Self-adaptive promotion robot driving training method based on capability grading and task decomposition

Abstract

The invention discloses a self-adaptive promotion robot driving training method based on capability grading and task decomposition, and aims to overcome the defects that in the prior art, test taking is emphasized, the capability is light, feedback is extensive, and a path is rigid. According to the method, a pyramid type three-layer teaching framework is adopted, and basic vehicle control training, atomic operation fragment strengthening and comprehensive process training are sequentially carried out. The system collects operation data in real time, constructs a dynamic ability portrait based on indexes such as vehicle speed standard deviation, calculates a comprehensive score through a weighting formula, realizes self-adaptive promotion in combination with a threshold value, and supports personalized point location marking and flexible splitting training. According to the invention, transformation from test taking to ability is realized, training precision and efficiency are improved, individual demands of trainees are adapted, teaching resource configuration is optimized, and the method is suitable for intelligent driving training scenes.

Description

Technical Field

[0001] This invention relates to the field of intelligent teaching and robot control technology, specifically to an adaptive progressive robot driving training method based on ability grading and task decomposition. Background Technology

[0002] With the continuous growth of motor vehicle ownership, the demand for driver training is increasing. The industry is upgrading towards intelligence and standardization. Robotic driver training systems, with their advantages of repeatability and consistency, are gradually replacing some traditional manual teaching methods and have become the technological development trend in the field of driver training.

[0003] Traditional driving skills training, especially for standardized subjects like reversing into a parking space and parallel parking, has long employed a training model centered on fixed reference points and with passing the test as the direct goal. Under this model, whether using traditional human instructors or early robotic instructors, the teaching content strictly adheres to the driving test syllabus, forming a fixed process: after getting into the car, the student selects the target test subject and mechanically repeats standardized instructions such as "turn the steering wheel all the way when your shoulder aligns with the reference line" as preset by the system or instructor. The essence of this method is to simplify continuous driving skills requiring comprehensive driving feel (i.e., "car sense") into a series of discrete, mechanical action commands. While this method has a certain short-term efficiency in passing standardized tests, its core drawback is that students are only trained in conditioned reflexes of "following instructions," without developing a fundamental perception and control ability regarding vehicle dynamics and spatial orientation. This results in severely insufficient adaptability and weak skill transferability when there are slight changes in actual road environments, vehicle parameters, or site markings.

[0004] With technological advancements, improved solutions have emerged, such as "full-process simulation teaching," which is currently the closest technological implementation to this invention. Its core logic is to completely replicate the entire examination process for either the second or third stage of the driving test using a robotic system. Students must operate continuously from the starting point to the finish line, and the system provides a binary judgment of "pass" or "fail" based solely on outcome indicators such as whether the vehicle crosses the line or stops accurately. However, this solution still fails to solve the core problems of the traditional model and instead exposes new technological flaws: Lacking a foundation for developing core competencies and with a distorted teaching logic: This program fails to break down driving skills into tiers, directly guiding beginners to complex maneuvers like reversing into a parking space. Forced into lengthy, complex operations before mastering basic vehicle control skills such as precise steering wheel control and linear clutch engagement, students rely heavily on rote memorization of reference points, failing to grasp the causal relationship between maneuvering and vehicle posture. This hinders the development of genuine core driving skills, resulting in a severe disconnect between training effectiveness and actual driving demands.

[0005] The feedback granularity is too coarse, resulting in low learning efficiency: Since the system only evaluates the final result and does not subdivide and evaluate the intermediate operations during the training process, when a long process task fails (for example, when the fifth step of reversing into a parking space is crossed, the actual reason is that the steering wheel was turned too early in the first step or the steering wheel was turned back too late in the third step), the learner cannot establish a connection between the final failure and the specific intermediate operation, making it difficult to locate their own weak points. They can only get into ineffective repetitive practice of "making mistakes from beginning to end without knowing where they went wrong", resulting in a steep learning curve and a lot of training time being wasted on meaningless repetition.

[0006] The teaching approach is rigid and lacks personalization: all students follow the same fixed process for training, without considering individual differences in driving talent, learning ability, and spatial awareness. For students with a good foundation and strong comprehension, repetitive training of already mastered content wastes time and teaching resources; for students with a weak foundation, the inability to complete the process may lead to frustration and even the development of incorrect operating habits, making it impossible to keep up with the uniform pace or receive targeted reinforcement training.

