Methods, devices and electronic equipment, and storage media for trajectory risk assessment of autonomous driving

By using a risk assessment method based on expert trajectory data and autonomous vehicle planned trajectory, combined with trajectory prediction results, the risk assessment of the RSS model is dynamically adjusted, which solves the problem of the RSS model being too conservative and limited in autonomous driving, and improves the safety and flexibility of the system.

CN123104634APending Publication Date: 2026-11-17MUSHROOM CHELIAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611267778.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing RSS models may be too conservative in autonomous driving, resulting in vehicles that are not flexible or efficient enough, and they have limitations in complex traffic scenarios, making it difficult to effectively combine them with predictive models.

Method used

By identifying the risk weights of the RSS model based on expert trajectory data, using the self-planned trajectory as a reference coordinate system, and combining the trajectory prediction results to conduct time-series risk assessment, the risk assessment method is dynamically adjusted to improve the accuracy and flexibility of risk assessment.

Benefits of technology

It improves the safety and reliability of autonomous driving systems, enables more accurate assessment of interaction risks with other vehicles, allows for flexible responses to complex traffic situations, and enhances driving safety and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104634A_ABST
    Figure CN123104634A_ABST
Patent Text Reader

Abstract

This application discloses a trajectory risk assessment method, device, electronic device, and storage medium for autonomous driving. The method includes identifying risk weights in the RSS model of the autonomous vehicle based on expert trajectory data to obtain a first risk assessment result; using the planned trajectory of the autonomous vehicle as a reference coordinate system to assess the interaction risk between the autonomous vehicle and other vehicles to obtain a second risk assessment result; performing a temporal risk assessment of the autonomous vehicle's RSS model based on the autonomous vehicle trajectory prediction result to obtain a third risk assessment result; if any of the first, second, or third risk assessment results indicates a collision risk, then the autonomous vehicle is alerted to the risk or an obstacle avoidance action is triggered. This application can improve the safety and reliability of autonomous driving systems.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to a method, device, electronic device, and storage medium for assessing the trajectory risk of autonomous driving. Background Technology

[0002] The RSS model, or "Responsibility Sensitive Safety," is a mathematical framework proposed by Mobileye. It aims to ensure safe decision-making in autonomous vehicles by formalizing safety common sense and responsibility allocation in human driving. Its core objective is to ensure that autonomous vehicles will never intentionally initiate a collision and can react appropriately to potential hazards to avoid accidents.

[0003] The longitudinal risk in the RSS model primarily stems from inconsistent motion between vehicles, such as sudden braking by the vehicle in front. The RSS model mitigates this risk by calculating a minimum safe distance, formalized based on parameters such as the maximum braking acceleration of the vehicle in front, the reaction time of the vehicle behind, and its braking capacity. The lateral risk in the RSS model involves potential conflicts between vehicles in the lateral direction, such as other vehicles suddenly changing lanes or traveling in opposite directions. The RSS model defines a lateral safe distance, calculating the minimum distance to avoid a collision using limiting assumptions (such as two vehicles approaching each other at maximum lateral acceleration and then braking with minimum deceleration). If the lateral distance falls below a threshold, the system may adjust its direction or brake to maintain a safe boundary.

[0004] As can be seen from the above, risk assessment is crucial in autonomous driving. The RSS model provides autonomous vehicles with a clear set of safety guidelines for decision-making and behavior planning, ensuring that the vehicle can take safe actions in various traffic conditions. However, the RSS model may be too conservative, causing autonomous vehicles to act inflexibly or inefficiently in certain situations, even after careful parameter tuning. Summary of the Invention

[0005] This application provides a method, device, electronic device, and storage medium for assessing trajectory risks in autonomous driving systems, thereby improving the safety and reliability of autonomous driving systems.

[0006] The embodiments of this application adopt the following technical solutions:

[0007] In a first aspect, embodiments of this application provide a method for trajectory risk assessment of autonomous driving, characterized in that the method includes:

[0008] Based on expert trajectory data, the risk weights in the vehicle's RSS model are identified to obtain the first risk assessment result;

[0009] Using the vehicle's planned trajectory as a reference coordinate system, a risk assessment of the interaction between the vehicle and other vehicles is conducted, resulting in a second risk assessment result.

[0010] Based on the vehicle trajectory prediction results, a time-series risk assessment of the vehicle's RSS model is performed to obtain the third risk assessment result.

[0011] If any of the first, second, or third risk assessment results indicate a collision risk, the vehicle will be alerted to the risk or an obstacle avoidance maneuver will be triggered.

