to find the decision-making behavior execution with the lowest segmented weighted total cost, thereby improving the intelligence and scalability of the decision-making behavior evaluation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of the method flow of the present invention;

[0038] Figure 2 This is a schematic diagram of the full semantic-level decision-making behavior of the present invention;

[0039] Figure 3 This is a schematic diagram of the minimum longitudinal safety distance in the case of a vehicle in front of the present invention;

[0040] Figure 4 This is a schematic diagram of the minimum lateral safety distance when there are vehicles around in the present invention. DETAILED DESCRIPTION

[0041] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0042] Example 1: Figure 1 or Figure 2 As shown, a multi-dimensional evaluation method for decision-making behavior in multiple driving modes is provided, the method comprising:

[0043] U1. The autonomous driving system's behavior decision module generates a comprehensive semantic-level decision behavior set based on temporal, lateral, and longitudinal behavior.

[0044] U2. Input the full set of semantic-level decision-making behaviors into the rigid fuse evaluation layer of the multi-dimensional evaluation model for behavioral decision-making in multiple driving modes for safety and compliance evaluation. If the evaluation fails, the decision-making behavior is directly judged as infeasible. If the evaluation meets the standards, the decision-making behavior is judged as feasible and proceeds to step U3.

[0045] U3. Input the semantic-level decision-making behavior from the rigid fuse evaluation layer into the segmented weighted evaluation layer of the multi-dimensional evaluation model for behavioral decision-making applicable to multiple driving modes. This layer evaluates the efficiency and comfort dimensions and calculates the segmented weighted total cost.

[0046] U4. Traverse the cost of the decision behaviors in the decision behavior group and find the decision behavior execution with the lowest weighted total cost.

[0047] In this embodiment, if Figure 3As shown, in step U2, the rigid fuse evaluation layer evaluates the decision-making behavior to build a responsibility-sensitive strong safety model. The responsibility-sensitive strong safety model includes the longitudinal safety distance maintenance principle, the lateral safety distance maintenance principle and the minimum safety distance unchanged lane principle. The longitudinal minimum safety distance calculation function d of the longitudinal safety distance maintenance principle lon,min for,

[0048] ,

[0049] Among them, v f is the longitudinal velocity of the preceding vehicle, v r is the longitudinal speed of the following vehicle, t lon is the longitudinal minimum reaction time, β lon,min is the minimum longitudinal braking acceleration, β lon,max is the maximum longitudinal braking acceleration, a lon,max is the maximum longitudinal acceleration.

[0050] In this embodiment, if Figure 4 As shown, the lateral minimum safety distance calculation function d of the lateral safety distance maintenance principle is lat,min for,

[0051] ,

[0052] Among them, v ego,lat is the lateral safety distance based on the lateral speed of the vehicle, v other,lat is the lateral speed of the vehicle, t lat is the minimum lateral reaction time, β lat,min is the minimum lateral vehicle deceleration.

[0053] In this embodiment, the minimum safety distance is unchanged in principle to further constrain the lateral behavior of the decision-making behavior of changing lanes, and a preset longitudinal safety margin γ is set. lat and lateral safety margin γ lon , when the calculated minimum longitudinal safety distance d lon,min ≤γ lat When the lane change decision behavior is evaluated as infeasible; when the lateral minimum safety distance d is calculated lon,min ≤γ lon , the lane-changing decision behavior is evaluated as infeasible.

[0054] In this embodiment, in the rigid fuse evaluation layer, the compliance dimension considers all traffic regulations to evaluate the decision-making behavior, including the compliance of lateral lane change behavior, longitudinal acceleration behavior and longitudinal uniform speed behavior. If the lane line within a certain longitudinal range of the lane where the vehicle is located is a solid line, the left lane line of the target lane for left lane change is a solid line, or the right lane edge line of the target lane for right lane change is a solid line, the corresponding lane change decision behavior is evaluated as infeasible. If the longitudinal behavior is acceleration behavior or uniform speed behavior, but the vehicle speed has reached the maximum speed limit of the vehicle lane or the target lane, the corresponding decision behavior is evaluated as infeasible.

