Vehicle control method and device, vehicle and storage medium

Abstract

The embodiment of the invention provides a vehicle control method and device, a vehicle and a storage medium, and the method comprises the steps: carrying out the forward simulation of a plurality of driving strategies of the vehicle based on an environment model of the vehicle, and obtaining a plurality of predicted driving tracks, the forward simulation is used for representing a process of simulating a future driving track of the vehicle; based on the plurality of predicted driving trajectories and first predicted driving trajectories of other vehicles in the environment, the plurality of driving strategies are evaluated, target scoring results of the plurality of driving strategies are obtained, and the target scoring results are used for representing influence degrees of the plurality of driving strategies; selecting a target driving strategy from the plurality of driving strategies based on the target scoring results of the plurality of driving strategies; and controlling the vehicle based on the target driving strategy. According to the invention, the technical problem of low vehicle control accuracy in related technologies is solved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of vehicles, and particularly relate to a vehicle control method and device, a vehicle and a storage medium. BACKGROUND

[0002] Current intelligent driving technologies mainly include rule-based state machines and artificial intelligence-based data-driven methods. The former is not good at handling probabilistic game events, and the latter is flexible but has high computational cost. In particular, in the active lane changing scenario, both methods are difficult to accurately control the speed and position of the vehicle, thereby resulting in low control accuracy of the vehicle in related technologies. SUMMARY

[0003] Embodiments of the present application provide a vehicle control method and device, a vehicle and a storage medium, aiming to improve the technical problem of low control accuracy of the vehicle in related technologies.

[0004] According to an aspect of embodiments of the present application, a vehicle control method is provided, including: performing forward simulation on a plurality of driving strategies of a vehicle based on an environment model of the vehicle, to obtain a plurality of predicted driving trajectories, wherein the environment model is used to represent environmental factors of an environment in which the vehicle is currently located, and the forward simulation is used to represent a process of simulating a future driving trajectory of the vehicle; evaluating the plurality of driving strategies based on the plurality of predicted driving trajectories and a first predicted driving trajectory of another vehicle in the environment, to obtain target score results of the plurality of driving strategies, wherein the target score results are used to represent the influence degree of the plurality of driving strategies; selecting a target driving strategy from the plurality of driving strategies based on the target score results of the plurality of driving strategies, wherein the target driving strategy is a driving strategy with the highest target score result in the plurality of driving strategies; and controlling the vehicle based on the target driving strategy.

[0005] By introducing forward simulation of multiple driving strategies, the comprehensiveness and foresight of vehicle decision-making are enhanced. Then, by scoring each strategy, automatic optimization is achieved, which improves the safety and efficiency of driving. Finally, based on the target driving strategy, control is implemented to ensure the quality and adaptability of driving decisions.

[0006] Furthermore, based on multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, the costs of multiple driving strategies are calculated to obtain target score results for multiple driving strategies. This includes: for a preset driving strategy among the multiple driving strategies, based on the second and first predicted driving trajectories among the multiple predicted driving trajectories, determining the target competition region corresponding to the preset driving strategy; evaluating the preset driving strategy based on the target competition region to obtain a first score result for the preset driving strategy, wherein the second predicted driving trajectory is the predicted driving trajectory corresponding to the preset driving strategy, and the target competition region is used to represent the area where the predicted driving trajectories of the vehicle intersect with those of other vehicles; for other driving strategies among the multiple driving strategies besides the preset driving strategy, evaluating the other driving strategies based on the second and first predicted driving trajectories to obtain a second score result for the other driving strategies; and obtaining the target score result for multiple driving strategies based on the first and second score results.

[0007] By analyzing vehicle behavior within a competitive area, pre-set driving strategies can be quantitatively evaluated, taking into account not only the physical constraints of the trajectory but also the game-theoretic relationships between vehicles. This helps to more accurately predict the effectiveness of strategies in game-theoretic scenarios such as lane changing and overtaking, ensuring that vehicles take the actions most likely to succeed with the lowest risk, and significantly enhancing the control accuracy of vehicles in high-risk scenarios.

[0008] Furthermore, based on the second predicted driving trajectory and the first predicted driving trajectory, the target competition area of ​​the preset driving strategy is determined, including: determining at least one intersection point based on the second predicted driving trajectory and the first predicted driving trajectory; determining the target intersection point of the at least one intersection point based on the intersection time information and intersection position information of the at least one intersection point, wherein the target intersection point is the intersection point with the greatest impact on the vehicle among the at least one intersection point; and constructing the target competition area based on the target intersection point.

[0009] By constructing target competition zones, we can focus on the most pressing points of conflict, thereby allocating computing resources more effectively during strategy evaluation and improving decision-making efficiency.

[0010] Furthermore, based on the target competition area, the preset driving strategy is evaluated to obtain a first score result of the preset driving strategy, including: determining the first time distribution of the vehicle and the second time distribution of other vehicles based on the preset acceleration range and the target competition area, wherein the first time distribution is used to represent the time distribution of the vehicle arriving at the target competition area, and the second time distribution is used to represent the time distribution of other vehicles arriving at the target competition area; the preset driving strategy is evaluated based on the first time distribution and the second time distribution to obtain a first score result of the preset driving strategy.

[0011] By analyzing the time distribution of vehicles and obstacles arriving at the target competition area, a deep evaluation method for preset driving strategies is proposed. This method considers the impact of speed changes on the probability of conflict and can dynamically adjust the strategy evaluation parameters based on vehicle acceleration, thereby more accurately predicting the outcome after strategy implementation. The time distribution-based evaluation mechanism helps to achieve rapid assessment of strategy risks and benefits with limited computing power, improving the reliability and efficiency of strategy selection.

[0012] Furthermore, based on the first time distribution and the second time distribution, the preset driving strategy is evaluated to obtain the first score result of the preset driving strategy, including: transforming the first time distribution to obtain a first normal distribution; transforming the second time distribution to obtain a second normal distribution; and evaluating the preset driving strategy based on the overlapping area of ​​the first normal distribution and the second normal distribution to obtain the first score result of the preset driving strategy.

[0013] By transforming the time distribution into a normal distribution, a simple yet effective strategy evaluation method is introduced. The use of a normal distribution simplifies the computation process, making it more suitable for real-time computing and resource-constrained in-vehicle systems. Analyzing the overlapping region of two normal distributions allows for a direct assessment of the probability of a vehicle encountering another vehicle in a target competition area. This enables adjustments to driving strategies to reduce conflict risk, thereby improving the accuracy and reliability of the evaluation results.

[0014] Further, based on the overlapping region of the first normal distribution and the second normal distribution, the preset driving strategy is evaluated to obtain a first score result of the preset driving strategy, including: if the area of ​​the overlapping region is greater than a first preset value, the current score result of the preset driving strategy is determined as the first score result of the preset driving strategy; if the area of ​​the overlapping region is less than or equal to the first preset value and greater than or equal to a second preset value, the control state of the vehicle is evaluated based on the first acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy; if the area of ​​the overlapping region is less than or equal to the second preset value, the control state of the vehicle is evaluated based on the second acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy, wherein the second acceleration is less than the first acceleration.

[0015] By setting preset values ​​to evaluate driving strategies in a graded manner, intelligent strategy adjustment is achieved. This allows for the automatic selection of appropriate acceleration strategies based on the degree of overlap in competitive areas, enabling rapid response to high-risk situations and the adoption of more economical and comfortable driving modes in medium- and low-risk conditions.

[0016] Furthermore, the method also includes: acquiring environmental perception data of the vehicle and road map data of the road where the vehicle is currently located; and constructing an environmental model based on the environmental perception data and road map data.

[0017] By building an environmental model, not only is the vehicle's understanding of its surroundings enhanced, but also the strategy evaluation becomes more accurate and the decision-making more rational.

[0018] According to another aspect of the embodiments of this application, a vehicle control device is provided, comprising: a simulation module, configured to perform forward simulation on multiple driving strategies of the vehicle based on an environmental model of the vehicle, to obtain multiple predicted driving trajectories, wherein the environmental model is used to represent environmental factors of the current environment of the vehicle, and the forward simulation is used to represent the process of simulating the future driving trajectory of the vehicle; a calculation module, configured to perform cost calculation on the multiple driving strategies based on the multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, to obtain target score results of the multiple driving strategies, wherein the cost calculation is used to quantify the degree of influence of the multiple driving strategies, and the target score results are used to represent the degree of influence of the multiple driving strategies; a selection module, configured to select a target driving strategy from the multiple driving strategies based on the target score results of the multiple driving strategies, wherein the target driving strategy is the driving strategy with the highest target score result among the multiple driving strategies; and a control module, configured to control the vehicle based on the target driving strategy.

[0019] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a memory for storing a computer program; and a processor for executing the program stored in the memory, wherein the program executes the control method of the vehicle when it runs.