[0007] The shortcomings of the existing technologies mentioned above result in insufficient practicality and effectiveness of robot driving training. This not only lowers the pass rate of driving tests but also discourages trainees from driving independently on public roads even if they pass the test, due to a lack of solid core driving skills. Therefore, the driving training industry urgently needs a robot driving training method that can start by building basic driving skills, break down complex tasks into detailed steps, and dynamically adjust the teaching progress and content based on the trainee's real-time performance. This would address the core contradictions of existing technologies, such as "emphasizing test-taking skills over practical abilities," "emphasizing results over process," and "emphasizing uniformity over individualization," thus balancing training efficiency and skill practicality. Summary of the Invention

[0008] The purpose of this invention is to overcome the core technical defects of existing robot driving training methods, which emphasize test-taking over ability, results over process, and uniformity over individualization. It provides an adaptive, progressive robot driving training method based on ability grading and task decomposition. By constructing a hierarchical, progressive teaching logic that includes basic abilities, task decomposition reinforcement, and comprehensive application, driving skills are decoupled to first help trainees solidify core basic skills and establish a realistic driving experience (vehicle feel). Then, by breaking down complex comprehensive subjects into independent atomic operation segments and constructing a closed-loop training mechanism, feedback and guidance are provided down to specific operational steps. Simultaneously, a trainee ability profile and quantitative evaluation system based on multi-dimensional data such as vehicle stability, spatial perception, and operational accuracy are established. An adaptive progression logic is constructed to dynamically adjust the teaching path, ultimately achieving a dual improvement in driving training efficiency and skill practicality. This ensures trainees pass driving tests efficiently while cultivating their core driving abilities to adapt to real-world road environments. Furthermore, it optimizes the allocation of teaching resources, meets the personalized learning needs of trainees with different skill levels, and promotes the transformation of driving training from "test-oriented" to "ability-oriented."

[0009] To achieve the above objectives, the technical solution of this invention is: an adaptive progressive robot driving training method based on ability grading and task decomposition, wherein the method is executed based on a three-layer teaching control system, the three-layer teaching control system comprising a basic ability training layer, a task decomposition reinforcement layer, and a comprehensive business logic layer arranged sequentially; the method includes the following steps: S1 controls the trainee to perform vehicle control ability training that is decoupled from specific exam subjects at the basic ability training layer. S2, in the task decomposition and reinforcement layer, divides the target comprehensive driving subject into multiple independent atomic operation segments according to the operation nodes, and controls the trainee to perform cyclic reinforcement training on the selected atomic operation segments. S3 controls the trainee to complete the training process of the target comprehensive driving subject at the integrated business logic layer. During the execution of steps S1 to S3 above, the student's operation data is collected in real time, and the student's ability profile, which includes multiple quantitative ability indicators, is dynamically calculated and updated based on the operation data. Based on the quantitative ability indicators in the trainee's ability profile, the system automatically determines whether the trainee meets the conditions for advancing from the current training level to the next training level; when the determination result is met, the system automatically controls the trainee to advance to the next training level.

[0010] Furthermore, the basic skills training program includes at least one of the following: starting and stopping practice, straight driving practice, and reversing direction adjustment practice; The starting and stopping exercise is configured to involve multiple vehicle starts and stops within a short distance to train the trainee's basic vehicle control skills. The straight-line driving exercise is configured to train trainees to control the vehicle onto any target straight line using a deduced four-step steering wheel operation method.

[0011] Furthermore, the target comprehensive driving subject is divided into multiple independent atomic operation segments according to the operation nodes. Specifically, the right reverse parking subject is broken down into multiple atomic operation segments corresponding to the timing of turning the steering wheel to full, the timing of straightening the steering wheel, and the timing of stopping.

[0012] Furthermore, the selected atomic operation segments are subjected to iterative reinforcement training, specifically including: constructing an independent state machine for each atomic operation segment, wherein the state machine sequentially includes an initialization state, an operation execution state, an operation evaluation state, and a system reset state, and achieving iterative training for a single skill point by running the state machine.

[0013] Furthermore, the quantitative capability indicators include vehicle control stability score, spatial perception score, and operational accuracy score; The vehicle control stability score is calculated based on the standard deviation of vehicle speed and the smoothness of steering wheel angular velocity during the training process; the spatial perception score is calculated based on the vehicle's centering rate within the lane. The operation accuracy score is calculated based on the average error between the actual position and the preset position of the key point during the training of the atomic operation segment, and the key point includes the parking point.