[0055] In this embodiment, a multi-dimensional behavioral decision-making evaluation model designed by the present invention, applicable to multiple driving modes, is used to perform a rigid-fuse evaluation layer on all semantic-level decision-making behaviors. Since the multi-dimensional evaluation includes multiple dimensions, such as safety, compliance, comfort, efficiency, and navigation, to simplify the complexity of the evaluation paradigm, these dimensions are divided into multiple levels, including a rigid-fuse evaluation layer and a weighted-subdivision evaluation layer. The rigid-fuse evaluation layer evaluates the safety and compliance dimensions. If the evaluation does not meet the standards, the decision-making behavior is directly judged as infeasible, and no further weighted-subdivision evaluation is performed.

[0056] At the rigid circuit breaker evaluation layer, the safety dimension designed a responsibility-sensitive strong safety model to evaluate decision-making behavior. The responsibility-sensitive strong safety model has three safety principles: maintaining a longitudinal safe distance, maintaining a lateral safe distance, and maintaining a minimum safe distance without changing lanes.

[0057] Meanwhile, at the rigid fuse evaluation layer, the compliance dimension considers all traffic regulations to evaluate decision-making behaviors, including compliance with lateral lane changes, longitudinal acceleration, and longitudinal constant speed. If the lane line within a certain longitudinal range of the ego vehicle's lane is a solid line, if the left lane line of the target lane is a solid line for a left lane change, or if the right lane edge of the target lane is a solid line for a right lane change, the corresponding lane change decision is considered infeasible. If the longitudinal behavior is acceleration or constant speed, but the ego vehicle's speed has reached the maximum speed limit of the ego vehicle lane or the target lane, the corresponding decision is considered infeasible.

[0058] Example 2: Based on the multi-dimensional evaluation method for decision-making behavior in multiple driving modes in Example 1, the present invention is further illustrated and described below.

[0059] like Figure 1 or Figure 2 As shown, a multi-dimensional evaluation method for decision-making behavior in multiple driving modes is provided, the method comprising:

[0060] U1. The autonomous driving system's behavior decision module generates a comprehensive semantic-level decision behavior set based on temporal, lateral, and longitudinal behavior.

[0061] U2. Input the full set of semantic-level decision-making behaviors into the rigid fuse evaluation layer of the multi-dimensional evaluation model for behavioral decision-making in multiple driving modes for safety and compliance evaluation. If the evaluation fails, the decision-making behavior is directly judged as infeasible. If the evaluation meets the standards, the decision-making behavior is judged as feasible and proceeds to step U3.

[0062] U3. Input the semantic-level decision-making behavior from the rigid fuse evaluation layer into the segmented weighted evaluation layer of the multi-dimensional evaluation model for behavioral decision-making applicable to multiple driving modes. This layer evaluates the efficiency and comfort dimensions and calculates the segmented weighted total cost.

[0063] U4. Traverse the cost of the decision behaviors in the decision behavior group and find the decision behavior execution with the lowest weighted total cost.

[0064] In this embodiment, in step U3, the total efficiency cost function F of the efficiency dimension efficiency for,

[0065] ,

[0066] Among them, F e,v is the cost of the vehicle to reach the desired speed, F e,n is the cost of the ego vehicle reaching the desired driving lane, Δv1 is the speed difference between the ego vehicle exceeding the speed limit and the desired speed, Δv2 is the speed difference between the ego vehicle failing to reach the desired speed and the desired speed, Δv3 is the speed difference between the low speed ahead and the ego vehicle, Δv3 is the speed difference between the high speed ahead and the desired speed of the ego vehicle, y1 is the ego vehicle overspeeding discount factor, y2 is the ego vehicle failing to reach the desired speed discount factor, y3 is the front vehicle blocking the ego vehicle at low speed, y4 is the front vehicle blocking the ego vehicle from reaching the desired speed discount factor, v5 is the lane change speed, y5 is the lane change penalty discount factor, and y6 is the navigation recommended lane change reward discount factor; the total comfort cost function F of the comfort dimension is: comfort for,

[0067] F comfort =F c,lon +F c,lat , where F c,lon is the longitudinal comfort cost of the coarse trajectory in the time domain, F c,lat is the lateral comfort cost of the coarse trajectory in the time domain.