[0020] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to execute the above-described vehicle control method. Attached Figure Description

[0021] Figure 1 This is a flowchart of a vehicle control method according to an embodiment of this application;

[0022] Figure 2 This is a framework diagram of a multi-strategy forward simulation and strategy evaluation according to an embodiment of this application;

[0023] Figure 3 This is a flowchart of a right-of-way estimation method according to an embodiment of this application;

[0024] Figure 4 This is a schematic diagram of a vehicle control device according to an embodiment of this application;

[0025] Figure 5 This is a structural diagram of the vehicle provided in the embodiments of this application. Detailed Implementation

[0026] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the following detailed description is provided in conjunction with embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] Terminology Explanation

[0028] Forward simulation: Based on the current vehicle state and environmental information, a series of mathematical models and algorithms are used to predict the vehicle's trajectory over a future period. It is a forward-looking analysis tool that helps vehicles anticipate various situations that may arise when taking a certain action strategy, including potential conflicts with surrounding vehicles or other obstacles, as well as changes in their own trajectory.

[0029] Right-of-way game theory refers to the interaction and decision-making process between two or more vehicles sharing limited road resources in a road traffic environment, based on their respective action strategies and road use rights. In the decision-making logic of autonomous vehicles, the introduction of right-of-way game theory aims to resolve traffic conflicts between vehicles and between vehicles and pedestrians, thereby improving the safety and decision-making efficiency of autonomous driving systems in complex traffic scenarios.

[0030] This application provides a vehicle control method, comprising: performing forward simulation on multiple driving strategies of the vehicle based on an environmental model to obtain multiple predicted driving trajectories, wherein the environmental model represents the environmental factors of the current environment in which the vehicle is located, and the forward simulation represents the process of simulating the future driving trajectory of the vehicle; evaluating the multiple driving strategies based on the multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment to obtain target score results for the multiple driving strategies, wherein the target score results represent the degree of influence of the multiple driving strategies; selecting a target driving strategy from the multiple driving strategies based on the target score results of the multiple driving strategies, wherein the target driving strategy is the driving strategy with the highest target score result among the multiple driving strategies; and controlling the vehicle based on the target driving strategy.

[0031] The vehicle control method provided in this application achieves the following technical effects:

[0032] This application first performs forward simulations on multiple driving strategies based on the vehicle's environmental model, obtaining multiple predicted driving trajectories, thus enabling the determination of the future driving trajectory corresponding to each driving strategy. Subsequently, based on the predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, each driving strategy is quantitatively evaluated, resulting in a target score. This ensures that the vehicle selects the driving strategy most suitable for the current traffic situation. This process avoids the limitations of traditional single-strategy evaluation, making strategy selection more comprehensive and reasonable. Finally, the target driving strategy selected based on the target score is used to control the vehicle, achieving precise driving under complex conditions. This application, by combining forward simulation and multi-strategy evaluation, achieves the goal of pre-evaluating and adjusting vehicle driving strategies, improving the control accuracy of vehicles in complex traffic environments, and thus solving the technical problem of low vehicle control accuracy in related technologies.

[0033] This application provides a vehicle control method. Figure 1 This is a flowchart of a vehicle control method according to an embodiment of this application. Please refer to it. Figure 1 This includes the following steps:

[0034] Step S102: Based on the vehicle's environment model, perform forward simulations on multiple driving strategies of the vehicle to obtain multiple predicted driving trajectories. The environment model is used to represent the environmental factors of the current environment of the vehicle, and the forward simulation is used to represent the process of simulating the future driving trajectory of the vehicle.

[0035] The aforementioned vehicles can refer to vehicles equipped with intelligent driving systems. Vehicle types can include, but are not limited to, passenger cars, commercial vehicles, special vehicles, and other vehicles with autonomous driving functions. The specific vehicle type needs to be determined based on the actual situation. In this application, the vehicle is the executor of active game-theoretic trajectory planning. Its goal is to select appropriate driving trajectories in different driving scenarios based on environmental models and driving strategies to improve lane change success rates and driving safety.

[0036] The aforementioned environmental model refers to a digital representation of static and dynamic elements in the vehicle's operating environment. Types of environmental models include, but are not limited to, map-based models (containing road geometry information), perception-based models (dynamic obstacle positions and speeds), and integrated models (combining map and perception information). The specific environmental model needs to be determined based on actual requirements. Environmental models can provide decision-making support for vehicles, helping them understand and predict changes in the surrounding environment, thereby enabling them to formulate reasonable driving strategies and trajectory planning.

[0037] The aforementioned driving strategies refer to a set of behavioral rules for a vehicle in different driving scenarios. Driving strategies may include, but are not limited to, maintaining lane position, changing lanes, detouring, accelerating, and decelerating; the specific driving strategy needs to be determined based on the actual driving scenario. Different driving strategies correspond to different driving behaviors, and driving strategies can be used to ensure that the vehicle can make safe, legal, and effective decisions in various situations to reach the predetermined destination.

[0038] The aforementioned forward simulation refers to a technique for simulating the future behavior of a vehicle, predicting its trajectory over a future period based on the current vehicle state and the selected driving strategy. Forward simulation can utilize vehicle dynamics models, environmental models, and obstacle behavior models to generate a series of possible future vehicle states, including position, speed, and acceleration. Forward simulation can be used to evaluate the feasibility and safety of different strategies, assisting the system in making decisions. This application uses forward simulation to evaluate vehicle behavior under different driving strategies, combining right-of-way and game-theoretic probabilities to select the most suitable driving trajectory.

[0039] The predicted driving trajectory mentioned above refers to the result of forward simulation, which depicts the vehicle's future position, speed, acceleration, and other motion states under the guidance of a certain strategy. The predicted driving trajectory can include, but is not limited to, straight-line trajectories, curved trajectories, acceleration trajectories, and deceleration trajectories. The specific predicted driving trajectory needs to be determined based on the specific driving strategy and road conditions. The predicted driving trajectory can be used to evaluate the safety and feasibility of the driving strategy.

[0040] Predicting a driving trajectory is more than just a simple path description; it also includes time-related information—when and where the vehicle will arrive. This is crucial for determining the vehicle's relative position and potential conflicts with other road users. By using predicted trajectories obtained through forward simulation, the vehicle's decision-making system can better understand future events and make more efficient driving decisions.

[0041] In one optional embodiment, a vehicle environment model is first constructed. This model integrates real-time data acquired by the vehicle's perception system with prior map information, covering road characteristics, traffic signals, and the status of other dynamic obstacles, accurately reflecting the complex traffic environment in which the vehicle is currently located. Next, based on this environment model, forward simulations are performed on various driving strategies the vehicle might adopt, such as lane keeping, lane changing, or detours. This forward simulation stage essentially simulates the vehicle's future trajectory. By setting different parameters such as acceleration and steering angle, the dynamic changes of the vehicle when executing each strategy are predicted, resulting in a series of predicted trajectories. These predicted trajectories not only include the vehicle's own position and speed information but also consider its relative positional relationships with other traffic participants, providing rich and intuitive data support for subsequent risk assessment, strategy evaluation, and trajectory selection. Through these steps, the possible outcomes of various driving strategies can be fully understood before the vehicle makes actual decisions, allowing for a more scientific and rational selection of the optimal path, ensuring driving safety, comfort, and efficiency. This method is particularly suitable for common traffic congestion and lane-changing scenarios in urban traffic. Through a deep understanding of the environment and accurate prediction of future driving trajectories, it effectively improves the autonomous vehicle's ability to actively engage in complex traffic environments and increases its success rate.

[0042] For example, suppose a vehicle, the owner, is traveling near a complex intersection. Ahead lies a dynamic obstacle, another vehicle, traveling in the same direction. Both vehicles may change lanes to the left into the same target lane to avoid the upcoming traffic light. First, based on the vehicle's perception data and high-precision map information, an environmental model is constructed, including lane markings, traffic light status, and the position and speed of the dynamic obstacle (the other vehicle). This model is generated in the owner's coordinate system, ensuring the accuracy of all environmental factors' position and dynamic information relative to the owner.

[0043] Next, several driving strategies for the vehicle were defined, such as lane keeping (K), lane changing (L), and detour (L-OA). For lane keeping strategy K, the vehicle will stay in the current lane; for lane changing strategy L, the vehicle will attempt to change lanes to the left to enter the target lane; and for detour strategy L-OA, when the lane changing strategy fails, the vehicle will attempt to detour around the dynamic obstacle to continue moving forward.

[0044] Based on these strategies, forward simulation is performed on the vehicle. For example, when executing lane change strategy L, different acceleration sampling values ​​are set, such as within the range of comfortable acceleration and deceleration, to calculate the vehicle's future trajectory under different accelerations. These trajectories will include information such as the vehicle's position, speed, and acceleration to reflect the vehicle's motion state under different strategies. Simultaneously, these trajectories are evaluated to check whether they collide with other obstacles in the environment (such as other vehicles) and whether they violate traffic rules.

