A driving intention adaptive method, system, device and readable storage medium based on vehicle driving instruction

Abstract

The application discloses a driving intention adaptive method and system based on vehicle driving instructions, equipment and a readable storage medium, and the method comprises the following steps: analyzing driving instructions to generate a structured driving intention containing an immutable core semantic field and a reconfigurable execution field; generating a candidate trajectory based on the intention and performing consistency evaluation; if the evaluation does not meet the execution condition, locking the core field and modifying the execution field to determine the target driving intention; and controlling the vehicle according to the target intention. The application solves the problem that the existing system cannot retain the core target and flexibly adjust the parameters when the instruction conflicts with the environment, avoids mechanically executing dangerous instructions or directly rejecting, improves the instruction completion rate and safety, and realizes the adaptive convergence of the driving intention under the safety constraint.

Description

Technical Field

[0001] This application relates to the field of autonomous driving, specifically to a method, system, device, and readable storage medium for adaptive driving intention based on vehicle driving instructions. Background Technology

[0002] With the continuous development of intelligent cockpits and autonomous driving technologies, natural language interaction, due to its convenience and intuitiveness, is gradually becoming an important interaction method in human-machine co-driving scenarios. Users expect to directly express their driving intentions through natural language commands, and the vehicle system can interpret these intentions and control the vehicle to perform corresponding driving actions to meet personalized driving needs.

[0003] In related technologies, natural language driving command processing systems typically employ a combination of intent recognition and trajectory planning. The system receives the user's natural language input, converts it into a specific driving task or target state, then generates a planned trajectory based on vehicle environmental information, and controls the vehicle to drive along that trajectory.

[0004] However, when existing natural language driving systems conflict with the physical environment, they cannot flexibly adjust execution parameters while preserving the user's core driving objectives. This often results in the system mechanically executing dangerous commands or directly rejecting commands, reducing the command execution rate and safety. Summary of the Invention

[0005] This application provides a driving intention adaptive method, system, device, and readable storage medium based on vehicle driving instructions. It can solve the technical problem that existing natural language driving systems cannot flexibly adjust execution parameters while preserving the user's core driving goals when there is a conflict between instructions and the physical environment. This often results in the system only being able to mechanically execute dangerous instructions or directly reject instructions, which reduces the execution rate and safety of instructions.

[0006] In a first aspect, embodiments of this application provide a driving intention adaptation method based on vehicle driving commands, the driving intention adaptation method based on vehicle driving commands includes: The obtained driving instructions are parsed to generate a structured driving intent, wherein the structured driving intent includes at least an immutable core semantic field and a reconfigurable execution field; Based on the structured driving intent, candidate trajectories are generated through a planning model; The consistency between the candidate trajectory and the structured driving intention is evaluated to obtain a consistency evaluation result; If the consistency assessment result does not meet the execution conditions and is of the correctable type, then the immutable core semantic field is locked, and the reconfigurable execution field is corrected to determine the target driving intention; The vehicle is controlled according to the target driving intention to achieve driving intention adaptation.

[0007] In conjunction with the first aspect, in one implementation, generating candidate trajectories based on the structured driving intent using a planning model includes: The structured driving intent is converted into input conditions and then input into the planning model; One or more candidate trajectories are generated using a planning model; The planning model includes at least a diffusion model, and the condition inputs include condition vectors or constraint parameters.

[0008] In conjunction with the first aspect, in one implementation, the step of performing a consistency assessment on the candidate trajectory and the structured driving intention to obtain a consistency assessment result includes: Evaluate the consistency between the candidate trajectory and the structured driving intent in semantic, safety, and dynamic dimensions; Output the consistency assessment result, which includes a consistency score and a reason code for inconsistency; The inconsistency reason code includes the inconsistency reason and the correctable type.

[0009] In conjunction with the first aspect, in one implementation, if the consistency assessment result does not meet the execution conditions and is of a correctable type, then the immutable core semantic field is locked, and the reconfigurable execution field is corrected to determine the target driving intention, including: The consistency assessment results include a consistency score and a reason code for inconsistency. When the consistency score does not meet the execution conditions and the inconsistency reason code belongs to the correctable type, the immutable core semantic field is locked based on the field attribute. The target execution field to be corrected is determined based on the inconsistency reason code; Adjust the parameter values ​​of the target execution field to obtain the reconstructed structured driving intent; The reconstructed structured driving intent is determined as the target driving intent.