[0014] Furthermore, the system calculates the trainees' overall ability score through weighted summation. The calculation formula is as follows: in, , , The preset weighting coefficients, The vehicle control stability score is given. The spatial perception score is given. The accuracy of the operation is scored.

[0015] Furthermore, the advancement criteria are configured as follows: when a trainee's quantitative ability index score in the basic ability training layer meets the first threshold condition, it is determined that the trainee meets the conditions for advancement from the basic ability training layer to the task decomposition enhancement layer. When a trainee's pass rate for all atomic operation segments included in the target comprehensive driving subject in the task decomposition enhancement layer meets the second threshold condition, it is determined that the trainee meets the condition for promotion from the task decomposition enhancement layer to the comprehensive business logic layer.

[0016] Furthermore, the first threshold condition is represented by the comprehensive ability score. The second threshold condition is expressed as a pass rate greater than 90%; where This is the preset threshold for advancing to the first level.

[0017] Furthermore, the steps for determining promotion criteria and controlling promotion are executed in a semi-automatic manner. Specifically, the system generates promotion suggestions based on the quantitative ability indicators and sends them to the coach. After receiving a manual confirmation instruction from the coach, the promotion operation is then executed.

[0018] Furthermore, instead of dividing atomic operation segments by operation node, a method of decomposing by operation action type can be adopted; when splitting by operation action type, the comprehensive driving subject is divided into clutch operation segment, steering wheel control segment, and vehicle speed adjustment segment.

[0019] Furthermore, during the training of atomic operation segments, trainees can independently mark personalized visual reference points. The system records these personalized visual reference points and associates them with the trainee's ability profile. In subsequent training, the system adapts these personalized visual reference points to provide operational guidance to trainees.

[0020] Furthermore, the adaptive adjustment strategy for loading training content includes: For training content corresponding to indicators whose scores in the trainee's ability profile are below the preset threshold, increase the training duration and number of cycles. For training content corresponding to indicators with scores higher than a preset threshold, reduce the number of repeated training sessions. For trainees with outstanding basic abilities, they can skip the basic ability training layer directly and have the system unlock the task decomposition enhancement layer and begin training.

[0021] The present invention, by adopting the above technical solution, has at least the following beneficial effects: This invention achieves a fundamental shift from "exam-oriented training" to "ability cultivation." By constructing a layered and progressive teaching architecture consisting of a "basic ability training layer - task decomposition and reinforcement layer - comprehensive business logic layer," it systematically decouples and reconstructs driving skills, effectively solving the core defects of existing technologies and possessing significant technological advancements and practical value. Specifically, the beneficial effects of this invention are reflected in the following aspects: First, this invention reverses the shortcomings of traditional teaching methods by addressing the root of the problem and solidifying learners' transferable core driving skills. It guides learners to prioritize basic skills training that is decoupled from specific exam subjects, specifically strengthening fundamental vehicle control skills such as precise speed control, steering correction, starting and stopping, and straight-line driving. This helps learners establish a causal understanding of vehicle operation and dynamic response, cultivating a genuine "driving feel" and completely abandoning the mechanical training model of rote memorization of fixed points found in existing technologies. After mastering the underlying control logic, learners can then advance to more complex training projects, truly understanding the essence of operation. This not only facilitates efficient passing of standardized exams but also significantly enhances their confidence and ability to independently handle complex real-world road conditions in the future, achieving a core transformation in driver training from "exam-oriented" to "ability-oriented."

[0022] Secondly, this invention significantly improves the accuracy and efficiency of overcoming difficulties and training skills. By dividing long, comprehensive driving courses such as reversing into parking spaces into several independent "atomic operation segments" (such as the timing of turning the steering wheel fully, the timing of straightening the steering wheel, and the timing of stopping), and constructing a closed-loop state machine for each segment containing "initialization-execution-evaluation-reset" states, it supports high-density, targeted, and cyclical reinforcement training for single weak skill points. This "sliced" training mode breaks the inefficient cycle of traditional teaching where "one mistake leads to starting over," refining the feedback granularity to specific operational steps. This allows trainees to quickly locate and correct errors, concentrating training time and resources on core weak areas. Compared to the coarse-grained result evaluation of existing technologies, training efficiency is improved by orders of magnitude.