[0045] Forward simulation yields a set of predicted driving trajectories for the vehicle under different driving strategies. These trajectories not only reflect the vehicle's future position but also take into account comfort, safety, and adherence to traffic rules, providing rich information for subsequent decision-making. This ensures the vehicle makes appropriate choices when executing driving strategies, thereby improving lane-changing success rates and overall driving safety. This process, by simulating the vehicle's behavior under different strategies, effectively predicts possible driving outcomes, providing a solid foundation for intelligent vehicle decision-making.

[0046] Step S104: Based on multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, multiple driving strategies are evaluated to obtain target score results for multiple driving strategies, wherein the target score results are used to represent the degree of influence of multiple driving strategies.

[0047] The aforementioned first predicted driving trajectory refers to the trajectory formed by predicting the possible future driving paths of other vehicles (i.e., other vehicles) in the current scenario. This prediction is based on information such as the current position, speed, acceleration, and direction of travel of other vehicles, combined with historical behavior patterns and environmental factors (such as traffic rules and road conditions), and is achieved through vehicle dynamics models and behavior prediction models. The first predicted driving trajectory can be used to provide the vehicle with prior knowledge about the dynamics of other vehicles for its own decision-making and planning, helping the vehicle to anticipate possible traffic conditions and thus make safer and more reasonable driving decisions.

[0048] The aforementioned target score can refer to the interaction between the vehicle and other vehicles in the environment (mainly the vehicle corresponding to the first predicted driving trajectory) when the vehicle adopts different driving strategies, as well as the probability and cost of achieving its own driving goal. Factors considered by the scoring system include, but are not limited to, safety (collision avoidance), comfort (smooth driving, avoiding sudden acceleration or deceleration), efficiency (travel time, fuel consumption, etc.), and compliance (following traffic rules). The target score can be used to quantify the advantages and disadvantages of different driving strategies, providing data support for the vehicle to select appropriate strategies to achieve safe, efficient, and comfortable driving.

[0049] In one optional embodiment, multiple driving strategies are quantitatively evaluated based on the vehicle's multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, resulting in a target score for each driving strategy. These scores reflect the safety, efficiency, and comfort of each driving strategy when considering interactions with other vehicles, thus providing a basis for subsequent decision-making. This method allows for the systematic analysis and quantification of the effects of different driving strategies, ensuring that the ultimately selected strategy effectively addresses the challenges of complex traffic environments while meeting the objectives.

[0050] For example, in a typical urban traffic jam lane-changing scenario, the vehicle is in a lane traveling at a medium speed and its goal is to change lanes to an adjacent empty lane to avoid slower-moving vehicles ahead. Currently, several driving strategies for the vehicle have been determined, such as driving strategies 1, 2, and 3.

[0051] Next, based on multiple predicted driving trajectories for each strategy, and other vehicles in the environment, such as the first predicted driving trajectory of the vehicle ahead in the target lane, the multiple driving strategies 1, 2, and 3 are evaluated. That is, the probability of winning or losing in the game with other vehicles when the vehicle uses strategies 1, 2, and 3 is calculated respectively. The calculation of the probability of winning or losing is related to the determination of right-of-way; the party with stronger right-of-way is more likely to successfully change lanes or maintain its lane.

[0052] Finally, based on these win / loss probabilities, a target score is set. For example, if the win / loss probability of Strategy 2 is higher than 0.8, the target score is higher because it means a higher success rate for lane changes. If the win / loss probability of Strategy 2 is lower than 0.2, the target score is lower, and it may be necessary to re-evaluate the strategy or choose Strategy C. If the win / loss probability of Strategy 2 is close to 0.5, the target score may be moderate. In this case, comfort and efficiency need to be further considered, and the acceleration sampling strategy needs to be adjusted.

[0053] The entire process, through quantitative analysis and evaluation, ensures that the intelligent driving system makes more rational decisions during lane changes, taking into account both right-of-way dynamics and driving efficiency and safety. The numerical values ​​presented above are for illustrative purposes only; specific values ​​need to be determined based on actual conditions and are not limited here.

[0054] Step S106: Based on the target score results of multiple driving strategies, select the target driving strategy from the multiple driving strategies, wherein the target driving strategy is the driving strategy with the highest target score result among the multiple driving strategies.

[0055] The aforementioned target driving strategy refers to a strategy selected from multiple driving strategies based on specific scoring criteria. This strategy is determined by evaluating the potential consequences of each strategy's execution. By calculating the target score for each strategy, the strategy that best meets the preset objectives (such as safety, comfort, and efficiency) is identified. The role of the target driving strategy is to guide the intelligent driving system in making appropriate decisions in complex driving environments. In lane-changing scenarios, the target driving strategy can help the vehicle choose an efficient path for lane changes while ensuring safety, avoiding traffic congestion, and considering driving comfort and the driving status of other road users. It can also improve lane-changing success rates, reduce waiting time, and enhance the overall driving experience.

[0056] In one optional embodiment, the driving strategy with the highest target score among multiple driving strategies is determined as the target driving strategy by comparing the target score results of the multiple driving strategies. This process ensures that the selected strategy is superior, enabling efficient road use while guaranteeing driving safety and comfort. Especially in complex lane-changing scenarios, the selection of the target driving strategy can significantly improve the win rate of active game and effectively guide the vehicle to make appropriate dynamic adjustments.

[0057] This score-based selection mechanism not only demonstrates the intelligence and flexibility of the algorithm, but also simplifies the decision-making process and avoids complex improvement calculations in high-dimensional space, thereby reducing the computing power requirements for real-time processing. This allows the method to run stably in resource-constrained vehicle environments, demonstrating good practicality and wide applicability.

[0058] For example, to select a target driving strategy from multiple driving strategies, firstly, based on the vehicle's current state and environmental information, a predefined set of driving strategies, such as lane keeping (K), lane changing (L), and detour (L-OA), is used. Forward simulation is performed on each strategy to generate the corresponding future driving trajectory of the vehicle. Next, these trajectories are evaluated in detail, i.e., a target score is calculated for each trajectory. This score comprehensively considers multiple dimensions such as safety, comfort, and efficiency. Specifically, the target score is a comprehensive evaluation calculated from the cost functions of each dimension, such as the cost of avoiding collisions, the cost of maintaining comfortable driving, and the cost of driving efficiency. Finally, the target score results of all strategies are compared, and the strategy with the highest target score is selected as the target driving strategy. This means that the driving trajectory under this strategy performs best in terms of safety, comfort, and efficiency, and will guide subsequent trajectory adjustments and vehicle control to achieve better driving behavior.

[0059] For example, suppose there are three strategies: Strategy A is to change lanes to the left lane, Strategy B is to maintain the current lane, and Strategy C is to change lanes to the right lane. After forward simulation for each strategy, the target score results are 85 for Strategy A, 78 for Strategy B, and 82 for Strategy C. Based on the target score results, Strategy A has the highest score. Therefore, Strategy A is selected as the target driving strategy, instructing the vehicle to change lanes to the left lane for a better driving experience. This process ensures that the algorithm can make decisions quickly and accurately with limited computing resources, improving the success rate and safety of active lane changing in driving scenarios.

[0060] Step S108: Control the vehicle based on the target driving strategy.

[0061] In one alternative embodiment, vehicle operation is controlled based on a defined target driving strategy, enabling the vehicle to anticipate potential conflict points and take measures in advance to avoid collisions with other vehicles or pedestrians, thereby significantly reducing the risk of accidents, improving road safety and traffic flow, enhancing the riding experience, and improving resource utilization efficiency.

[0062] This application first performs forward simulations on multiple driving strategies based on the vehicle's environmental model, obtaining multiple predicted driving trajectories, thus enabling the determination of the future driving trajectory corresponding to each driving strategy. Subsequently, based on the predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, each driving strategy is quantitatively evaluated, resulting in a target score. This ensures that the vehicle selects the driving strategy most suitable for the current traffic situation. This process avoids the limitations of traditional single-strategy evaluation, making strategy selection more comprehensive and reasonable. Finally, the target driving strategy selected based on the target score is used to control the vehicle, achieving precise driving under complex conditions. This application, by combining forward simulation and multi-strategy evaluation, achieves the goal of pre-evaluating and adjusting vehicle driving strategies, improving the control accuracy of vehicles in complex traffic environments, and thus solving the technical problem of low vehicle control accuracy in related technologies.

[0063] Optionally, step S104 includes: for a preset driving strategy among multiple driving strategies, determining a target competition area corresponding to the preset driving strategy based on a second predicted driving trajectory and a first predicted driving trajectory among multiple predicted driving trajectories; evaluating the preset driving strategy based on the target competition area to obtain a first score result for the preset driving strategy, wherein the second predicted driving trajectory is the predicted driving trajectory corresponding to the preset driving strategy, and the target competition area is used to represent the area where the predicted driving trajectories of the vehicle intersect with those of other vehicles; for other driving strategies besides the preset driving strategy among multiple driving strategies, evaluating the other driving strategies based on the second predicted driving trajectory and the first predicted driving trajectory to obtain a second score result for the other driving strategies; and obtaining a target score result for multiple driving strategies based on the first score result and the second score result.