[0010] In conjunction with the first aspect, in one implementation, the step of locking the immutable core semantic field based on field attributes, determining the target execution field to be corrected based on the inconsistency reason code, adjusting the parameter values ​​of the target execution field, and obtaining the reconstructed structured driving intent includes: A field locking mask is generated based on the field attributes, and the immutable core semantic field is locked based on the field locking mask; Based on the inconsistency reason code, analyze the failure characteristics of the candidate trajectory and the current state of the vehicle to determine the target execution field to be corrected; Based on the target execution field to be corrected, a field update amount is generated, and the parameter value of the target execution field is corrected based on the field update amount to obtain the reconstructed structured driving intent.

[0011] In conjunction with the first aspect, in one implementation, the parsed driving instructions are used to generate a structured driving intent, including: Semantic parsing is performed on the acquired natural language driving instructions to extract intent elements; The structured driving intent is generated based on the intent elements; Among them, the immutable core semantic fields represent the driving action targets, including changing lanes, following other vehicles, or parking; Reconfigurable execution fields represent execution parameters, including target speed, time window, or safety margin.

[0012] In conjunction with the first aspect, in one implementation, after performing a consistency assessment on the candidate trajectory and the structured driving intention to obtain a consistency assessment result, the method further includes: If the consistency assessment result meets the execution conditions, then the structured driving intention is determined as the target driving intention; The vehicle is controlled according to the target driving intention to achieve driving intention adaptation.

[0013] Secondly, embodiments of this application provide a driving intention adaptive system based on vehicle driving commands, the system comprising: The intent parsing module is used to parse the acquired driving instructions and generate structured driving intents, wherein the structured driving intents include at least immutable core semantic fields and reconfigurable execution fields; The trajectory planning module is used to generate candidate trajectories based on the structured driving intent through a planning model; The consistency assessment module is used to assess the consistency between the candidate trajectory and the structured driving intention, and obtain the consistency assessment result. The intent reconstruction module is used to lock the immutable core semantic field and correct the reconstructable execution field if the consistency evaluation result does not meet the execution conditions and belongs to the correctable type, so as to determine the target driving intent. The vehicle control module is used to control the vehicle according to the target driving intention in order to achieve adaptive driving intention.

[0014] Thirdly, embodiments of this application provide a driving intention adaptation device based on vehicle driving instructions. The driving intention adaptation device based on vehicle driving instructions includes a processor, a memory, and a driving intention adaptation program based on vehicle driving instructions stored in the memory and executable by the processor. When the driving intention adaptation program based on vehicle driving instructions is executed by the processor, it implements the steps of the driving intention adaptation method based on vehicle driving instructions as described in any of the preceding claims.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a driving intention adaptation program based on vehicle driving instructions, wherein when the driving intention adaptation program based on vehicle driving instructions is executed by a processor, it implements the steps of the driving intention adaptation method based on vehicle driving instructions as described in any of the preceding claims.

[0016] The beneficial effects of the technical solutions provided in this application include: By parsing the obtained driving commands, a structured driving intent is generated. This structured driving intent includes at least immutable core semantic fields and reconfigurable execution fields. Based on the structured driving intent, a planning model generates candidate trajectories. The consistency between the candidate trajectories and the structured driving intent is evaluated. If the consistency evaluation result does not meet the execution conditions and is of a correctable type, the immutable core semantic fields are locked, and the reconfigurable execution fields are corrected to determine the target driving intent. The vehicle is then controlled according to the target driving intent to achieve adaptive driving intent. This solves the problem in existing natural language driving systems where, when commands conflict with the physical environment, the system cannot retain user input. The technical problem of flexibly adjusting execution parameters under the premise of core driving objectives often leads to the system only being able to mechanically execute dangerous commands or directly reject commands, reducing the execution rate and safety of commands. This application solves the problem by dividing driving intention into immutable core semantic fields and reconfigurable execution fields, and locking the core fields and only modifying the execution fields when the consistency assessment does not meet the execution conditions. This allows the system to retain the user's fundamental driving objectives while adaptively adjusting execution parameters when facing environmental conflicts, avoiding dangers caused by forced execution or experience degradation caused by direct rejection. This achieves closed-loop convergence of driving intention under safety constraints, improving command completion rate and driving safety. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the adaptive driving intention method based on vehicle driving commands in this application. Figure 2 This is a schematic diagram of the adaptive process of the driving intention adaptive method based on vehicle driving instructions in this application; Figure 3This is a schematic diagram of the functional modules of the adaptive driving intention system based on vehicle driving commands in this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application are described below with reference to the accompanying drawings and related embodiments. In the description of the embodiments of this application, the terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to limit the application. As used in the specification and appended claims of this application, the singular expressions "a," "the," "the," "the," and "this" are intended to also include expressions such as "one or more," unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, "at least one" and "one or more" refer to one or more (including two). The term "and / or" is used to describe the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0019] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized. The term "connection" includes direct connections and indirect connections, unless otherwise stated. "First" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated.