[0012] In some embodiments, the identification of risk weights in the vehicle RSS model based on expert trajectory data to obtain a first risk assessment result includes:

[0013] Obtain the lateral and longitudinal safe distances of the vehicle during its driving process;

[0014] Determine whether the actual lateral distance and the actual longitudinal distance are less than the corresponding lateral safety distance and the corresponding longitudinal safety distance, respectively.

[0015] If so, then calculate the risk score data for that scenario;

[0016] Based on the safe driving performance of the expert trajectory data under the corresponding risk score data, the lateral safety risk weight and longitudinal safety risk weight in the RSS model are determined by minimizing the risk loss function, so as to obtain the first risk assessment result.

[0017] In some embodiments, the step of using the vehicle's planned trajectory as a reference coordinate system to perform risk assessment and obtain a second risk assessment result includes:

[0018] Establish a coordinate system for the vehicle's planned trajectory with the planned trajectory of the vehicle as the X-axis;

[0019] Transform the motion state parameters of the vehicle and other vehicles from the road coordinate system to the trajectory coordinate system;

[0020] In the trajectory coordinate system, the lateral velocity component of the vehicle is 0;

[0021] The lateral and longitudinal safe distances are calculated based on the transformed motion state parameters of the other vehicle, and it is determined whether the RSS safety constraint is met, so as to obtain the second risk assessment result.

[0022] In some embodiments, the step of using the vehicle's planned trajectory as a reference coordinate system to perform risk assessment and obtain a second risk assessment result includes:

[0023] In the road coordinate system, it is determined that both the lateral RSS constraint and the longitudinal RSS constraint of the vehicle and other vehicles have been violated;

[0024] Switch the coordinate system to the vehicle's planned trajectory coordinate system;

[0025] The risk is recalculated in the coordinate system of the vehicle's planned trajectory. If it is determined that the vehicle still maintains a safety margin in the lateral direction, the second risk assessment result indicates that there is currently no risk of lateral collision.

[0026] In some embodiments, the method further includes:

[0027] Determine whether the distance between your vehicle and other vehicles is less than a preset threshold;

[0028] If so, it belongs to a close-range interaction scenario, and the risk assessment result under the aforementioned vehicle trajectory coordinate system is adopted;

[0029] If not, then it belongs to a long-distance interaction scenario, and the risk assessment result under the road coordinate system is adopted.

[0030] In some embodiments, the temporal risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results to obtain a third risk assessment result includes:

[0031] Obtain the predicted trajectory of other vehicles; if it is predicted that other vehicles will perform a Cut-in action.

[0032] If at the current moment, the RSS model of the vehicle in both the road coordinate system and the vehicle trajectory coordinate system is determined to have violated safety constraints;

[0033] Predict the vehicle's position and speed within a preset time period in the future;

[0034] Based on the predicted state within the preset future time period, a time-series RSS model risk assessment is performed to calculate the collision probability or safe distance change trend at future moments, and a third risk assessment result is obtained.

[0035] In some embodiments, the temporal risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results to obtain a third risk assessment result includes:

[0036] Based on the time-series risk assessment results, the risk values ​​at future moments are weighted according to the time decay coefficient;

[0037] The weighted future risk value is combined with the current risk value to obtain the weighted average risk value;

[0038] Alternatively, the current risk assessment threshold can be dynamically adjusted based on the accumulation of future risks.

[0039] Secondly, embodiments of this application also provide a trajectory risk assessment device for autonomous driving, wherein the device includes:

[0040] The first risk assessment module is used to identify the risk weights in the vehicle's RSS model based on expert trajectory data, and obtain the first risk assessment result.

[0041] The second risk assessment module is used to assess the risk of interaction between the vehicle and other vehicles by using the vehicle's planned trajectory as a reference coordinate system, and obtain the second risk assessment result.

[0042] The third risk assessment module is used to conduct a time-series risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results, and obtain the third risk assessment results.

[0043] The prompting and triggering module is used to prompt the vehicle for a risk or trigger an obstacle avoidance action if any of the first, second or third risk assessment results indicate that there is a collision risk.

[0044] Thirdly, embodiments of this application also provide an electronic device, including: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the above-described method.

[0045] Fourthly, embodiments of this application also provide a computer-readable storage medium that stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform the above-described method.