[0064] The aforementioned preset driving strategy can refer to a series of predefined driving behavior patterns. Preset driving strategies may include, but are not limited to, lane keeping (K), lane changing (L), and detour (L-OA), etc. The specific preset driving strategy needs to be determined according to actual control requirements. The preset driving strategy can provide a behavioral benchmark, enabling the autonomous driving system to select the most appropriate strategy in different driving environments. In this application, the preset driving strategy may refer to a lane changing driving strategy.

[0065] The aforementioned second predicted driving trajectory can refer to the future driving trajectory generated by predicting the vehicle's motion state based on a preset driving strategy. This second predicted driving trajectory can be used to evaluate whether the vehicle's future movement path under the preset driving strategy is safe, comfortable, and efficient.

[0066] The aforementioned target competition area can refer to a specific geographical space, typically located at a point on a road, where the predicted trajectories of multiple vehicles may intersect, potentially leading to traffic conflicts. The target competition area can be determined by analyzing the intersection of the second predicted trajectory (this vehicle) and the first predicted trajectory (the other vehicle), which typically involves mathematical operations and logical judgments regarding the vehicles' predicted paths. Identifying the target competition area can be used to quantify and analyze the game-theoretic relationship between vehicles and other vehicles in lane-changing scenarios, with a particular focus on competition for right-of-way.

[0067] The first rating result mentioned above can refer to a comprehensive evaluation of the safety, comfort, and efficiency of the vehicle's future movement path under a preset driving strategy. The second rating result mentioned above can refer to a comprehensive evaluation of the safety, comfort, and efficiency of the vehicle's future movement path under other driving strategies. The first and second rating results are used to compare the advantages and disadvantages of different driving strategies within the target competition area, helping the system to select the appropriate driving strategy to execute.

[0068] In one optional embodiment, for multiple driving strategies for vehicle driving, the focus is first on a preset driving strategy. By analyzing the vehicle's second predicted driving trajectory and a predetermined first predicted driving trajectory under this strategy, the competition area where the two may intersect is precisely delineated. This area is considered the target competition area, which clarifies the spatial range of potential conflict points. Subsequently, using this target competition area as an evaluation benchmark, the preset driving strategy is deeply evaluated, generating a comprehensive performance index reflecting its performance in collision avoidance, driving efficiency, and comfort—a first score result. Next, for all other driving strategies besides the preset driving strategy, a comparative analysis of the second and first predicted driving trajectories is applied to independently evaluate these strategies, producing their respective second score results, reflecting their performance differences in similar situations. Finally, the first and second score results of all strategies are integrated, and through comparative analysis, the final target score result for each strategy is determined. This series of scores helps the driver or autonomous driving system make appropriate decisions in lane-changing scenarios, ensuring driving safety and efficiency.

[0069] This step takes different driving strategies and their predicted trajectories as input, and outputs a quantitatively evaluated strategy score. The entire process not only enhances the scientific nature of decision-making but also effectively improves the feasibility and success rate of lane-changing operations in complex traffic environments. By meticulously analyzing the dynamic interaction between the vehicle and surrounding obstacles, this method can ensure safety while also considering smooth driving and passenger experience, providing strong support for the proactive game-playing capabilities of intelligent driving systems.

[0070] For example, suppose an autonomous vehicle (autonomous vehicle) is operating on an urban road and faces the need to actively change lanes to improve traffic efficiency. At this time, the autonomous vehicle's decision-making and planning module has identified another vehicle (other vehicle) approaching in an adjacent lane, and the two may intersect in a certain area (hereinafter referred to as the target competition area). In order to more efficiently evaluate the feasibility of changing lanes, the system first calculates the predicted driving trajectory of the autonomous vehicle under the following conditions based on the current driving state and the preset driving strategy: maintaining the current lane (lane keeping strategy K), actively changing lanes to an adjacent lane (lane changing strategy L), and detour strategy (L-OA).

[0071] Next, using the second predicted driving trajectory planned by the autonomous vehicle (assuming it is the driving trajectory under lane change strategy L) and the first predicted driving trajectory of the other vehicle, the area where the driving paths of the two vehicles are about to intersect or have already intersected is calculated as the target competition area. Then, the preset driving strategy is evaluated based on the target competition area to obtain the first score result of the preset driving strategy.

[0072] Similarly, a similar evaluation process is performed for lane-changing strategy L and detour strategy L-OA, calculating the time it takes for the vehicle to reach the target competition area and the probability of obtaining priority right-of-way under these strategies, thus obtaining a second score. This process involves adjusting the sampling of vehicle acceleration under different strategies to ensure the accuracy of the evaluation results.

[0073] Finally, the first and second scoring results are comprehensively considered, and a target score is obtained for all driving strategies through weighted averaging or other suitable methods. The scoring criteria not only include the probability of success in right-of-way negotiations but also cover multiple dimensions such as driving comfort, safety, and efficiency. Based on this comprehensive score, the decision-making module will select the driving strategy with the highest score and execute the corresponding lane-changing action. In this embodiment, if strategy L has the highest score, it indicates that changing lanes is the best choice, and the system will execute the lane-changing operation while ensuring the safety and rationality of the operation.

[0074] This process enables rapid and accurate evaluation of various possible driving decisions under limited computing power, especially in high-density traffic environments, significantly improving the success rate of active lane changes while ensuring driving safety and experience. This approach allows for lightweight decision-making and planning, better adapting to the actual needs of modern intelligent driving systems.

[0075] Optionally, based on the second predicted driving trajectory and the first predicted driving trajectory, the target competition area of ​​the preset driving strategy is determined, including: determining at least one intersection point based on the second predicted driving trajectory and the first predicted driving trajectory; determining the target intersection point of the at least one intersection point based on the intersection time information and intersection position information of the at least one intersection point, wherein the target intersection point is the intersection point with the greatest impact on the vehicle among the at least one intersection point; and constructing the target competition area based on the target intersection point.

[0076] The aforementioned intersection point can refer to a point in space where the predicted driving trajectory of the vehicle intersects with the predicted driving trajectory of another vehicle at some future point in time. Intersection points can be used to identify potential collision risk points and are the basis for lane-change game analysis.

[0077] The aforementioned intersection time information refers to the specific time at which the predicted trajectories of the two vehicles meet at the intersection point. This intersection time information can be a specific instant or a time interval; the specific intersection time information needs to be determined based on prediction accuracy and data availability. Intersection time information can be used to assess the probability of both vehicles reaching the intersection point at a given time, thereby determining right-of-way and the outcome of the game.

[0078] The aforementioned intersection location information refers to the specific location of the intersection point within the road coordinate system. Intersection location information can be on a straight road, or at a curve, intersection, exit, etc. Different intersection locations have different impacts on lane-changing decisions, and the specific intersection location information needs to be determined based on the actual situation. Intersection location information is used to determine whether the intersection point is located in a critical driving area, such as a lane merging point or a crossroads; intersection events at these locations often require more careful handling.

[0079] The aforementioned target intersection point can refer to one or more points selected from at least one intersection point that have the greatest impact on the vehicle's lane-changing decision. The target intersection point can be determined based on various factors, such as its distance from the vehicle's current position, the urgency of the intersection time, and the complexity of the road structure where the intersection is located. Specific determining factors need to be determined according to actual needs. Selecting the target intersection point helps the system focus its attention and computing resources on analyzing and processing events most likely to affect the success rate of lane changes.

[0080] In one alternative embodiment, based on a second predicted driving trajectory and a first predicted driving trajectory, all possible intersections, i.e., at least one intersection, are analyzed and identified. These points signify future collision risks or interaction opportunities. Then, the intersection time information and intersection location information of each of the at least one intersection are comprehensively considered. The intersection time information reflects the estimated time of arrival at the intersection for the vehicle and other vehicles, while the intersection location information indicates the exact location of the intersection on the road. By analyzing this information, it is determined which intersections have the greatest impact on the vehicle's driving strategy, thereby identifying the target intersection. This concept emphasizes selecting the most critical one among many interaction points, providing a more accurate basis for subsequent decision-making.

[0081] Finally, based on the target intersection, a target competition zone is constructed. This target competition zone is a specific geographical area encompassing the target intersection and its surrounding areas that may affect lane-changing safety and efficiency. Determining the target competition zone provides autonomous vehicles with precise spatial coordinates for decision-making, enabling them to adjust driving strategies accordingly, such as gaining right-of-way through speed control or avoiding dangerous areas. This series of steps effectively transforms complex road conditions and vehicle dynamics into actionable decision-making information, enhancing the autonomous vehicle's ability to navigate lane-changing scenarios and ensuring safe and smooth driving.