[0020] In the embodiments of this application, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplarily" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0021] The steps involved in the methods provided in this application are merely examples, and not all steps are mandatory, nor are all information or message contents required. They can be added or removed as needed during use. In this application, the same step, or steps or messages with the same function, can be referenced and learned from each other in different embodiments.

[0022] The driving intention adaptation method based on vehicle driving commands provided in this application can be applied to vehicles with autonomous driving capabilities. These vehicles are equipped with an onboard computing platform, which can also be referred to as a vehicle control unit. Exemplarily, the onboard computing platform can be a domain controller, a central computing unit, or an embedded industrial control computer, or other devices that support vehicle control functions. The onboard computing platform includes a processor and a memory. The processor is used to call computer programs stored in the memory to execute the method provided in this application. The vehicle also includes a sensor module and an actuator. The sensor module is used to collect vehicle environmental information, and the actuator is used to drive the vehicle according to control commands.

[0023] The following description uses an intelligent passenger vehicle as an example to illustrate the adaptive driving intention method based on vehicle driving commands provided in this application, in conjunction with the accompanying drawings and specific application scenarios.

[0024] In one possible implementation, the driving intention adaptation method based on vehicle driving commands provided in this application includes the following steps: S10: Parse the obtained driving instructions and generate structured driving intentions.

[0025] Among them, structured driving intent refers to the machine-readable data format converted from natural language driving instructions. This invention forcibly divides its internal fields into two categories: immutable core semantic fields and reconfigurable execution fields. The immutable core semantic fields represent the user's most fundamental driving goal and are locked during the system's automatic correction process, which does not allow modification and ensures that the user's original intention is not violated. The reconfigurable execution fields represent the specific parameters for achieving the core goal and allow the system to make dynamic adjustments based on road conditions and safety constraints.

[0026] For example, the user's input of driving commands can be a user's trigger operation on the human-machine interface in the vehicle. The human-machine interface has at least a microphone, touch screen controls, etc., for receiving the user's natural language commands or point commands. When the vehicle detects that the user speaks a voice command during driving, it can receive the voice signal through the microphone and perform voice recognition.

[0027] In this embodiment, the immutable core semantic field represents the driving action target, including lane changing, following, or parking; the reconfigurable execution field represents the execution parameters, including target speed, time window, or safety margin; see Table 1 for the specific definitions and attributes of the immutable core semantic field, and see Table 2 for the specific definitions and attributes of the reconfigurable execution field.

[0028] Table 1: Definition Table of Immutable Core Semantic Fields

[0029] Table 2: Reconfigurable Execution Field Definition Table

[0030] This step also includes a field-constraint adaptation process, setting deterministic or parameterized field-constraint adapters for each type of reconfigurable execution field, converting its semantic or categorical values ​​into numerical constraints or conditional inputs usable by the planner. Specifically, there are three mapping methods: hard constraint mapping, which directly converts numerical or discretizable fields into numerical thresholds in the planner's cost function or boundary constraints, such as mapping target speed to a speed upper bound constraint; conditional predicate mapping, which converts conditional or time-based fields into Boolean-triggered predicates based on environmental observations, where candidate trajectories can only enter the consistency evaluation execution decision when the predicate is true, essentially serving as a gating condition for the planning loop; and condition vector or condition label mapping, which encodes semantic or preference-based fields into condition vectors or condition labels via a structured driving intent management module, inputting them into the end-to-end planning module through cross-attention or feature linear modulation condition injection. Explicit rule mapping is used for fields strongly related to safety or legality, ensuring verifiable and auditable decisions; implicit model mapping is used for semantic or preference-related fields, ensuring scenario generalization capability. The combination of these two forms a complete conversion link from semantics to trajectory constraints.

[0031] S20 generates candidate trajectories based on structured driving intentions through a planning model.

[0032] The system transforms structured driving intentions into input conditions that the planning model can understand. Specifically, the structured driving intentions can be encoded into condition vectors or constraint parameters and input into the planning model. The planning model includes at least a diffusion model, and the condition inputs include condition vectors or constraint parameters.

[0033] The specific content includes: the condition vector is used to represent the semantic features of the intention, and each dimension of the vector corresponds to a potential semantic dimension of the intention; the constraint parameters can be boundary values ​​of a numerical range, such as the speed constraint being the range from the minimum to the maximum value, and the acceleration constraint being the range from the minimum to the maximum value; the planning model generates one or more candidate trajectories based on the current vehicle's environmental observation data and input conditions. The candidate trajectory usually contains a series of time-state points, such as position, speed, and heading angle, and the trajectory has a preset time span and sampling frequency.