[0023] Furthermore, this invention achieves truly personalized and adaptive teaching, optimizing the overall allocation of teaching resources. By collecting student operation data in real time, and based on multi-dimensional indicators such as vehicle speed standard deviation, steering wheel angular velocity smoothness, parking space centering rate, and average operation error, the system dynamically calculates and constructs a quantitative "student ability profile," allowing the system to objectively and continuously assess the student's true level. Combined with preset advancement thresholds (such as comprehensive score > $Threshold_{Level_1}$, atomic operation segment pass rate > 90%), the system can automatically or semi-automatically execute the teaching level transition, truly achieving "teaching according to aptitude": for students with outstanding basic abilities, they are allowed to skip the basic ability training level or reduce repetitive training, avoiding waste of time and resources; for students with weak basic abilities, it ensures that they remain at the corresponding level for sufficient practice before reaching the target, preventing "forced growth" teaching. This mechanism not only meets the personalized learning needs of different students but also enables the efficient allocation of coaching resources and vehicle equipment according to the actual progress of students, improving the overall utilization rate of teaching resources.

[0024] Finally, this invention enhances system adaptability and student engagement, improving the practicality and flexibility of the teaching plan. It supports multiple task breakdown alternatives, including "decomposition by operation action type" and "decomposition by operation node," adaptable to different comprehensive driving subjects and student operating habits. Simultaneously, it allows students to independently mark and save personalized visual reference points during training. The system records these points and links them to the student's ability profile, integrating them into subsequent personalized guidance. This makes teaching more aligned with different students' cognitive habits and physical characteristics, enhancing the approachability and effectiveness of training. Furthermore, the system is compatible with both "fully automatic advancement" and a semi-automatic advancement mode with "instructor manual review assistance," ensuring intelligent teaching while accommodating necessary human judgment in teaching management, further improving the flexibility of the plan's scenario adaptation.

[0025] In summary, this invention constructs an efficient, accurate, and personalized closed-loop robot driving training system by organically integrating a tiered teaching framework, atomized task decomposition, quantitative competency profiling, and adaptive advancement logic. It effectively solves core problems in existing technologies such as inverted teaching logic, coarse feedback, and rigid pathways. This represents a key technological advancement in the field of driving training, moving from "exam-oriented" to "competency-oriented" approaches, and has broad application value. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in this embodiment or the prior art, the drawings used in the description of the embodiment or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this embodiment. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the adaptive advancement robot driving training method of the present invention. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this embodiment. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this embodiment as detailed in the appended claims.

[0029] like Figure 1As shown, this embodiment provides an adaptive progressive robot driving training method based on ability grading and task decomposition. The method is executed based on a three-layer teaching control system, which includes a basic ability training layer, a task decomposition reinforcement layer, and a comprehensive business logic layer arranged sequentially. The method includes the following steps: S1: In the basic ability training layer, the trainee is controlled to conduct vehicle control ability training decoupled from specific examination subjects; S2: In the task decomposition and reinforcement layer, the target comprehensive driving subject is divided into multiple independent atomic operation segments according to the operation nodes, and the trainee is controlled to perform cyclic reinforcement training on the selected atomic operation segments. S3: In the integrated business logic layer, control the trainee to conduct the complete process training of the target integrated driving subject; During the execution of steps S1 to S3 above, the student's operation data is collected in real time, and the student's ability profile, which includes multiple quantitative ability indicators, is dynamically calculated and updated based on the operation data. Based on the quantitative ability indicators in the student's ability profile, it is automatically determined whether the student meets the conditions for advancing from the current training level to the next training level. When the determination result is met, the student is automatically controlled to advance to the next training level.

[0030] This three-tiered progressive architecture breaks away from the single-mode, one-time training of existing technologies. By first establishing the basics, then breaking them down, and finally integrating them, it achieves the layered construction of driving skills. This avoids the frustration of beginners directly challenging complex projects, while allowing students with a solid foundation to advance efficiently. It solves the problem of the inverted order in existing technology teaching logic from an architectural perspective.