[0082] For example, imagine an autonomous vehicle (the self-driving car) driving on a busy city road, aiming to safely and efficiently switch to the adjacent lane on the right. First, its future trajectory is predicted based on the self-driving car's lane-keeping strategy, i.e., the second predicted trajectory. At the same time, by analyzing the driving patterns of surrounding vehicles, especially vehicles approaching in the right lane (other vehicles), the trajectory of other vehicles is predicted, which is considered the first predicted trajectory.

[0083] Next, the two sets of predicted trajectories are compared and analyzed to identify all possible intersections, i.e., at least one intersection. These intersections may be located at lane merging points, intersections, or any point where the paths of the two vehicles may overlap. For example, if another vehicle is accelerating to merge into the lane of the first vehicle, then the intersection of the two vehicles at the lane merging point in the next few seconds constitutes an intersection.

[0084] Subsequently, based on the intersection time information (i.e., the estimated time for both vehicles to arrive at the point) and intersection location information (i.e., the coordinates of the point on the road) at each intersection, the system determines which intersections have the greatest impact on the vehicle's lane-changing strategy; these are the target intersections. For example, if at a lane merging point, the vehicle expects to meet another vehicle in 3 seconds, and this point happens to be located on the vehicle's planned lane-changing path, then this intersection becomes the target intersection because it is directly related to whether the lane change will be successful.

[0085] Finally, a target competition zone is constructed around the target intersection. This zone includes not only the intersection itself but also the distances in front and behind it sufficient to influence lane-changing decisions. For example, the target competition zone might extend from 10 meters in front of the vehicle to 20 meters behind it. Within this zone, the right-of-way dynamics between the vehicle and other vehicles become crucial for decision-making. Through forward simulation analysis, the system can comprehensively consider the vehicle's speed adjustment, relative position to other vehicles, and time difference to evaluate the safety and success rate under different driving strategies. Ultimately, it selects the appropriate strategy to ensure successful lane changing within the target competition zone while avoiding potential collision risks.

[0086] This process combines predictive analytics, a comprehensive consideration of temporal and spatial information, and lightweight decision adjustments, effectively improving the lane-changing capabilities and safety of autonomous vehicles in complex traffic environments. The numerical values ​​presented above are for illustrative purposes only; specific values ​​should be determined based on actual conditions and are not limited here.

[0087] Optionally, based on the target competition area, the preset driving strategy is evaluated to obtain a first score result of the preset driving strategy, including: determining a first time distribution of the vehicle and a second time distribution of other vehicles based on the preset acceleration range and the target competition area, wherein the first time distribution is used to represent the time distribution of the vehicle arriving at the target competition area, and the second time distribution is used to represent the time distribution of other vehicles arriving at the target competition area; and evaluating the preset driving strategy based on the first time distribution and the second time distribution to obtain a first score result of the preset driving strategy.

[0088] The aforementioned preset acceleration range refers to a range of possible acceleration values ​​used when evaluating lane-changing strategies. These acceleration values ​​cover a continuous interval from minimum to maximum comfort acceleration. For example, the system might set a preset acceleration range from -2 m / s² (maximum comfort limit for deceleration) to 3 m / s² (maximum comfort limit for acceleration) to generate different speed sampling strategies. The specific preset acceleration range needs to be determined according to actual needs and is not limited here. The preset acceleration range can be used to allow the system to simulate the vehicle's driving state under different acceleration conditions, thereby better evaluating the vehicle's dynamic response and right-of-way acquisition capabilities within the target competition area, providing diverse references for decision-making.

[0089] The aforementioned first time distribution refers to the possible time distribution of the vehicle's arrival at the target competition area, obtained through forward simulation based on a preset acceleration range. It depicts a probability density function, expressing the probability that the vehicle might reach the target competition area under different acceleration strategies. The first time distribution can be a normal distribution, a uniform distribution, or any other probability distribution suitable for describing time uncertainty; the specific first time distribution needs to be determined according to actual needs. In this application, a normal distribution can be used. The first time distribution can provide a quantitative basis for evaluating right-of-way games under vehicle lane-changing strategies. By analyzing the vehicle's time distribution within the target competition area, the probability of the vehicle obtaining priority right-of-way under different driving strategies can be understood, thereby adjusting decisions to adjust lane-changing operations.

[0090] The aforementioned second time distribution can refer to the expected time distribution of other vehicles (such as other cars) arriving at the target competition area, obtained through predictive analysis. Similar to the first time distribution, the second time distribution is also a probability density function, showing the probability of other cars reaching the target competition area at any given time. The second time distribution can be a normal distribution, a uniform distribution, or any other probability distribution suitable for describing time uncertainty; the specific second time distribution needs to be determined according to actual needs. In this application, a normal distribution can be used. The second time distribution helps the system understand the dynamics of other cars within the target competition area, thereby determining possible collisions or intersections between the system and other cars. By comparing the first and second time distributions, the system can estimate the outcome of the game when the two cars meet in the competition area.

[0091] In one optional embodiment, when planning the lane-changing strategy for an autonomous vehicle, an acceleration range is first pre-defined, i.e., a preset acceleration range, which includes the acceleration and deceleration range of the vehicle under comfort constraints. Next, using this preset acceleration range, the time it takes for the vehicle to reach the previously defined target competition area is simulated and calculated, thus obtaining the vehicle's first time distribution, which is a probability distribution describing the possible arrival times of the vehicle in the competition area. Simultaneously, the driving predictions of other surrounding vehicles (such as other vehicles) are analyzed to estimate their second time distribution for arriving at the same target competition area. This distribution also reflects the possible time distribution of other vehicles.

[0092] Subsequently, the first and second time distributions are compared and analyzed. By calculating the relative relationship between the two distributions, such as the degree of overlap, the difference in medians, or the average arrival time, the relative position and right-of-way status of the vehicle and other vehicles within the target competition area are assessed. This assessment is essentially a game theory analysis, aiming to determine, through probabilistic modeling, whether the vehicle can obtain priority right-of-way in the target competition area when executing specific preset driving strategies (such as lane keeping, accelerating to overtake, or slowing down to yield), or what strategy should be adopted to maximize the success rate of lane changing.

[0093] Based on the results of this game theory analysis, a first score for the preset driving strategy is generated. This score not only reflects the feasibility of the lane-changing strategy within the target competition area but also comprehensively considers the strategy's safety, comfort, and efficiency, providing a quantitative basis for determining the appropriate lane-changing plan. The entire process achieves lightweight decision-making and planning by establishing a probabilistic model between vehicle dynamics and the predicted behavior of surrounding traffic participants. Especially in urban congestion environments, it can significantly improve the accuracy of lane-changing decisions and vehicle traffic efficiency.

[0094] Optionally, the preset driving strategy is evaluated based on the first time distribution and the second time distribution to obtain a first score result of the preset driving strategy, including: transforming the first time distribution to obtain a first normal distribution; transforming the second time distribution to obtain a second normal distribution; and evaluating the preset driving strategy based on the overlapping area of ​​the first normal distribution and the second normal distribution to obtain a first score result of the preset driving strategy.

[0095] The aforementioned first normal distribution can refer to the normal distribution form of the first time distribution. The first normal distribution can be used to describe the probabilistic characteristics of the time it takes for a vehicle to reach a specified area under different acceleration strategies. Its center position (mean) and diffusion degree (standard deviation) reflect the dynamic performance and uncertainty of the vehicle.

[0096] The aforementioned second normal distribution can refer to the normal distribution form of the second time distribution. The second normal distribution can be used to represent the time probability of another vehicle being within the target area.

[0097] Normal distribution transformation standardizes what might otherwise be an irregular time distribution, facilitating mathematical operations and probability calculations. By comparing the first and second normal distributions, the temporal relationships between the driver and other vehicles can be intuitively analyzed, providing a probabilistic basis for decision-making. These distribution comparisons help assess the right-of-way situation and lane-changing probability within a target competitive area. For example, when the mean of the normal distribution of the driver's arrival time is less than that of other vehicles, it means the driver has a time advantage in the game; conversely, a greater mean means a greater time advantage indicates a disadvantage.

[0098] The aforementioned overlapping region can refer to the portion where the first normal distribution and the second normal distribution intersect on the time axis, that is, the area within which both the driver and other vehicles may arrive at the target competition area within the same time period. The overlapping region can include, but is not limited to, completely non-overlapping (i.e., no time conflict), partially overlapping (indicating the possibility of a game within a limited time window), and completely overlapping (high-risk scenarios requiring urgent decision-making), etc. The specific overlapping region needs to be determined based on the actual situation. By calculating the area or degree of overlap of the overlapping region, the system can estimate the probability of a conflict between the driver and other vehicles when executing a specific driving strategy, thereby guiding the driver to adjust its strategy to improve the success rate of lane changes.