[0046] The at least one technical solution adopted in this application embodiment can achieve the following beneficial effects: Based on expert trajectory data, the risk weights in the RSS model of the autonomous vehicle are identified to obtain a first risk assessment result; the autonomous vehicle's planned trajectory is used as a reference coordinate system to assess the interaction between the autonomous vehicle and other vehicles, resulting in a second risk assessment result; based on the autonomous vehicle's trajectory prediction results, a temporal risk assessment of the autonomous vehicle's RSS model is performed to obtain a third risk assessment result. After the above risk assessments, if any of the first, second, or third risk assessment results indicates a collision risk, a risk warning is issued to the autonomous vehicle or an obstacle avoidance action is triggered. The method of identifying the weights of lateral and longitudinal safety risks in the RSS model based on expert trajectories in the risk assessment can not only identify the expert's understanding of lateral and longitudinal safety, but also transform this understanding into behavioral guidelines for the autonomous driving system, thereby improving the system's safety and reliability. The second risk assessment method, which combines the lane coordinate system with the vehicle trajectory coordinate system, can provide more accurate and relevant information for autonomous vehicles, thereby improving safety when interacting with other vehicles. The third risk assessment method, which combines the risk assessment results with predictions, combines the RSS model with the prediction model and adopts methods such as dynamic weight adjustment and comprehensive risk assessment. This allows the autonomous driving system to cope with complex traffic situations more flexibly, improving driving safety and efficiency. Attached Figure Description

[0047] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0048] Figure 1 This is a flowchart illustrating the trajectory risk assessment method for autonomous driving in the embodiments of this application;

[0049] Figure 2 This is a schematic diagram of the trajectory risk assessment device for autonomous driving in the embodiments of this application;

[0050] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] The inventors discovered that the RSS model still has the following related problems when used for trajectory risk assessment:

[0053] (1) When the RSS model is used on roads (non-intersection environment), it uses a distance calculation method based on the road coordinate system, namely the Frenet coordinates which are commonly used in autonomous driving. Although the road coordinate system can simplify the problem in some cases, it may have some limitations in complex traffic scenarios and dynamic environments.

[0054] (2) The RSS model tends to make worst-case assumptions about the behavior of other traffic participants, which makes the risk assessment based on RSS tend to be conservative, thus affecting its application in the real world.

[0055] (3) In real-world scenarios, there are often instances where the RSS model's safe distance is violated at a single time point, but these violations are frequently observed in the driving trajectories of human experts. This is because human experts are more accurate in predicting future scenarios. Therefore, there is still a lack of sufficient research on how to combine the RSS model with prediction.

[0056] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0057] This application provides a method for trajectory risk assessment in autonomous driving, such as... Figure 1 The diagram illustrates a flowchart of an autonomous driving trajectory risk assessment method in this application embodiment. The method includes at least the following steps S110 to S140:

[0058] This application is actually divided into multiple risk assessment scenarios and indicates whether there is a collision risk.

[0059] Step S110: Based on expert trajectory data, identify the risk weights in the vehicle's RSS model to obtain the first risk assessment result.

[0060] Expert trajectory data is used to model the importance of lateral and longitudinal safety in the RSS model. Since the risk assessment of the RSS model is calculated separately for lateral and longitudinal risks, it is necessary to estimate the proportion of lateral and longitudinal risks in the overall risk, that is, to identify the risk weights in the RSS model of the vehicle.

[0061] Understandably, expert trajectory data is a high-quality demonstration data in autonomous driving systems. It not only records the spatiotemporal location sequence of moving objects, but also integrates high-level semantic information or correct behavioral demonstrations.

[0062] Step S120: Using the self-planned trajectory as a reference coordinate system, a risk assessment is performed on the interaction between the self-vehicle and other vehicles to obtain a second risk assessment result.

[0063] It should be noted that there is no strict order between steps S120 and S110; they can be executed sequentially or in parallel. This application embodiment does not impose any specific limitations.

[0064] When assessing trajectory risk from the lane coordinate system to the vehicle trajectory coordinate system, the origin / reference benchmark of the coordinate system is bound to the vehicle's planned trajectory (the road that the autonomous vehicle plans to take). In other words, the risk calculation is transformed from the original road coordinate system to the vehicle's planned trajectory coordinate system, which assumes that the vehicle will be on the planned trajectory at future moments, and the lateral speed of the vehicle in the planned trajectory coordinate system is 0.

[0065] Under the vehicle's trajectory coordinate system, the interaction risk between the vehicle and other vehicles is assessed to obtain the second risk assessment result.

[0066] Step S130: Based on the vehicle trajectory prediction results, perform a time-series risk assessment of the vehicle's RSS model to obtain the third risk assessment result.

[0067] Using a combination of prediction and RSS model for time-series safety modeling, if neither of the two assumptions in step S120 or step S110 can yield a risk assessment result, but when another vehicle is suddenly cut-in, the RSS model violates both the horizontal and vertical safety risks, then a risk warning needs to be issued, i.e., whether takeover is necessary needs to be considered.

[0068] Step S140: If any of the first, second, or third risk assessment results indicates a collision risk, then the vehicle is alerted to the risk or an obstacle avoidance action is triggered.