[0034] S30 performs a consistency assessment on the candidate trajectory and the structured driving intention to obtain a consistency assessment result.

[0035] After generating candidate trajectories, the system does not execute them directly, but first performs a consistency assessment. The assessment is mainly carried out in multiple dimensions, including semantic consistency, whether the candidate trajectory achieves the target defined by the immutable core semantic fields; safety consistency, whether the candidate trajectory complies with vehicle dynamics constraints and traffic regulations, and whether there is a collision risk; and dynamic consistency, whether the acceleration and jerk of the candidate trajectory are within the safety threshold of the vehicle actuator.

[0036] The consistency assessment results include a consistency score and a non-consistency reason code. The non-consistency reason code includes the reason for the non-consistency and the type of correction. The consistency score quantifies the degree of matching between the trajectory and the intent, as well as the security; a higher score indicates better performance. The specific formula for calculating the consistency score is S:

[0037] in, The semantic consistency score indicates whether the action type corresponding to the candidate trajectory is consistent with the semantics of the core action. The target hit score indicates whether the trajectory meets the constraints of the target object, target lane, or target location. The rule consistency score indicates whether the trajectory satisfies traffic rules, road topology, and navigation constraints. The safety consistency score indicates whether collision risk, minimum distance, and comfort boundaries are met. The dynamic executability score indicates whether the trajectory satisfies the vehicle's kinematics and dynamics boundaries. The temporal consistency score indicates whether the timing of trajectory execution meets the time window requirements. For each weight, the best option is selected that satisfies... ; Without providing driving tasks or scenarios, the weights of each component can be dynamically adjusted. For example, in a parking task, the weight of each component can be increased. , and The weight; in lane change tasks, it can improve , and The weight; in navigation tasks, it can improve , and The weight.

[0038] Based on consistency score A tiered judgment mechanism is preferred:

[0039] in, The threshold for normal execution. To conservatively enforce the threshold, the reason code R and the consistency score S jointly determine the subsequent control logic: if S is low but R is of the correctable type, the intention refactoring process is initiated; if R is of the uncorrectable type, refactoring is no longer performed, but instead, execution rejection, clarification request, or safe rollback is triggered.

[0040] The inconsistency cause code is generated using a two-level hybrid mechanism. The first level is the rule layer, which is primary, interpretable, and auditable. In addition to outputting multiple sub-item scores, the consistency assessment also outputs the corresponding specific diagnostic quantities, such as the measured collision time, target lane deviation, rule boundary crossing flag, dynamic boundary crossing magnitude, and time deviation. The system predefines deterministic mapping rules from sub-item score defect patterns to cause codes. For example, a specific cause code is corresponding to a safety sub-item below the threshold and a measured collision time below the safety lower bound; a specific cause code is corresponding to a low object sub-item and a candidate target entity count not equal to 1; a specific cause code is corresponding to a low rule sub-item and a lane or area boundary crossing flag is detected; and so on, covering multiple cause codes. The second level is the learning layer, which assists in adjudication. When multiple sub-items are simultaneously low, or when diagnostic features overlap or conflict, making it impossible for the rule layer to make a unique judgment, a classification model is introduced with diagnostic feature vectors as input and cause code category and confidence level as output for fine-grained adjudication. The key cause codes involving safety or legality are always given a definitive conclusion by the rule layer. The learning layer is only used to adjudicate semantic ambiguity, multi-factor coupling and other fuzzy boundary situations, and does not independently determine the judgment result of uncorrectable categories, so as to meet the requirements of automotive-grade functional safety for decision traceability. For the specific definition and classification of inconsistent cause codes, please refer to Table 3.

[0041] Table 3: Definition Table of Inconsistency Reason Codes

[0042] R1 to R9 are typically correctable types, allowing the system to attempt to update reconfigurable execution fields while maintaining immutable core semantic fields. R10 is an uncorrectable type, indicating an irreconcilable conflict between the user's core objective and current traffic rules, safety boundaries, or physical executability. In this case, the system will not proceed with the intent reconfiguration process but will directly trigger a safety rollback. For rule conflicts represented by R4, the system can further determine whether they are correctable based on the nature of the conflict. For example, "the target parking spot is located in a no-parking zone" can be reconfigured by searching for nearby legal parking areas; while "requiring a vehicle to cross a solid line to overtake" is an uncorrectable conflict and should be rejected.

[0043] S40. If the consistency assessment result does not meet the execution conditions and is of the correctable type, then the immutable core semantic field is locked and the reconfigurable execution field is corrected to determine the target driving intention.