[0099] In one optional embodiment, the first time distribution of the vehicle's (i.e., the self-vehicle) expected arrival time in the area is first converted into a normal distribution, namely the first normal distribution. This conversion is based on the vehicle's dynamic characteristics and a defined acceleration range. The normal distribution clearly shows the approximate arrival time of the self-vehicle in the target area and its probability distribution over time, including the average arrival time and temporal dispersion. Simultaneously, a similar process is applied to the second time distribution of other vehicles (other vehicles) arriving in the same target area, converting it into a second normal distribution. The second normal distribution also describes a set of probabilistic characteristics, indicating the possible arrival times of other vehicles in the target area and the concentration of their distribution.

[0100] Next, the overlapping region between the first and second normal distributions is analyzed. This overlapping region represents the common probability interval between the autonomous vehicle and other vehicles that might arrive at the target area within the same time window. By calculating and analyzing this overlapping region, the preset driving strategy is comprehensively evaluated, generating a first score. This score integrates the safety, feasibility, and efficiency of the strategy, specifically considering adjusting acceleration to reduce or increase the overlapping region to improve the autonomous vehicle's chances of gaining priority right-of-way in the target area. The score intuitively reflects the expected performance of the autonomous vehicle in lane-changing games under a specific driving strategy, thereby assisting the system in making appropriate choices among various strategy options, ensuring that the vehicle can improve lane-changing success rate and driving efficiency while maintaining safety. This process simplifies complex vehicle dynamics and game theory into mathematical operations based on normal distribution, greatly enhancing the decision-making ability and reaction speed of the autonomous driving system in real-world traffic environments.

[0101] For example, imagine an autonomous vehicle driving on an urban expressway, planning to change lanes from the left lane to the right lane to avoid a slow-moving vehicle ahead. The system first uses forward simulation to calculate the time distribution of the vehicle's arrival at the target competition zone (here, the merging point of the two lanes) based on a preset acceleration range (e.g., from a comfortable deceleration of -3 m / s² to a comfortable acceleration of 3 m / s²), i.e., the first time distribution.

[0102] To perform more precise analysis, the initial time distribution is then transformed into a normal distribution, known as the first normal distribution. Assume the parameters of the transformed normal distribution are a mean μ1 = 5 seconds (indicating the vehicle is most likely to arrive at the competition zone in 5 seconds) and a standard deviation σ1 = 1 second (reflecting the volatility of the time prediction). This means the vehicle has a high probability of arriving at the target area between 4 and 6 seconds.

[0103] Meanwhile, a similar prediction is made for a potential other vehicle in the right lane, yielding its arrival time distribution, known as the second time distribution, which is then transformed into a normal distribution, called the second normal distribution. Assuming this distribution has a mean of μ² = 5.5 seconds and a standard deviation of σ² = 1 second, it indicates that the other vehicle may arrive slightly later, but still within 4.5 to 6.5 seconds of the same area.

[0104] Subsequently, the overlapping region between the first and second normal distributions is analyzed. In this example, the overlapping region is between 4.5 and 6 seconds, which is the time interval during which both the driver and other vehicles may simultaneously appear in the target competition area. The system's evaluation strategy considers the size of this overlapping region, as well as the relative positions of the driver and other vehicles upon arrival, to calculate the success rate and risk of the preset driving strategy.

[0105] Ultimately, based on the above assessment, a first score is generated. This score reflects not only the safety and efficiency of the preset driving strategy but also its chances of success in right-of-way negotiations. For example, if the adjusted acceleration strategy can significantly reduce the overlapping area, the score will be higher, indicating that the strategy is more conducive to successful lane changes while ensuring safety. In this way, driving strategies can be intelligently adjusted according to real-time traffic conditions, improving decision-making quality and lane-change success rate in complex traffic environments. The above values ​​are for illustrative purposes only; specific values ​​need to be determined based on actual conditions and are not limited here.

[0106] Optionally, the preset driving strategy is evaluated based on the overlapping region of the first normal distribution and the second normal distribution to obtain a first score result of the preset driving strategy, including: if the area of ​​the overlapping region is greater than a first preset value, the current score result of the preset driving strategy is determined as the first score result of the preset driving strategy; if the area of ​​the overlapping region is less than or equal to the first preset value and greater than or equal to a second preset value, the control state of the vehicle is evaluated based on the first acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy; if the area of ​​the overlapping region is less than or equal to the second preset value, the control state of the vehicle is evaluated based on the second acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy, wherein the second acceleration is less than the first acceleration.

[0107] The aforementioned first preset value can refer to a predefined overlap area threshold. This first preset value can be used to assess whether the overlap area between the first and second normal distributions reaches a level requiring an urgent adjustment of the driving strategy. Typically, a higher first preset value indicates a higher risk of time overlap or high competitive pressure in lane-changing games. In this application, the first preset value can be set to 0.8, 0.9, 1, etc., and the specific first preset value needs to be determined based on actual requirements.

[0108] The aforementioned second preset value can refer to another predefined overlap area threshold. Compared to the first preset value, the second preset value is lower and is used to distinguish between medium-risk and low-risk situations. When the overlap area is lower than the first preset value but overlap still exists, a more detailed assessment and corresponding fine-tuning of the driving strategy are initiated. In this application, the second preset value can be set to 0.55, 0.5, 0.45, etc., and the specific second preset value needs to be determined according to actual needs.

[0109] The first and second preset values ​​are set based on a variety of factors, including but not limited to vehicle speed, road conditions, traffic flow, and safety standards. These preset values ​​may vary in different application scenarios and environments, and may even be dynamically adjusted to adapt to constantly changing traffic conditions.

[0110] The current rating result mentioned above may refer to the result obtained during the initial assessment of the driving strategy, reflecting the immediate risk level based on the area of ​​the overlapping region. It is directly related to the size of the overlapping region and a first preset value, used to quickly determine whether immediate action is needed to adjust the strategy.

[0111] The aforementioned first acceleration can refer to the acceleration adjustment value used to quickly change the vehicle's state and reduce time overlap with other vehicles when facing moderate risk (i.e., the area of ​​the overlapping region is less than or equal to a first preset value and greater than or equal to a second preset value). This acceleration is usually set as an acceleration value to ensure immediate response. The first acceleration can be used to minimize the overlap time with other vehicles in the target competition area by rapidly adjusting the vehicle speed, such as by significantly accelerating, when the system detects a moderate-risk lane-changing scenario, thereby improving the safety of lane-changing operations. This adjustment strategy is usually activated in emergency situations to respond immediately to changes in the external environment and avoid collisions. In this application, the first acceleration can be an acceleration of 1 m / s², but the specific first acceleration needs to be determined according to actual control requirements and is not limited here.

[0112] The aforementioned second acceleration can refer to the acceleration used when the overlapping area is at a low risk level (i.e., less than or equal to a second preset value). Its value is less than the first acceleration, aiming to adjust the vehicle's driving state in a gentler manner, thereby avoiding potential collision risks without affecting driving comfort and efficiency. The second acceleration can be used in low-risk scenarios to maintain fine-tuning of the vehicle's state, preventing accidental collisions caused by low-probability events. By using a smaller acceleration value, the system improves safety while avoiding aggressive driving operations, ensuring smooth vehicle operation and a good passenger experience. In this application, the second acceleration can be a deceleration of 1 m / s², but the specific second acceleration needs to be determined according to actual control requirements and is not limited here.

[0113] In one optional embodiment, firstly, the area of ​​the overlapping region between the first normal distribution (the vehicle) and the second normal distribution (other vehicles) is determined. This area directly reflects the probability that the vehicle and other vehicles will appear in the competing area at the same time, i.e., the risk level in the lane-changing game. If the area of ​​the overlapping region exceeds a first preset value, it indicates that the lane-changing operation faces a high risk. The system will no longer further evaluate the impact of acceleration adjustment, but will directly determine the current score result of the preset driving strategy as its first score result, triggering an emergency strategy adjustment to ensure the safety of vehicle driving.

[0114] Next, when the overlapping area shrinks to below the first preset value but remains above the second preset value, it indicates that the lane change risk is at a moderate level. In this case, a first acceleration is introduced for a more detailed evaluation. That is, by simulating the vehicle's control state under the first acceleration, the first score result of the preset driving strategy is recalculated. The selection of the first acceleration reflects the system's intention to avoid collisions while maintaining driving comfort and efficiency as much as possible by moderately adjusting the vehicle speed.

[0115] Finally, if the overlapping area further decreases, falling to or below the second preset value, it indicates that the risk of lane-changing has been reduced to a low level. At this point, the driving strategy is re-evaluated using a second acceleration in a more conservative manner, yielding the final first score result. The value of the second acceleration is smaller than that of the first acceleration, intended to fine-tune the vehicle's state, ensuring a smooth transition even in low-risk environments, avoiding unnecessary acceleration and deceleration, and maintaining a good road driving experience.