[0031] As one implementation method, the basic ability training layer in this embodiment includes at least one of the following: starting and stopping practice, straight driving practice, and reversing direction adjustment practice. The starting and stopping practice is configured to involve multiple vehicle starts and stops within a short distance, focusing on training the learner's control over clutch engagement and precise braking force. This helps learners establish a basic correlation between "operation action and vehicle response," preventing basic issues such as stalling during starts or excessively abrupt stops from affecting subsequent complex training. The straight driving practice is configured to train learners through a summarized four-step steering wheel operation method. This method is specifically implemented through four standardized steps: aligning the initial vehicle direction, making minor corrections to driving deviations, maintaining a stable straight driving posture, and finally fine-tuning the direction to return to the optimal driving state, controlling the vehicle to any target straight line. This specifically strengthens the smoothness of steering wheel operation and the ability to center the vehicle, abandoning the mechanical driving habits that rely on fixed reference points in existing technologies, and cultivating the learner's core ability for dynamic adjustment. This setup solidifies vehicle control skills from the ground up, providing capability support for subsequent complex subject training and solving the skill gap problem caused by "forcing advancement despite weak foundations" in existing technologies.

[0032] As one implementation method, this embodiment divides the target comprehensive driving subject into multiple independent atomic operation segments according to the operation nodes. Specifically, it includes breaking down the right-hand reversing parking subject into multiple atomic operation segments corresponding to the timing of turning the steering wheel fully, the timing of straightening the steering wheel, and the timing of stopping. Each atomic operation segment focuses on a single skill difficulty. For example, the "timing of turning the steering wheel fully" segment only trains the trainee to judge and execute the accuracy of turning the steering wheel fully based on the relative position of the vehicle body and the parking line, eliminating other operational interference. This decomposition method breaks down the long and complex comprehensive subject into low-difficulty, focusable single skill points, solving the inefficient training problem of "one mistake and the whole thing has to start over" in the existing technology, allowing trainees to target their weaknesses and significantly improve the efficiency of overcoming difficulties.

[0033] As one implementation method, this embodiment performs cyclic reinforcement training on selected atomic operation segments, specifically including: constructing an independent state machine for each atomic operation segment, the state machine sequentially including an initialization state, an operation execution state, an operation evaluation state, and a system reset state. The initialization state is used to reset the vehicle and training scenario to the starting position and initial parameters of the segment, ensuring a consistent training environment each time; the operation execution state collects student operation data in real time and synchronously provides feedback on operation deviations; the operation evaluation state scores the single segment operation based on quantitative indicators, clearly informing the student of errors (e.g., "turning the steering wheel to full position 0.5 seconds too early"); the system reset state automatically resets to the initialization state after each segment training session, supporting cyclic training until the student masters the skill. This closed-loop state machine achieves seamless integration of "training-evaluation-correction-retraining," eliminating the need for students to manually reset or restart the entire process. This not only increases the training density of individual skill points but also helps students quickly correct errors through precise feedback. Compared to the "full-process fuzzy evaluation" of existing technologies, training accuracy and efficiency are significantly improved.

[0034] As one implementation method, the quantitative ability indicators in this embodiment include vehicle control stability score, spatial perception score, and operational accuracy score. The vehicle control stability score is calculated based on the standard deviation of vehicle speed and the smoothness of steering wheel angular velocity during training. A smaller standard deviation and smoother steering wheel angular velocity result in a higher score, accurately reflecting the learner's stable control ability over the vehicle's driving state. The spatial perception score is calculated based on the vehicle's centering rate within the lane, directly reflecting the learner's ability to judge and adjust the vehicle's position. The operational accuracy score is calculated based on the average error between the actual position and the preset position of key points operated on during atomic operation segment training. Key points include parking points; a smaller average error results in a higher score, focusing on the accuracy of operational actions. This three-dimensional quantitative indicator system transforms the learner's driving ability into objective data, replacing the subjective "qualified / unqualified" judgments of existing technologies. It provides precise data support for subsequent adaptive advancement and personalized adjustments, ensuring the fairness and scientific nature of the ability assessment.

[0035] As one implementation method, in this embodiment, the system calculates the trainee's comprehensive ability score by weighted summation, and the calculation formula is as follows: .in, , , These are preset weighting coefficients, which can be dynamically adjusted according to the training focus (e.g., to improve the skills of novice trainees). Weighting, strengthening vehicle control stability training). The vehicle control stability score is given. The spatial perception score is given. The accuracy of the operation is scored. This formula achieves a scientific integration of multi-dimensional indicators, which not only comprehensively reflects the student's overall driving ability, but also has flexible adaptability, which can meet the training needs of different training scenarios and different student groups. Compared with the evaluation of a single indicator, it can more objectively present the student's ability shortcomings and strengths.