[0044] In a preferred embodiment, intent reconstruction is initiated when the following conditions are met: the consistency score between the candidate trajectory and the structured driving intent is lower than the normal execution threshold; the current trajectory is not suitable for direct execution or can only be executed in a conservative manner; the reason code is of the correctable type; there is no irreconcilable conflict between the immutable core semantic fields and the current traffic rules and safety constraints; and the current iteration number has not exceeded the maximum reconstruction number.

[0045] First, determine correctability: check the inconsistency reason code. If the reason code is of an uncorrectable type, such as an irreconcilable conflict between the core objectives, trigger a safety rollback strategy, such as requesting manual takeover or shutdown; if it is of a correctable type, proceed to the next step. Secondly, field locking is performed: a field locking mask is generated based on the field attributes, and immutable core semantic fields are locked based on the field locking mask. For immutable core semantic fields, the mask is set to a prohibited modification state; for reconfigurable execution fields, the mask is set to a permitted modification state. Then, the correction strategy is determined: based on the inconsistency reason code, the failure characteristics of the candidate trajectory and the current state of the vehicle are analyzed to determine the target execution field to be corrected. To improve the interpretability and feasibility of the intent reconstruction process, this invention establishes a mapping relationship between reason codes and reconstruction actions. This mapping relationship is used to guide the intent reconstruction module to determine the field to be updated and the update strategy; see Table 4 for the specific reason code-reconstruction action mapping relationship.

[0046] Table 4: Reason Code-Refactoring Action Mapping Table

[0047] Next, parameter adjustments are performed: Field update values ​​are generated based on the target execution field to be corrected, and the parameter values ​​of the target execution field are corrected based on these update values ​​to obtain the reconstructed structured driving intent. For correctable cause codes, corresponding reconstructable execution field update functions are predefined. These functions take the diagnostic values ​​obtained from the analysis of non-executable root causes as input and output specific field update values. Update functions are divided into two categories based on the physical dimensions of the fields. Both are deterministic calculations, neither using gradient descent nor allowing the language model to freely generate values ​​within the closed loop. Analytical updates are used for fields with clear physical or kinematic meaning, such as safety margin, time window, and target speed. For example, taking insufficient safety margin as an example, the system solves for a new time window based on the current collision time measurement, safety threshold, and relative speed and distance of oncoming vehicles in the target lane, according to kinematic relationships. That is, the specific waiting time is obtained by solving the closed-form kinematic equation of the time required for oncoming vehicles to pass through the safety window, rather than a fixed increment. Similarly, the target velocity is updated according to the velocity upper bound obtained from the inverse solution of safety constraints, and the safety margin is updated according to the margin inverse solution formula. Numerical search updates are used for spatial fields such as target pose and spatial tolerance that cannot be directly solved in a closed loop. Within the feasible region constraints given by the candidate trajectory failure features, a restricted local search is adopted, such as sampling and searching around the original target area at a preset step size or direction, to find the nearest feasible value that satisfies the consistency evaluation constraints. The language model is only used in the natural language parsing stage of step S10 and does not participate in the field update calculation at the numerical level within the closed loop, so as to ensure the determinism and reproducibility of the numerical calculation process and meet the automotive-grade real-time and verifiability requirements. Finally, the reconstructed intent is generated: the reconstructed structured driving intent is determined as the target driving intent.

[0048] Optionally, to prevent infinite refactoring loops, the system sets refactoring termination conditions and a safe rollback mechanism. The system maintains a refactoring iteration counter for each natural language instruction processing session, in conjunction with field change monitoring and cause code oscillation detection. Counter initialization: When the language intent parsing module receives a new natural language driving instruction, the counter is reset to its initial value. Counter increment: Each time the intent refactoring module generates a new structured driving intent, re-enters the planning loop, and completes a new consistency assessment, the counter is incremented by 1. Hard termination: When the counter reaches the preset maximum number of refactoring iterations, regardless of the current consistency score, the refactoring process is immediately stopped and the system transitions to the decision-making and rollback module. Early termination of convergence stagnation: Record the changes in fields of the structured driving intent before and after each reconstruction. When the changes in fields of two consecutive reconstructions are both lower than the preset threshold, even if the counter is less than the maximum number of reconstructions, it is determined to be a convergence stagnation and a safety rollback is triggered in advance. Oscillation detection: Maintain a cause code sequence window with a length equal to the maximum number of reconstructions. If the cause codes repeatedly alternate in recent reconstructions and the corresponding field update directions are contradictory, it is determined to be a potential dead loop mode, and the reconstruction is immediately terminated and a safety rollback is triggered. Counter reset condition: The counter is only reset when the system receives a brand new natural language driving instruction, or when the user actively confirms or reissues the current immutable core semantic field.

[0049] S50, if the consistency assessment result meets the execution conditions, the structured driving intention is determined as the target driving intention.