[0068] In this embodiment, the longitudinal comfort cost F of the time domain rough trajectory is c,lon for,

[0069] ,

[0070] The lateral comfort cost F of the time domain rough trajectory c,lat for,

[0071] ,

[0072] Among them, n lat is the number of trajectory points exceeding the index threshold in the horizontal direction, n lon is the number of longitudinal trajectory points exceeding the index threshold, c lat is the lateral comfort threshold, c lon is the longitudinal comfort threshold, and N is the total number of trajectory points.

[0073] In this embodiment, the function F of the weighted total cost of the subdivision cost for,

[0074] F cost =δ efficiency F efficiency +δ comfort F comfort ,

[0075] Among them, δ efficiency is the efficiency weight, δ comfort To adapt to different driving needs of users, a variety of driving modes are designed for comfort weight, including conservative mode, aggressive mode and balanced mode. Based on the driving mode selected by the user, the efficiency weight δ is dynamically adjusted. efficiency and comfort weight δ comfort .

[0076] In this embodiment, the present invention provides a multi-dimensional evaluation system for decision-making behavior in multiple driving modes, including a computer device programmed or configured to execute the steps of any one of the multi-dimensional evaluation methods for decision-making behavior in multiple driving modes.

[0077] In this embodiment, the present invention provides a computer-readable storage medium storing a computer program programmed or configured to execute any one of the multi-dimensional evaluation methods for decision-making behaviors in multiple driving modes.

[0078] Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0079] In summary, the present invention not only solves the problem of inconsistent evaluation paradigms and evaluation dimensions of decision-making behavior evaluation models in multiple scenarios, but also greatly simplifies the complexity of decision-making behavior evaluation and improves the accuracy and efficiency of decision-making behavior evaluation.

[0080] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A multi-dimensional evaluation method for decision-making behavior in multiple driving modes, characterized by: The method comprises: U1. The autonomous driving system's behavior decision module generates a comprehensive semantic-level decision behavior set based on temporal, lateral, and longitudinal behavior. U2. Input the full set of semantic-level decision-making behaviors into the rigid fuse evaluation layer of the multi-dimensional evaluation model for behavioral decision-making in multiple driving modes for safety and compliance evaluation. If the evaluation fails, the decision-making behavior is directly judged as infeasible. If the evaluation meets the standards, the decision-making behavior is judged as feasible and proceeds to step U3. U3. Input the semantic-level decision-making behavior from the rigid fuse evaluation layer into the segmented weighted evaluation layer of the multi-dimensional evaluation model for behavioral decision-making applicable to multiple driving modes. This layer evaluates the efficiency and comfort dimensions and calculates the segmented weighted total cost. U4. Traverse the cost of the decision behaviors in the decision behavior group and find the decision behavior execution with the lowest weighted total cost.

2. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 1 is characterized in that: In step U2, the rigid fuse evaluation layer evaluates the decision-making behavior to build a responsibility-sensitive strong safety model. The responsibility-sensitive strong safety model includes the longitudinal safety distance maintenance principle, the lateral safety distance maintenance principle and the minimum safety distance unchanged lane principle. The longitudinal minimum safety distance calculation function d of the longitudinal safety distance maintenance principle is lon,min for, ， Among them, v f is the longitudinal velocity of the preceding vehicle, v r is the longitudinal speed of the following vehicle, t lon is the longitudinal minimum reaction time, β lon,min is the minimum longitudinal braking acceleration, β lon,max is the maximum longitudinal braking acceleration, a lon,max is the maximum longitudinal acceleration.

3. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 2 is characterized by: The lateral minimum safety distance calculation function d of the lateral safety distance maintenance principle lat,min for, ， Among them, v ego,lat is the lateral safety distance based on the lateral speed of the vehicle, v other,lat is the lateral speed of the vehicle, t lat is the minimum lateral reaction time, β lat,min is the minimum lateral vehicle deceleration.

4. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 2, characterized in that: The principle of the minimum safety distance remains unchanged is a further safety constraint on the lateral behavior of the lane change decision-making behavior, and a preset longitudinal safety margin γ is set. lat and lateral safety margin γ lon , when the calculated minimum longitudinal safety distance d lon,min ≤γ lat When the lane change decision behavior is evaluated as infeasible; when the lateral minimum safety distance d is calculated lon,min ≤γ lon , the lane-changing decision behavior is evaluated as infeasible.

5. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 2 is characterized by: In the rigid fuse evaluation layer, the compliance dimension considers all traffic regulations to evaluate the decision-making behavior, including the compliance of lateral lane change behavior, longitudinal acceleration behavior and longitudinal uniform speed behavior. If the lane line within a certain longitudinal range of the lane where the vehicle is located is a solid line, the left lane line of the target lane for left lane change is a solid line, or the right lane edge line of the target lane for right lane change is a solid line, the corresponding lane change decision behavior is evaluated as infeasible. If the longitudinal behavior is acceleration or uniform speed behavior, but the vehicle speed has reached the maximum speed limit of the vehicle lane or the target lane, the corresponding decision behavior is evaluated as infeasible.

6. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 1 is characterized in that: In step U3, the total efficiency cost function F of the efficiency dimension efficiency for, ， Among them, F e,v is the cost of the vehicle to reach the desired speed, F e,n is the cost of the ego vehicle reaching the desired driving lane, Δv1 is the speed difference between the ego vehicle exceeding the speed limit and the desired speed, Δv2 is the speed difference between the ego vehicle failing to reach the desired speed and the desired speed, Δv3 is the speed difference between the low speed ahead and the ego vehicle, Δv3 is the speed difference between the high speed ahead and the desired speed of the ego vehicle, y1 is the ego vehicle overspeeding discount factor, y2 is the ego vehicle failing to reach the desired speed discount factor, y3 is the front vehicle blocking the ego vehicle at low speed, y4 is the front vehicle blocking the ego vehicle from reaching the desired speed discount factor, v5 is the lane change speed, y5 is the lane change penalty discount factor, and y6 is the navigation recommended lane change reward discount factor; the total comfort cost function F of the comfort dimension is: comfort for, F comfort =F c,lon +F c,lat , where F c,lon is the longitudinal comfort cost of the coarse trajectory in the time domain, F c,lat is the lateral comfort cost of the coarse trajectory in the time domain.

7. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 6 is characterized by: The longitudinal comfort cost F of the time domain rough trajectory c,lon for, ， The lateral comfort cost F of the time domain rough trajectory c,lat for, ， Among them, n lat is the number of trajectory points exceeding the index threshold in the horizontal direction, n lon is the number of longitudinal trajectory points exceeding the index threshold, c lat is the lateral comfort threshold, c lon is the longitudinal comfort threshold, and N is the total number of trajectory points.

8. The multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to claim 6 is characterized by: The weighted total cost of the segment is a function of F cost for, F cost =d efficiency F efficiency +d comfort F comfort , Among them, δ efficiency is the efficiency weight, δ comfort To adapt to different driving needs of users, a variety of driving modes are designed for comfort weight, including conservative mode, aggressive mode and balanced mode. Based on the driving mode selected by the user, the efficiency weight δ is dynamically adjusted. efficiency and comfort weight δ comfort .

9. A multi-dimensional evaluation system for decision-making behavior in multiple driving modes, including a computer device, characterized in that: The computer device is programmed or configured to execute the steps of the multi-dimensional evaluation method for decision-making behavior in multiple driving modes according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program programmed or configured to execute the multi-dimensional evaluation method for decision-making behaviors in multiple driving modes according to any one of claims 1 to 8.

Images