[0116] The entire process demonstrates the flexibility and hierarchical progression of the autonomous driving system's decision-making. Based on lane-changing scenarios with different risk levels, the system adjusts the vehicle's driving strategy in a timely and appropriate manner, effectively balancing the relationship between driving safety, comfort, and efficiency, and ensuring that intelligent driving vehicles can still make reasonable and safe driving decisions in complex and ever-changing road environments.

[0117] For example, suppose that when evaluating the game state during a lane change, the area of ​​the overlapping region is found to be A square units. The first preset value is known to be 0.5 square units, and the second preset value is known to be 0.1 square units.

[0118] If the overlapping area A square units is greater than the first preset value of 0.5 square units, it indicates a high risk of collision, and the current score of the preset driving strategy is determined as the first score of the preset driving strategy. If the overlapping area A square units is less than or equal to the first preset value of 0.5 square units, but greater than or equal to the second preset value of 0.1 square units, the system will perform a more refined driving strategy evaluation based on the first acceleration (e.g., +1 m / s²), adjusting the vehicle speed to reduce the risk of collision. If the overlapping area A square units is less than or equal to the second preset value of 0.1 square units, it indicates an extremely low risk of collision, and the second acceleration (e.g., -1 m / s²) is activated to make smaller adjustments to the vehicle status to maintain smooth driving and passenger experience.

[0119] This tiered acceleration adjustment mechanism enables autonomous driving systems to not only effectively respond to emergencies but also maintain good driving quality under normal circumstances, making it a crucial component of intelligent vehicle decision-making and planning. The values ​​above are for illustrative purposes only; specific values ​​need to be determined based on actual conditions and are not limited here.

[0120] Optionally, the method further includes: acquiring environmental perception data of the vehicle and road map data of the road where the vehicle is currently located; and constructing an environmental model based on the environmental perception data and the road map data.

[0121] The aforementioned environmental perception data refers to information collected by onboard sensors (such as radar, lidar, and cameras) to depict the real-time state of the vehicle's surroundings. This data includes, but is not limited to, the position, speed, and direction of other vehicles on the road, the presence of pedestrians, cyclists, static obstacles (such as trees and curbs), and the position of dynamic obstacles (such as construction zones and temporary traffic signs). Specific environmental perception data needs to be determined based on the actual environment. Environmental perception data enables vehicles to monitor surrounding traffic participants and obstacles in real time, thereby predicting their movement trends and identifying potential risks, which is crucial for safe driving and effective decision-making.

[0122] The aforementioned road map data refers to map data of the roads the vehicle travels on. Road map data may include, but is not limited to, static map information, dynamic map information, rule information, and auxiliary information. The specific road map data needs to be determined based on actual needs. Road map data can be used to provide autonomous vehicles with information about road infrastructure and layout, helping the vehicle understand its current environment, enabling it to navigate correctly at complex intersections and make relevant driving decisions.

[0123] In one optional embodiment, environmental perception data from vehicle-mounted sensors is first collected, and detailed map data of the current road, i.e., road map data, is simultaneously retrieved. Based on the environmental perception data and road map data, a comprehensive model is created that includes the vehicle's current position, the target driving path, and the relative positions and motion states of surrounding traffic participants. This is the environmental model. This environmental model not only helps the vehicle understand and predict changes in the surrounding environment but also provides the vehicle with the information needed to formulate safe and efficient path planning, ensuring that autonomous vehicles can make informed decisions, such as when to change lanes, which driving direction to follow, and how to respond to unexpected traffic situations, thereby achieving a safe and smooth driving experience.

[0124] In one alternative embodiment, Figure 2 This is a framework diagram of multi-policy forward simulation and policy evaluation according to an embodiment of this application, such as... Figure 2 As shown, the framework mainly consists of three modules: environment modeling and pre-decision making, multi-strategy forward simulation, and strategy evaluation and trajectory selection. After environment modeling and pre-decision making, a coarse screening of the environment and obstacles is obtained; then, multi-strategy forward simulation is performed to obtain the strategy sequence and cost; finally, strategy evaluation and trajectory selection are performed.

[0125] Specifically, environmental modeling and pre-decision making involves combining perception and prior map results to generate the environment in the vehicle's coordinate system, including lane lines, traffic lights, and dynamic obstacles, and then making decisions about the decision space and key obstacles based on the vehicle's driving direction. Multi-strategy forward simulation involves performing forward simulations for each driving strategy defined by prior rules, such as lane keeping (K), lane changing (L), and detour (L-OA), calculating the future driving trajectory and corresponding obstacle behavior for each strategy. Strategy evaluation and trajectory selection involves calculating the cost and overall score of the forward simulation trajectories under different driving strategies, selecting safe, comfortable, and efficient trajectories, and then providing these to downstream trajectory improvement and control.

[0126] Figure 3 This is a flowchart of a right-of-way estimation method according to an embodiment of this application, such as... Figure 3 As shown, the process begins by calculating the competition zone; then, the time distribution of the vehicle and other vehicles to the competition zone is calculated; next, the outcome of the game is calculated; the acceleration sampling of the current strategy is adjusted according to the game outcome; the vehicle's speed and position are adjusted; and finally, the feasibility and success rate of the candidate strategy L are changed.

[0127] Specifically, in step one, the competition target area is calculated: the competition area is calculated based on the collision position of the vehicle's planned trajectory in the previous frame and the predicted trajectory of other vehicles in the current frame. s_o and s_e represent the distances from other vehicles to the competition area and from the vehicle to the competition area, respectively. The above distances are based on the vehicle's (frenet) coordinate system.

[0128] Step two involves estimating the arrival time distribution. Based on two longitudinal strategy assumptions—comfortable deceleration and comfortable acceleration—let a_min and a_max represent the minimum and maximum acceleration of the vehicle's control strategy, respectively. The arrival time distributions can be calculated as t_e = [t_min_e, t_max_e] and t_o = [t_min_o, t_max_o]. t_e represents the time distribution range for the vehicle to reach the specified competitive area after considering both comfortable deceleration and comfortable acceleration strategies. t_min_e is the earliest possible arrival time when using maximum acceleration (i.e., comfortable acceleration), while t_max_e is the latest possible arrival time when using minimum acceleration (i.e., comfortable deceleration). This yields a time window, i.e., the time period during which the vehicle may arrive at a specific location. t_o refers to the time distribution of other vehicles; t_min_o and t_max_o are the earliest and latest possible arrival times of other vehicles in the same competitive area under the comfortable acceleration and comfortable deceleration strategies, respectively. By comparing these two sets of time distributions, autonomous driving systems can assess the feasibility of lane changes or other decisions, as well as the probability of successfully executing these decisions under different acceleration strategies.

[0129] Step 3: Compare the magnitudes of the two time distributions to derive the probability distribution of the game's outcome. Here, based on the assumption that the distribution is normal, without introducing other costs for complex probability calculations, a simple weighted average calculation of t_ave, i.e., the median value of the distribution, can be used to calculate the probability of winning or losing, representing whether the vehicle has the right-of-way.

[0130] Step four: Adjust the vehicle's strategy based on the right-of-way game result. By changing the vehicle's speed under the current strategy K, adjust the vehicle's position to improve the success rate of lane changing. If the game result is 1, maintain the vehicle's speed sampling strategy K; if it is 0.5, adjust the vehicle's speed sampling strategy under strategy K, sampling initial acceleration at 1 m / s², and finally determining the final control acceleration based on factors such as whether there is a following vehicle and speed limits; if it is 0, adjust the vehicle's speed sampling strategy under strategy K, sampling initial acceleration at 1 m / s², and finally determining the final control acceleration based on factors such as whether there are vehicles behind and the distance between vehicles.

[0131] Ultimately, the feasibility and success rate of adjusting the vehicle speed and changing the candidate strategy L (by performing forward simulation and scoring the forward simulation trajectory for comfort, safety, efficiency, etc.) are assessed, reusing the original algorithm framework. The numerical values ​​in the above process are for illustrative purposes only; specific values ​​need to be adjusted according to actual needs, and are not limited here.

[0132] This application also provides a vehicle control device. Figure 4 This is a schematic diagram of a vehicle control device according to an embodiment of this application, such as...Figure 4 As shown, the device 40 includes: a simulation module 402, a calculation module 404, a selection module 406, and a control module 408.

[0133] The simulation module 402 is used to perform forward simulations of multiple driving strategies for the vehicle based on the vehicle's environment model, obtaining multiple predicted driving trajectories. The environment model represents the environmental factors of the vehicle's current environment, and the forward simulation represents the process of simulating the vehicle's future driving trajectory. The calculation module 404 is used to calculate the costs of multiple driving strategies based on the multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, obtaining target score results for multiple driving strategies. The cost calculation quantifies the influence of multiple driving strategies, and the target score results represent the influence of multiple driving strategies. The selection module 406 is used to select a target driving strategy from the multiple driving strategies based on the target score results. The target driving strategy is the driving strategy with the highest target score among the multiple driving strategies. The control module 408 is used to control the vehicle based on the target driving strategy.