[0050] If the consistency assessment result directly meets the execution conditions in the consistency assessment step, such as the score being higher than the threshold, then there is no need for reconstruction, and the initial structured driving intention can be directly used as the target driving intention.

[0051] S60 controls the vehicle according to the target driving intention to achieve driving intention adaptation.

[0052] Once the target driving intent is determined, whether it is the initially generated structured driving intent (from S50) or the reconstructed intent (from S40), the system inputs it back into the planning model to generate the final trajectory, or directly calculates control commands such as steering angle, throttle, and braking based on the intent and sends them to the vehicle chassis actuators to control the vehicle's movement; when the system cannot obtain a candidate trajectory that meets the execution conditions through intent reconstruction, it adopts the corresponding rollback strategy according to the failure type.

[0053] Rollback strategies include, but are not limited to: if intent resolution fails, request the user to restate or maintain the current safety state; if the target object cannot be bound, request the user to confirm the target object or adopt a conservative default target; if the target area is unreachable, search for the nearest legally reachable area, and refuse execution if there is no alternative area; if rule conflicts cannot be corrected, refuse the original execution expression and provide the reason; if the safety margin is consistently insufficient, slow down and wait, maintain the lane, or request takeover; if dynamics are not feasible, reduce speed and replan, and rollback if still not feasible; if language ambiguity cannot be resolved, request user confirmation or adopt the system's default safety policy; if the maximum number of reconstructions is reached, execute a conservative trajectory or trigger a minimum risk policy; if the core objectives are irreconcilable and conflicting, refuse execution, request manual takeover, or stop with minimum risk.

[0054] The rollback mechanism follows these principles: it does not violate traffic rules and safety boundaries due to user language commands; it does not directly execute high-risk trajectories in a state of low confidence or low consistency; it does not arbitrarily change the user's core driving objectives; it prioritizes maintaining the vehicle's safe state when reconstruction cannot be completed; and it can provide feedback on alternative execution results or request confirmation from the user when a legal alternative solution is available.

[0055] To illustrate this embodiment more clearly, a specific scenario will be described below.

[0056] Scenario: The vehicle is driving on the highway, and the user gives the voice command: "Overtake immediately."

[0057] S10 parsing steps: The system parses the instructions and generates structured driving intentions. Immutable core semantic fields: Action: Overtaking, Target: Left lane; Reconfigurable execution fields: Time window: Now, Target speed: 120km / h, Safety margin: 10 meters.

[0058] S20 planning steps: The planning model generates candidate trajectories based on the perception data that there is a vehicle driving close to the left front.

[0059] S30 Assessment Steps: Consistency assessment revealed that, based on the time window: Execution now carries a high risk of collision. Consistency score: Below the threshold. Inconsistency reason code: Insufficient safety margin; Type: Correctable.

[0060] S40 Reconstruction Steps: Lock Core Fields: Action and objective remain unchanged; overtaking is mandatory; left lane must be moved. Correct Execution Fields: Based on the cause code, the system determines to adjust the time window and safety margin. Use analytical updates, calculating the new time window based on collision time measurements. Post-reconstruction Intent: Action: Overtaking; Objective: Left lane; Time window: 3 seconds later; Target speed: 120 km / h; Safety margin: 20 meters.

[0061] S60 control steps: Based on the reconstructed intent, the trajectory is replanned. At this point, the trajectory is safe and feasible, and the vehicle control executes the overtaking maneuver after 3 seconds. Through the above process, the system not only satisfies the user's core demand for "overtaking" but also avoids dangerous execution, achieving adaptive convergence of intent.

[0062] The method embodiments provided by this application have been described above. The system embodiments provided by this application will be described below. It should be understood that the description of the system embodiments corresponds to the description of the method embodiments. Therefore, for content not described in detail, please refer to the method embodiments above. For the sake of brevity, it will not be repeated here.

[0063] This application provides a driving intention adaptive system based on vehicle driving commands. The system includes: an intention parsing module, a trajectory planning module, a consistency evaluation module, an intention reconstruction module, and a vehicle control module.

[0064] The intent parsing module is used to parse the acquired driving instructions and generate structured driving intents. The structured driving intents include at least immutable core semantic fields and reconfigurable execution fields.

[0065] The trajectory planning module is used to generate candidate trajectories based on structured driving intentions through a planning model.

[0066] The consistency assessment module is used to assess the consistency between candidate trajectories and structured driving intentions, and obtain the consistency assessment results.

[0067] The intent reconstructing module is used to lock immutable core semantic fields and correct reconstructable execution fields if the consistency assessment result does not meet the execution conditions and belongs to the correctable type, so as to determine the target driving intent.

[0068] The vehicle control module is used to control the vehicle according to the target driving intention in order to achieve adaptive driving intention.