[0036] As one implementation method, the advancement conditions in this embodiment are configured as follows: when a trainee's quantitative ability index score in the basic ability training layer meets a first threshold condition, they are determined to meet the conditions for advancement from the basic ability training layer to the task decomposition enhancement layer; when a trainee's training pass rate for all atomic operation segments included in the target comprehensive driving subject in the task decomposition enhancement layer meets a second threshold condition, they are determined to meet the conditions for advancement from the task decomposition enhancement layer to the comprehensive business logic layer. This hierarchical advancement logic ensures that trainees must solidly master the skills of the current level before entering the next level, avoiding the problem of forced advancement due to insufficient ability caused by the "uniform progress" of existing technologies. It guarantees training quality from a process perspective and achieves a scientific teaching rhythm of "advancement upon meeting ability standards."

[0037] In one implementation method, the first threshold condition in this embodiment is represented by the comprehensive ability score. The second threshold condition is expressed as a pass rate greater than 90%; where The preset first-level advancement threshold can be flexibly set according to the basic level of the student group (such as beginners and those retaking the exam). A 90% pass rate threshold ensures that students achieve mastery of each atomic operation segment, avoiding superficial training; while Its configurability allows the system to adapt to different training scenarios, ensuring both the rigor of training standards and the flexibility of scenario adaptation, thus solving the problem that existing technology upgrade standards are fixed and cannot adapt to diverse needs.

[0038] As one implementation method, the steps for determining promotion conditions and controlling promotion in this embodiment are executed in a semi-automatic manner. Specifically, the system generates promotion suggestions based on the quantitative ability indicators and sends them to the coach's end. At the same time, it synchronizes the student's training data, ability profile, and typical operation videos. The coach conducts manual review based on the data and experience. After receiving a manual confirmation instruction from the coach's end, the promotion operation is then executed. This method balances the objectivity of intelligent assessment with the flexibility of human experience. In special cases (such as when a student's data meets the standards but their operating habits have potential problems), the coach can intervene and adjust in a timely manner, avoiding the problem of "data meeting the standards but insufficient ability" that may occur with purely automatic promotion, thus improving the reliability and controllability of teaching management.

[0039] As one implementation method, the method of segmenting atomic operation fragments by operation node in this embodiment can be replaced by decomposing by operation action type. When splitting by operation action type, the comprehensive driving subject is divided into clutch operation fragment, steering wheel control fragment, and vehicle speed adjustment fragment. The two splitting methods can be flexibly switched according to the learner's learning habits. For example, for learners with weak spatial perception ability, splitting by operation node (focusing on position judgment) is used; for learners with non-standard operation actions, splitting by operation action type (focusing on action standardization) is used. This alternative solution enriches the dimensions of task decomposition, making teaching more suitable for the different learners' ability weaknesses and learning habits, improving the system's adaptability and personalization level, and solving the problem that existing technical training models are rigid and cannot adapt to diverse learning needs.

[0040] As one implementation method, in this embodiment, trainees can independently mark personalized visual reference points during atomic operation segment training. The system records these personalized visual reference points and associates them with the trainee's ability profile. In subsequent training, the system adapts these personalized visual reference points to provide operational guidance to the trainees. Different trainees have different heights and sitting postures, and fixed reference points cannot fit all trainees. This setting allows trainees to determine reference points according to their own situation (such as using the specific alignment relationship between their shoulders and the library line as a reference point). The system pushes targeted operation prompts based on this reference point, which not only retains the auxiliary role of reference points, but also solves the problem of poor adaptability caused by the "one-size-fits-all" approach of fixed reference points in existing technologies, improving the comfort and accuracy of trainee operations and enhancing training participation.