[0069] If any of the above risk assessment results indicate a collision risk, a risk warning will be issued to the vehicle or the vehicle's obstacle avoidance action will be triggered.

[0070] It is understandable that the vehicle's obstacle avoidance actions include emergency braking control: when the system detects a collision risk, it implements deceleration or stopping through the automatic emergency braking (AEB) function to avoid or mitigate the collision.

[0071] It is understandable that the obstacle avoidance actions of a vehicle include: emergency steering control: through the automatic emergency steering (AES) system, the vehicle actively adjusts its direction to avoid obstacles when braking is insufficient or the scene is limited, usually based on model predictive control (MPC) to optimize the path.

[0072] ​​

[0073] Using the above method, the risk weights in the autonomous vehicle's RSS model are identified based on expert trajectory data, yielding the first risk assessment result. The method of identifying the weights of RSS lateral and longitudinal safety risks in risk assessment based on expert trajectories not only identifies the experts' understanding of lateral and longitudinal safety but also transforms this understanding into behavioral guidelines for the autonomous driving system, thereby improving the system's safety and reliability.

[0074] Using the method described above, the vehicle's planned trajectory is used as a reference coordinate system to assess the risk of interactions between the vehicle and other vehicles, resulting in a second risk assessment result. Combining the lane coordinate system with the vehicle's trajectory coordinate system can provide more accurate and relevant information for autonomous vehicles, thereby improving safety during interactions with other vehicles.

[0075] Using the above method, a temporal risk assessment of the vehicle's RSS model is performed based on the vehicle trajectory prediction results, yielding a third risk assessment result. This temporal risk assessment method, combining the RSS model with the prediction model and employing dynamic weight adjustment and comprehensive risk assessment, allows the autonomous driving system to more flexibly cope with complex traffic situations, improving driving safety and efficiency.

[0076] In one embodiment of this application, the step of identifying risk weights in the vehicle RSS model based on expert trajectory data to obtain a first risk assessment result includes: obtaining the lateral and longitudinal safety distances of the vehicle during driving; determining whether the actual lateral and longitudinal distances are less than the corresponding lateral and longitudinal safety distances, respectively; if so, statistically analyzing the risk score data for that scenario; and determining the lateral and longitudinal safety risk weights in the RSS model based on the safe driving performance of the expert trajectory data under the corresponding risk score data, using a minimized risk loss function, to obtain the first risk assessment result.

[0077] Since the risk assessment of the RSS model calculates lateral and longitudinal risks separately, it is necessary to estimate the proportion of lateral and longitudinal risks in the overall risk. It is typically assumed that the vehicle is in a risky state only when both lateral and longitudinal safety are violated. Statistics are compiled on cases where both lateral and longitudinal safety are violated. The performance of expert driving trajectories under different risk scores is analyzed to find the pattern that minimizes the risk score, i.e., the safe driving performance based on the expert trajectory data under the corresponding risk score data. The weights of lateral and longitudinal safety risks in the RSS model are determined using the minimized risk loss function. These patterns are then used as the weights of RSS lateral and longitudinal safety risks in the risk assessment. The weights of RSS lateral and longitudinal safety risks in the risk assessment are identified based on the expert trajectories.

[0078] In one embodiment of this application, the step of using the self-planned trajectory as a reference coordinate system to perform risk assessment and obtain a second risk assessment result includes: establishing a self-planned trajectory coordinate system with the self-planned trajectory as the X-axis; transforming the motion state parameters of the self-car and other vehicles from the road coordinate system to the trajectory coordinate system; in the trajectory coordinate system, the lateral velocity component of the self-car is 0; calculating the lateral and longitudinal safety distances based on the transformed motion state parameters of other vehicles, and determining whether the RSS safety constraint is met, so as to obtain the second risk assessment result.

[0079] In the scenario of a vehicle overtaking another vehicle, under the road coordinate system, both the lateral and longitudinal RSS risks of the vehicle and other vehicles are violated. However, under the vehicle's trajectory coordinate system, even if the other vehicle changes lanes to the left with the maximum rate of curvature change, it can still maintain lateral safety with the vehicle after the reaction time by returning to its original lane. Therefore, in this embodiment, the risk calculation is transformed from the road coordinate system to the planned trajectory coordinate system, that is, it is assumed that at a future moment, the vehicle is on the planned trajectory, and the lateral velocity of the vehicle under the planned trajectory coordinate system is 0.