[0069] Optionally, the trajectory planning module is specifically used to: convert structured driving intentions into input conditions and input them into a planning model; generate one or more candidate trajectories through the planning model; wherein the planning model includes at least a diffusion model, and the condition inputs include condition vectors or constraint parameters.

[0070] Optionally, the consistency assessment module is specifically used to: assess the consistency between candidate trajectories and structured driving intentions in semantic, safety, and dynamic dimensions; output consistency assessment results, which include consistency scores and inconsistency reason codes; wherein, the inconsistency reason codes include inconsistency reasons and correctable types.

[0071] Optionally, the intent reconstruction module is specifically used for: locking immutable core semantic fields based on field attributes when the consistency score does not meet the execution conditions and the inconsistency reason code belongs to the correctable type; determining the target execution field to be corrected based on the inconsistency reason code; adjusting the parameter value of the target execution field to obtain the reconstructed structured driving intent; and determining the reconstructed structured driving intent as the target driving intent.

[0072] Optionally, the intent reconstruction module is also specifically used for: generating a field locking mask based on field attributes, locking immutable core semantic fields based on the field locking mask; analyzing the failure characteristics of candidate trajectories and the current state of the vehicle based on the inconsistency reason code, and determining the target execution field to be corrected; generating a field update amount based on the target execution field to be corrected, and correcting the parameter value of the target execution field based on the field update amount, to obtain the reconstructed structured driving intent.

[0073] Optionally, the intent parsing module is specifically used for: performing semantic parsing on the acquired natural language driving instructions to extract intent elements; generating structured driving intents based on intent elements; wherein, immutable core semantic fields represent driving action targets, including lane changing, following, or stopping; and reconfigurable execution fields represent execution parameters, including target speed, time window, or safety margin.

[0074] Optionally, the system also includes a decision module, which determines the structured driving intention as the target driving intention if the consistency evaluation result meets the execution conditions; and controls the vehicle according to the target driving intention to achieve driving intention adaptation.

[0075] It should be understood that the description of the system embodiments can refer to the relevant descriptions of the various method embodiments above. The implementation principle and technical effect are similar to those of the method embodiments above, and will not be repeated here.

[0076] It should be understood that a "module" in a system can be implemented in hardware, software, or by hardware executing corresponding software. For example, a "module" can refer to application-specific integrated circuits, electronic circuits, processors and memory for executing one or more software or firmware programs, integrated logic circuits, and / or other suitable components that support the described functions.

[0077] The foregoing described the method and system embodiments provided in this application, and the following describes the device embodiments provided in this application.

[0078] This application provides a driving intention adaptation device based on vehicle driving instructions. The driving intention adaptation device based on vehicle driving instructions includes a processor, a memory, and a driving intention adaptation program based on vehicle driving instructions stored in the memory and executable by the processor. When the driving intention adaptation program based on vehicle driving instructions is executed by the processor, it implements the steps of the driving intention adaptation method based on vehicle driving instructions as described in any of the above method embodiments.

[0079] It should be understood that the processor mentioned in the embodiments of this application can be a central processing unit, or it can be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0080] It should also be understood that the memory mentioned in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory can be read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, or flash memory. Volatile memory can be random access memory used as an external cache. By way of example, but not limitation, many forms of random access memory are available, such as static random access memory, dynamic random access memory, synchronous dynamic random access memory, double data rate synchronous dynamic random access memory, enhanced synchronous dynamic random access memory, synchronously linked dynamic random access memory, and direct memory bus random access memory.

[0081] The foregoing described embodiments of the methods, systems, and devices provided in this application; the following will describe embodiments of the computer-readable storage media provided in this application.

[0082] This application provides a computer-readable storage medium storing a driving intention adaptation program based on vehicle driving instructions. When the driving intention adaptation program based on vehicle driving instructions is executed by a processor, it implements the steps of the driving intention adaptation method based on vehicle driving instructions as described in any of the above method embodiments.

[0083] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. In the above embodiments, the descriptions of each embodiment have different focuses; parts not described in detail or recorded in a certain embodiment can be referred to in the relevant descriptions of other embodiments.

[0084] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0085] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0086] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0087] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0088] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0089] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0090] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0091] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A driving intention adaptive method based on vehicle driving commands, characterized in that, The adaptive driving intention method based on vehicle driving commands includes: The obtained driving instructions are parsed to generate a structured driving intent, wherein the structured driving intent includes at least an immutable core semantic field and a reconfigurable execution field; Based on the structured driving intent, candidate trajectories are generated through a planning model; The consistency between the candidate trajectory and the structured driving intention is evaluated to obtain a consistency evaluation result; If the consistency assessment result does not meet the execution conditions and is of the correctable type, then the immutable core semantic field is locked, and the reconfigurable execution field is corrected to determine the target driving intention; The vehicle is controlled according to the target driving intention to achieve driving intention adaptation.