[0134] Optionally, the calculation module is used to determine the target competition region corresponding to the preset driving strategy among multiple driving strategies, based on the second predicted driving trajectory and the first predicted driving trajectory among multiple predicted driving trajectories; evaluate the preset driving strategy based on the target competition region to obtain a first score result for the preset driving strategy, wherein the second predicted driving trajectory is the predicted driving trajectory corresponding to the preset driving strategy, and the target competition region is used to represent the area where the predicted driving trajectories of the vehicle intersect with those of other vehicles; for other driving strategies among multiple driving strategies besides the preset driving strategy, evaluate the other driving strategies based on the second predicted driving trajectory and the first predicted driving trajectory to obtain a second score result for the other driving strategies; and obtain a target score result for multiple driving strategies based on the first score result and the second score result.

[0135] Optionally, the calculation module is further configured to determine at least one intersection based on the second predicted driving trajectory and the first predicted driving trajectory; determine the target intersection of the at least one intersection based on the intersection time information and intersection position information of the at least one intersection, wherein the target intersection is the intersection that has the greatest impact on the vehicle among the at least one intersection; and construct a target competition area based on the target intersection.

[0136] Optionally, the calculation module is further configured to determine a first time distribution of the vehicle and a second time distribution of other vehicles based on a preset acceleration range and a target competition area, wherein the first time distribution represents the time distribution of the vehicle arriving at the target competition area, and the second time distribution represents the time distribution of other vehicles arriving at the target competition area; and to evaluate the preset driving strategy based on the first time distribution and the second time distribution to obtain a first score result of the preset driving strategy.

[0137] Optionally, the calculation module is also used to transform the first time distribution to obtain a first normal distribution; transform the second time distribution to obtain a second normal distribution; and evaluate the preset driving strategy based on the overlapping area of ​​the first normal distribution and the second normal distribution to obtain a first score result of the preset driving strategy.

[0138] Optionally, the calculation module is further configured to: if the area of ​​the overlapping region is greater than a first preset value, determine the current score result of the preset driving strategy as the first score result of the preset driving strategy; if the area of ​​the overlapping region is less than or equal to the first preset value and greater than or equal to a second preset value, evaluate the control state of the vehicle based on the first acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy; if the area of ​​the overlapping region is less than or equal to the second preset value, evaluate the control state of the vehicle based on the second acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy, wherein the second acceleration is less than the first acceleration.

[0139] Optionally, the device is also used to acquire environmental perception data of the vehicle and road map data of the road where the vehicle is currently located; and to construct an environmental model based on the environmental perception data and road map data.

[0140] This application also provides a vehicle 50, please refer to... Figure 5 It includes a processor 510 and a memory 520, wherein the memory 510 is used to store computer programs; the processor 520 is used to execute the programs stored in the memory 510 to implement the vehicle control method described in any embodiment of this application.

[0141] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the vehicle control method described in any embodiment of this application.

[0142] In this application, "multiple" refers to two or more.

[0143] In this application, unless otherwise expressly defined, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0144] The terms “first,” “second,” “third,” “fourth,” etc., in this application (if any) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0145] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0146] Unless otherwise specified, all steps in this application may be performed sequentially or randomly. For example, if the method includes steps A and B, it means that the method may include steps A and B performed sequentially, or it may include steps B and A performed sequentially. For example, if the method may also include step C, it means that step C may be added to the method in any order. For example, the method may include steps A, B, and C, or it may include steps A, C, and B, or it may include steps C, A, and B, etc.

[0147] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for controlling a vehicle, characterized in that, include: Based on the vehicle's environment model, forward simulations are performed on multiple driving strategies of the vehicle to obtain multiple predicted driving trajectories. The environment model is used to represent the environmental factors of the current environment of the vehicle, and the forward simulation is used to represent the process of simulating the future driving trajectory of the vehicle. Based on the multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, the multiple driving strategies are evaluated respectively to obtain target score results for the multiple driving strategies, wherein the target score results are used to represent the degree of influence of the multiple driving strategies; Based on the target score results of the multiple driving strategies, a target driving strategy is selected from the multiple driving strategies, wherein the target driving strategy is the driving strategy with the highest target score result among the multiple driving strategies; The vehicle is controlled based on the target driving strategy.

2. The method according to claim 1, characterized in that, Based on the multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, the costs of the multiple driving strategies are calculated respectively to obtain the target score results of the multiple driving strategies, including: For a preset driving strategy among the plurality of driving strategies, a target competition area corresponding to the preset driving strategy is determined based on the second predicted driving trajectory and the first predicted driving trajectory among the plurality of predicted driving trajectories. Based on the target competition area, the preset driving strategy is evaluated to obtain a first score result of the preset driving strategy. The second predicted driving trajectory is the predicted driving trajectory corresponding to the preset driving strategy, and the target competition area is used to represent the area where the predicted driving trajectories of the vehicle intersect with those of other vehicles. For the driving strategies other than the preset driving strategy among the multiple driving strategies, the other driving strategies are evaluated based on the second predicted driving trajectory and the first predicted driving trajectory to obtain a second score result for the other driving strategies; Based on the first rating result and the second rating result, the target rating result of the multiple driving strategies is obtained.

3. The method according to claim 2, characterized in that, Based on the second predicted driving trajectory and the first predicted driving trajectory, the target competition area of ​​the preset driving strategy is determined, including: Based on the second predicted driving trajectory and the first predicted driving trajectory, at least one intersection point is determined; Based on the intersection time information and intersection location information of the at least one intersection, a target intersection is determined, wherein the target intersection is the intersection that has the greatest impact on the vehicle among the at least one intersection. The target competition region is constructed based on the target intersection point.

4. The method according to claim 2, characterized in that, Based on the target competition area, the preset driving strategy is evaluated to obtain a first score result for the preset driving strategy, including: Based on a preset acceleration range and a target competition area, a first time distribution of the vehicle and a second time distribution of the other vehicles are determined, wherein the first time distribution represents the time distribution of the vehicle arriving at the target competition area, and the second time distribution represents the time distribution of the other vehicles arriving at the target competition area; Based on the first time distribution and the second time distribution, the preset driving strategy is evaluated to obtain a first score result for the preset driving strategy.

5. The method according to claim 4, characterized in that, Based on the first time distribution and the second time distribution, the preset driving strategy is evaluated to obtain a first score result for the preset driving strategy, including: The first time distribution is transformed to obtain the first normal distribution; The second time distribution is transformed to obtain the second normal distribution; Based on the overlapping region of the first normal distribution and the second normal distribution, the preset driving strategy is evaluated to obtain a first score result of the preset driving strategy.

6. The method according to claim 5, characterized in that, Based on the overlapping region of the first normal distribution and the second normal distribution, the preset driving strategy is evaluated to obtain a first score result of the preset driving strategy, including: If the area of ​​the overlapping region is greater than a first preset value, the current rating result of the preset driving strategy is determined as the first rating result of the preset driving strategy; If the area of ​​the overlapping region is less than or equal to the first preset value and greater than or equal to the second preset value, the control state of the vehicle is evaluated based on the first acceleration and the preset driving strategy to obtain the first score result of the preset driving strategy. If the area of ​​the overlapping region is less than or equal to the second preset value, the control state of the vehicle is evaluated based on the second acceleration and the preset driving strategy to obtain a first score result of the preset driving strategy, wherein the second acceleration is less than the first acceleration.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: Acquire the vehicle's environmental perception data of the environment, as well as the road map data of the road where the vehicle is currently located; The environmental model is constructed based on the environmental perception data and the road map data.

8. A vehicle control device, characterized in that, include: The simulation module is used to perform forward simulations on multiple driving strategies of the vehicle based on the vehicle's environment model to obtain multiple predicted driving trajectories. The environment model is used to represent the environmental factors of the current environment of the vehicle, and the forward simulation is used to represent the process of simulating the future driving trajectory of the vehicle. The calculation module is used to calculate the cost of the multiple driving strategies based on the multiple predicted driving trajectories and the first predicted driving trajectories of other vehicles in the environment, and to obtain the target score results of the multiple driving strategies. The cost calculation is used to quantify the degree of influence of the multiple driving strategies, and the target score results are used to represent the degree of influence of the multiple driving strategies. The selection module is used to select a target driving strategy from the multiple driving strategies based on the target score results of the multiple driving strategies, wherein the target driving strategy is the driving strategy with the highest target score result among the multiple driving strategies; A control module is used to control the vehicle based on the target driving strategy.

9. A vehicle, characterized in that, Includes processor and memory, of which: Memory, used to store computer programs; A processor for executing a program stored in memory to implement the method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1 to 7.

Images