[0080] In one embodiment of this application, the step of using the vehicle's planned trajectory as a reference coordinate system to perform risk assessment and obtain a second risk assessment result includes: in the road coordinate system, if it is determined that both the lateral RSS constraints and longitudinal RSS constraints of the vehicle and other vehicles have been violated; switching the coordinate system to the vehicle's planned trajectory coordinate system; recalculating the risk in the vehicle's planned trajectory coordinate system; if it is determined that the vehicle still maintains a safety margin in the lateral direction, then the second risk assessment result indicates that there is currently no lateral collision risk.

[0081] When calculating the second risk assessment result, the premise is that, under the road coordinate system, if it is determined that the lateral RSS constraints and longitudinal RSS constraints of the vehicle and other vehicles have been violated, the coordinate system is switched to the vehicle's planned trajectory coordinate system, and the risk is recalculated under the vehicle's planned trajectory coordinate system.

[0082] Specifically, if it is determined that the vehicle still maintains a safety margin in the lateral direction, the second risk assessment result indicates that there is currently no risk of lateral collision. Conversely, if it is determined that there is no risk of lateral collision, then there is no risk of lateral collision.

[0083] In one embodiment of this application, the method further includes: determining whether the distance between the vehicle and other vehicles is less than a preset threshold; if so, it belongs to a close-range interaction scenario, and the risk assessment result under the vehicle trajectory coordinate system is adopted; if not, it belongs to a long-range interaction scenario, and the risk assessment result under the road coordinate system is adopted.

[0084] It is understandable that in close-range interactions, especially in lateral safety risk calculations, we place greater emphasis on risk assessments based on the vehicle's trajectory coordinate system. Similarly, in long-range interactions, especially in longitudinal safety risk calculations, we tend to use risk assessments based on the road coordinate system.

[0085] In one embodiment of this application, the step of performing a temporal risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction result to obtain a third risk assessment result includes: obtaining the predicted trajectory of another vehicle; if it is predicted that the other vehicle will perform a cut-in action; if at the current moment, the RSS model of the vehicle in both the road coordinate system and the vehicle trajectory coordinate system determines that a safety constraint violation has occurred; predicting the position and speed of the vehicle in a future preset time period; performing a temporal RSS model risk assessment based on the predicted state in the future preset time period, calculating the collision probability or safe distance change trend at future moments, and obtaining a third risk assessment result.

[0086] The RSS model pre-calculates the minimum lateral and longitudinal safe distance thresholds that a vehicle must meet to maintain safe driving. When the actual distance between the two vehicles is: actual lateral distance < lateral safe distance threshold and actual longitudinal distance < longitudinal safe distance threshold, both lateral and longitudinal RSS safety constraints are not met simultaneously, it is defined as an RSS risk violation, meaning the safety constraints have been breached and there is a potential collision hazard in the current operating condition.

[0087] The RSS model for the vehicle in both the road coordinate system and the vehicle trajectory coordinate system indicates a safety constraint violation. Specifically, this means that in a scenario where surrounding vehicles are predicted to cut into the vehicle's lane, the following scenarios are observed: 1) Road global coordinate system: Calculating the lateral and longitudinal distances between the vehicle and the cutting vehicle using fixed road coordinates. If both measured lateral and longitudinal distances are less than the corresponding safety threshold of the RSS, the safety constraint fails, resulting in a risk violation; 2) Vehicle trajectory coordinate system: Constructing a local coordinate system based on the vehicle's planned driving route and recalculating the relative positions of the two vehicles. Even with this new coordinate system, the relative lateral and longitudinal distances between the two vehicles still fail to meet the minimum safety distance requirement of the RSS. Independent verification of both coordinate systems shows that both lateral and longitudinal safety conditions are not met, meaning that the RSS is violated in both coordinate systems.

[0088] When the other vehicle's cut-in action occurs, both the lateral and longitudinal safety margins (RSS) have been violated, even within the vehicle's trajectory coordinate system. Due to the existence of prediction, the risk continues to decrease. Based on the predicted trajectory of the other vehicle one second in the future, the RSS model risk for each future time point is calculated, and the future risk is averaged to the current time in an exponentially decreasing manner as the result of the risk assessment.

[0089] In one embodiment of this application, the step of performing a time-series risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results to obtain a third risk assessment result includes: weighting the risk values ​​at future times according to a time decay coefficient based on the time-series risk assessment results; fusing the weighted future risk values ​​with the current risk values ​​to obtain a weighted average risk value; or, dynamically adjusting the current risk judgment threshold based on the accumulation of future risks.

[0090] Based on the time-series risk assessment results, the risk values ​​at future times are weighted according to the time decay coefficient. In this embodiment of the application, a weighted average method is used to fuse the weighted future risk values ​​with the current risk values ​​to obtain a weighted average risk value.