2. The adaptive driving intention method based on vehicle driving commands according to claim 1, characterized in that, The step of generating candidate trajectories based on the structured driving intent through a planning model includes: The structured driving intent is converted into input conditions and then input into the planning model; One or more candidate trajectories are generated using a planning model; The planning model includes at least a diffusion model, and the condition inputs include condition vectors or constraint parameters.

3. The adaptive driving intention method based on vehicle driving commands according to claim 1, characterized in that, The process of evaluating the consistency between the candidate trajectory and the structured driving intention to obtain a consistency evaluation result includes: Evaluate the consistency between the candidate trajectory and the structured driving intent in semantic, safety, and dynamic dimensions; Output the consistency assessment result, which includes a consistency score and a reason code for inconsistency; The inconsistency reason code includes the inconsistency reason and the correctable type.

4. The adaptive driving intention method based on vehicle driving commands according to claim 1, characterized in that, If the consistency assessment result does not meet the execution conditions and is of a correctable type, then the immutable core semantic field is locked, and the reconfigurable execution field is corrected to determine the target driving intention, including: The consistency assessment results include a consistency score and a reason code for inconsistency. When the consistency score does not meet the execution conditions and the inconsistency reason code belongs to the correctable type, the immutable core semantic field is locked based on the field attribute. The target execution field to be corrected is determined based on the inconsistency reason code; Adjust the parameter values ​​of the target execution field to obtain the reconstructed structured driving intent; The reconstructed structured driving intent is determined as the target driving intent.

5. The adaptive driving intention method based on vehicle driving commands according to claim 4, characterized in that, The process of locking the immutable core semantic field based on field attributes, determining the target execution field to be corrected based on the inconsistency reason code, adjusting the parameter values ​​of the target execution field, and obtaining the reconstructed structured driving intent includes: A field locking mask is generated based on the field attributes, and the immutable core semantic field is locked based on the field locking mask; Based on the inconsistency reason code, analyze the failure characteristics of the candidate trajectory and the current state of the vehicle to determine the target execution field to be corrected; Based on the target execution field to be corrected, a field update amount is generated, and the parameter value of the target execution field is corrected based on the field update amount to obtain the reconstructed structured driving intent.

6. The adaptive driving intention method based on vehicle driving commands according to claim 1, characterized in that, The parsed driving instructions are used to generate structured driving intentions, including: Semantic parsing is performed on the acquired natural language driving instructions to extract intent elements; The structured driving intent is generated based on the intent elements; Among them, the immutable core semantic fields represent the driving action targets, including changing lanes, following other vehicles, or parking; Reconfigurable execution fields characterize execution parameters, including target speed, time window, or safety margin.

7. The adaptive driving intention method based on vehicle driving commands according to claim 1, characterized in that, After performing a consistency assessment on the candidate trajectory and the structured driving intention to obtain the consistency assessment result, the method further includes: If the consistency assessment result meets the execution conditions, then the structured driving intention is determined as the target driving intention; The vehicle is controlled according to the target driving intention to achieve driving intention adaptation.

8. A driving intention adaptive system based on vehicle driving commands, characterized in that, The system includes: The intent parsing module is used to parse the acquired driving instructions and generate structured driving intents, wherein the structured driving intents include at least immutable core semantic fields and reconfigurable execution fields; The trajectory planning module is used to generate candidate trajectories based on the structured driving intent through a planning model; The consistency assessment module is used to assess the consistency between the candidate trajectory and the structured driving intention, and obtain the consistency assessment result. The intent reconstruction module is used to lock the immutable core semantic field and correct the reconstructable execution field if the consistency evaluation result does not meet the execution conditions and belongs to the correctable type, so as to determine the target driving intent. The vehicle control module is used to control the vehicle according to the target driving intention in order to achieve adaptive driving intention.

9. A driving intention adaptive device based on vehicle driving commands, characterized in that, The vehicle driving instruction-based driving intention adaptation device includes a processor, a memory, and a vehicle driving instruction-based driving intention adaptation program stored in the memory and executable by the processor, wherein when the vehicle driving instruction-based driving intention adaptation program is executed by the processor, it implements the steps of the vehicle driving instruction-based driving intention adaptation method as 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 driving intention adaptation program based on vehicle driving instructions, wherein when the driving intention adaptation program based on vehicle driving instructions is executed by a processor, it implements the steps of the driving intention adaptation method based on vehicle driving instructions as described in any one of claims 1 to 7.

Images