[0091] Preferably, the current risk assessment threshold can be dynamically adjusted based on the accumulation of future risks, i.e., a dynamic risk threshold assessment method can be provided.

[0092] This application also provides an autonomous driving trajectory risk assessment device 200, such as... Figure 2 As shown, a schematic diagram of the structure of an autonomous driving trajectory risk assessment device 200 in this application embodiment is provided. The autonomous driving trajectory risk assessment device 200 includes at least: a first risk assessment module 210, a second risk assessment module 220, a third risk assessment module 230, and a prompting and triggering module 240, wherein:

[0093] In one embodiment of this application, the first risk assessment module 210 is specifically used to: identify the risk weights in the RSS model of the vehicle based on expert trajectory data, and obtain a first risk assessment result.

[0094] Expert trajectory data is used to model the importance of horizontal and vertical safety in the RSS model. Since the risk assessment of the RSS model is calculated separately for horizontal and vertical risks, it is necessary to estimate the proportion of horizontal and vertical risks in the overall risk, that is, to identify the risk weights in the vehicle's RSS model.

[0095] Understandably, expert trajectory data is a high-quality demonstration data in autonomous driving systems. It not only records the spatiotemporal location sequence of moving objects, but also integrates high-level semantic information or correct behavioral demonstrations.

[0096] In one embodiment of this application, the second risk assessment module 220 is specifically used to: use the self-planned trajectory as a reference coordinate system to assess the risk of interaction between the self-vehicle and other vehicles, and obtain a second risk assessment result.

[0097] When assessing trajectory risk from the lane coordinate system to the vehicle trajectory coordinate system, the origin / reference benchmark of the coordinate system is bound to the vehicle's planned trajectory (the road that the autonomous vehicle plans to take). In other words, the risk calculation is transformed from the original road coordinate system to the vehicle's planned trajectory coordinate system, which assumes that the vehicle will be on the planned trajectory at future moments, and the lateral speed of the vehicle in the planned trajectory coordinate system is 0.

[0098] In one embodiment of this application, the third risk assessment module 230 is specifically used to: perform a time-series risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results, and obtain a third risk assessment result.

[0099] Using a combination of prediction and RSS model for time-series safety modeling, if neither of the above two assumptions can yield risk assessment results, but when another vehicle is suddenly cut-in, the RSS model violates both the horizontal and vertical safety risks, then a risk warning needs to be issued, i.e., whether takeover is necessary needs to be considered.

[0100] In one embodiment of this application, the prompting and triggering module 240 is specifically used to: if any one of the first, second or third risk assessment results indicates a collision risk, then prompt the vehicle for a risk or trigger an obstacle avoidance action.

[0101] If any of the above risk assessment results indicate a collision risk, a risk warning will be issued to the vehicle or the vehicle's obstacle avoidance action will be triggered.

[0102] It is understandable that the obstacle avoidance actions of a self-driving vehicle include: Emergency braking control: When the system detects a collision risk, it implements deceleration or stopping through the Automatic Emergency Braking (AEB) function to avoid or mitigate a collision. Emergency steering control: Through the Automatic Emergency Steering (AES) system, the vehicle actively adjusts its direction to avoid obstacles under conditions of insufficient braking or scene limitations, usually based on model predictive control (MPC) to optimize the path. Multi-sensor fusion perception: Utilizing sensors such as cameras, LiDAR, and millimeter-wave radar to detect obstacles in real time, and improving the robustness of detection and reducing false alarms through data fusion. Motion prediction and trajectory planning: Predicting the future trajectories of obstacles and the self-driving vehicle, and generating a collision-free path by combining environmental constraints; common methods include occupancy grid method, potential field method, or intelligent optimization algorithm. Dynamic adjustment and stability control: During obstacle avoidance, the system smoothly adjusts braking or steering force based on road condition estimates, and works with underlying control systems such as ABS and ESC to maintain vehicle stability.

[0103] In one embodiment of this application, the first risk assessment module 210 is further configured to:

[0104] Obtain the lateral and longitudinal safe distances of the vehicle during its driving process;

[0105] Determine whether the actual lateral distance and the actual longitudinal distance are less than the corresponding lateral safety distance and the corresponding longitudinal safety distance, respectively.

[0106] If so, then calculate the risk score data for that scenario;

[0107] Based on the safe driving performance of the expert trajectory data under the corresponding risk score data, the lateral safety risk weight and longitudinal safety risk weight in the RSS model are determined by minimizing the risk loss function, so as to obtain the first risk assessment result.

[0108] In one embodiment of this application, the second risk assessment module 220 is further configured to:

[0109] Establish a coordinate system for the vehicle's planned trajectory with the planned trajectory of the vehicle as the X-axis;

[0110] Transform the motion state parameters of the vehicle and other vehicles from the road coordinate system to the trajectory coordinate system;

[0111] In the trajectory coordinate system, the lateral velocity component of the vehicle is 0;

[0112] The lateral and longitudinal safe distances are calculated based on the transformed motion state parameters of the other vehicle, and it is determined whether the RSS safety constraint is met, so as to obtain the second risk assessment result.

[0113] In one embodiment of this application, the second risk assessment module 220 is further configured to:

[0114] In the road coordinate system, it is determined that both the lateral RSS constraint and the longitudinal RSS constraint of the vehicle and other vehicles have been violated;

[0115] Switch the coordinate system to the vehicle's planned trajectory coordinate system;

[0116] The risk is recalculated in the coordinate system of the vehicle's planned trajectory. If it is determined that the vehicle still maintains a safety margin in the lateral direction, the second risk assessment result indicates that there is currently no risk of lateral collision.

[0117] In one embodiment of this application, the second risk assessment module 220 is further configured to:

[0118] Determine whether the distance between your vehicle and other vehicles is less than a preset threshold;

[0119] If so, it belongs to a close-range interaction scenario, and the risk assessment result under the aforementioned vehicle trajectory coordinate system is adopted;

[0120] If not, then it belongs to a long-distance interaction scenario, and the risk assessment result under the road coordinate system is adopted.

[0121] In one embodiment of this application, the third risk assessment module 230 is further configured to:

[0122] Obtain the predicted trajectory of other vehicles; if it is predicted that other vehicles will perform a Cut-in action.

[0123] If at the current moment, the RSS model of the vehicle in both the road coordinate system and the vehicle trajectory coordinate system is determined to have violated safety constraints;

[0124] Predict the vehicle's position and speed within a preset time period in the future;

[0125] Based on the predicted state within the preset future time period, a time-series RSS model risk assessment is performed to calculate the collision probability or safe distance change trend at future moments, and a third risk assessment result is obtained.

[0126] In one embodiment of this application, the third risk assessment module 230 is further configured to:

[0127] Based on the time-series risk assessment results, the risk values ​​at future moments are weighted according to the time decay coefficient;

[0128] The weighted future risk value is combined with the current risk value to obtain the weighted average risk value;

[0129] Alternatively, the current risk assessment threshold can be dynamically adjusted based on the accumulation of future risks.

[0130] It is understood that the above-mentioned trajectory risk assessment device for autonomous driving can realize each step of the trajectory risk assessment method for autonomous driving provided in the foregoing embodiments. The relevant explanations of the trajectory risk assessment method for autonomous driving are applicable to the trajectory risk assessment device for autonomous driving, and will not be repeated here.

[0131] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Please refer to it. Figure 3 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.

[0132] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0133] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0134] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming a trajectory risk assessment device for autonomous driving at the logical level. The processor executes the program stored in memory and specifically performs the following operations:

[0135] Based on expert trajectory data, the risk weights in the vehicle's RSS model are identified to obtain the first risk assessment result;

[0136] Using the vehicle's planned trajectory as a reference coordinate system, a risk assessment of the interaction between the vehicle and other vehicles is conducted, resulting in a second risk assessment result.

[0137] Based on the vehicle trajectory prediction results, a time-series risk assessment of the vehicle's RSS model is performed to obtain the third risk assessment result.

[0138] If any of the first, second, or third risk assessment results indicate a collision risk, the vehicle will be alerted to the risk or an obstacle avoidance maneuver will be triggered.

[0139] The above is as stated in this application. Figure 1The method executed by the trajectory risk assessment device for autonomous driving disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0140] The electronic device can also perform Figure 1 The method for implementing trajectory risk assessment devices for autonomous driving, and the realization of trajectory risk assessment devices for autonomous driving in... Figure 1 The functions of the embodiments shown are not described in detail here.

[0141] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by an electronic device including multiple applications, enable the electronic device to perform... Figure 1 The method executed by the trajectory risk assessment device for autonomous driving in the illustrated embodiment is specifically used to perform:

[0142] Based on expert trajectory data, the risk weights in the vehicle's RSS model are identified to obtain the first risk assessment result;

[0143] Using the vehicle's planned trajectory as a reference coordinate system, a risk assessment of the interaction between the vehicle and other vehicles is conducted, resulting in a second risk assessment result.

[0144] Based on the vehicle trajectory prediction results, a time-series risk assessment of the vehicle's RSS model is performed to obtain the third risk assessment result.

[0145] If any of the first, second, or third risk assessment results indicate a collision risk, the vehicle will be alerted to the risk or an obstacle avoidance maneuver will be triggered.

[0146] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0147] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0148] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0149] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0150] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0151] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0152] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0153] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0154] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0155] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for trajectory risk assessment in autonomous driving, characterized in that, The method includes: Based on expert trajectory data, the risk weights in the vehicle's RSS model are identified to obtain the first risk assessment result; Using the vehicle's planned trajectory as a reference coordinate system, a risk assessment of the interaction between the vehicle and other vehicles is conducted, resulting in a second risk assessment result. Based on the vehicle trajectory prediction results, a time-series risk assessment of the vehicle's RSS model is performed to obtain the third risk assessment result. If any of the first, second, or third risk assessment results indicate a collision risk, the vehicle will be alerted to the risk or an obstacle avoidance maneuver will be triggered.

2. The method according to claim 1, characterized in that, The process of identifying risk weights in the vehicle RSS model based on expert trajectory data to obtain the first risk assessment result includes: Obtain the lateral and longitudinal safe distances of the vehicle during its driving process; Determine whether the actual lateral distance and the actual longitudinal distance are less than the corresponding lateral safety distance and the corresponding longitudinal safety distance, respectively. If so, then calculate the risk score data for that scenario; Based on the safe driving performance of the expert trajectory data under the corresponding risk score data, the lateral safety risk weight and longitudinal safety risk weight in the RSS model are determined by minimizing the risk loss function, so as to obtain the first risk assessment result.

3. The method according to claim 1, characterized in that, The risk assessment, which uses the vehicle's planned trajectory as a reference coordinate system, yields a second risk assessment result, including: Establish a coordinate system for the vehicle's planned trajectory with the planned trajectory of the vehicle as the X-axis; Transform the motion state parameters of the vehicle and other vehicles from the road coordinate system to the trajectory coordinate system; In the trajectory coordinate system, the lateral velocity component of the vehicle is 0; The lateral and longitudinal safe distances are calculated based on the transformed motion state parameters of the other vehicle, and it is determined whether the RSS safety constraint is met, so as to obtain the second risk assessment result.

4. The method according to claim 3, characterized in that, The risk assessment, which uses the vehicle's planned trajectory as a reference coordinate system, yields a second risk assessment result, including: In the road coordinate system, it is determined that both the lateral RSS constraint and the longitudinal RSS constraint of the vehicle and other vehicles have been violated; Switch the coordinate system to the vehicle's planned trajectory coordinate system; The risk is recalculated in the coordinate system of the vehicle's planned trajectory. If it is determined that the vehicle still maintains a safety margin in the lateral direction, the second risk assessment result indicates that there is currently no risk of lateral collision.

5. The method according to claim 4, characterized in that, The method further includes: Determine whether the distance between your vehicle and other vehicles is less than a preset threshold; If so, it belongs to a close-range interaction scenario, and the risk assessment result under the aforementioned vehicle trajectory coordinate system is adopted; If not, then it belongs to a long-distance interaction scenario, and the risk assessment result under the road coordinate system is adopted.

6. The method according to claim 1, characterized in that, The time-series risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results yields a third risk assessment result, including: Obtain the predicted trajectory of other vehicles; if it is predicted that other vehicles will perform a Cut-in action. If at the current moment, the RSS model of the vehicle in both the road coordinate system and the vehicle trajectory coordinate system is determined to have violated safety constraints; Predict the vehicle's position and speed within a preset time period in the future; Based on the predicted state within the preset future time period, a time-series RSS model risk assessment is performed to calculate the collision probability or safe distance change trend at future moments, and a third risk assessment result is obtained.

7. The method according to claim 6, characterized in that, The time-series risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results yields a third risk assessment result, including: Based on the time-series risk assessment results, the risk values ​​at future moments are weighted according to the time decay coefficient; The weighted future risk value is combined with the current risk value to obtain the weighted average risk value; Alternatively, the current risk assessment threshold can be dynamically adjusted based on the accumulation of future risks.

8. A trajectory risk assessment device for autonomous driving, wherein, The device includes: The first risk assessment module is used to identify the risk weights in the vehicle's RSS model based on expert trajectory data, and obtain the first risk assessment result. The second risk assessment module is used to assess the risk of interaction between the vehicle and other vehicles by using the vehicle's planned trajectory as a reference coordinate system, and obtain the second risk assessment result. The third risk assessment module is used to conduct a time-series risk assessment of the vehicle's RSS model based on the vehicle trajectory prediction results, and obtain the third risk assessment results. The prompting and triggering module is used to prompt the vehicle for a risk or trigger an obstacle avoidance action if any of the first, second or third risk assessment results indicate that there is a collision risk.

9. An electronic device, comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of applications, cause the electronic device to perform the method of any one of claims 1 to 7.