[0041] As one implementation method, the adaptive adjustment strategy for loading training content in this embodiment includes: increasing the training duration and number of cycles for training content corresponding to indicators with scores below a preset threshold in the student's ability profile, thus strengthening weak areas; reducing the number of repeated training sessions for training content corresponding to indicators with scores above the preset threshold, avoiding ineffective time consumption; and allowing students with outstanding basic abilities to skip the basic ability training layer directly, with the system unlocking the task decomposition reinforcement layer and starting training. This strategy achieves true "personalized instruction," ensuring that students with weak foundations fully address their shortcomings while saving time and improving efficiency for students with outstanding abilities. It also optimizes the allocation of coaching resources and vehicle equipment, avoiding resource waste and solving the resource allocation imbalance problem of "strong students wasting resources while weak students fall behind" caused by the unified teaching path in existing technologies.

[0042] Furthermore, the three-layer teaching control system described in this embodiment adopts a pyramid-shaped architecture, and the definitions and linkage logic of each layer are as follows: The first layer is the Fundamental Capability Layer, which focuses on pure vehicle control practice, stripped of specific exam requirements and not tied to test locations, concentrating on building fundamental vehicle control skills. Specific training exercises include starting and stopping drills, straight-line driving drills, and reversing direction adjustment drills. Starting and stopping drills strengthen students' basic vehicle starting and stopping skills through multiple short-distance starts and stops; straight-line driving drills cultivate students' ability to precisely control the vehicle to any target straight line through a four-step standardized steering wheel operation method, solidifying dynamic adjustment and directional control skills; and reversing direction adjustment drills specifically enhance students' core ability to correct vehicle posture during reversing. The second layer is the Decomposed Task Layer, which decomposes comprehensive driving skills such as right-hand reverse parking into several independent atomic segments based on operational nodes. Taking the right-hand reverse parking exercise as an example, it can be broken down into core atomic segments such as the timing of turning the steering wheel fully, the timing of straightening the steering wheel, and the timing of stopping. Trainees can conduct repeated reinforcement training on single operational timings, while independently marking and determining personalized visual reference points suitable for their own height and sitting posture. The system links these points to the trainee's ability profile, providing targeted guidance for subsequent training and enabling precise breakthroughs in skill difficulties. The third layer is the Integrated Subject Layer, which replicates the complete process of standard exams such as Subject 2 and Subject 3, echoing the existing full-process simulation teaching model. However, this layer has strict hierarchical linkage restrictions—the system only unlocks the teaching content of this layer after the student has achieved the training score of all atomic operation segments in the second layer. This ensures that students first master the decomposition skills and then carry out comprehensive applications, avoiding the logical flaw of the existing technology of "directly getting started with the whole process". To support the aforementioned pyramid-shaped architecture and hierarchical linkage, the system synchronously maintains a dynamically updated student profile. This profile includes three core quantitative indicators, all calculated based on real-time collected sensor data to ensure the objectivity of the assessment: First, the vehicle control stability score, calculated based on the standard deviation of vehicle speed and the smoothness of steering wheel angular velocity during training; a higher value indicates a more stable vehicle driving state. Second, the spatial perception score, calculated based on the vehicle's centering rate within the lane, directly reflecting the student's ability to judge and adjust the vehicle's spatial position. Third, the operational accuracy score, calculated based on the average error between the actual operational position and the preset position of key points such as parking points during atomic operation segment training; a smaller error indicates higher accuracy. The upgrade logic in this embodiment is based on the aforementioned quantitative indicators and level requirements: when a trainee's comprehensive ability score at the basic ability training level... When the system automatically distributes the teaching configuration file for the second layer (task decomposition reinforcement layer), it unlocks the training of atomic operation segments; when the trainee's pass rate for all atomic operation segments in the second layer is >90%, the system automatically unlocks the third layer (comprehensive business logic layer), allowing the full-process comprehensive training to be carried out. Based on the aforementioned technical features, this invention possesses significant advantages over existing technologies, which can be summarized as follows: First, it achieves a core transformation from "exam-oriented" to "ability-oriented." Through basic ability training, students first master fundamental skills such as "correcting direction" and "speed control," and then progress to more advanced subjects like reversing into a parking space. It eliminates reliance on rote memorization of reference points, instead focusing on understanding the causal relationship between vehicle trajectory and operation. This not only improves the pass rate but also cultivates students' ability to independently handle real road conditions, addressing the pain point of existing technologies where "licensed drivers are afraid to drive on the road." Second, it significantly improves the efficiency of overcoming difficulties. The training of atomic operation segments breaks the inefficient cycle of "one mistake and the whole thing starts over," allowing students to specifically strengthen their weak steps. The training density is 5-10 times higher than the existing full-process mode, significantly shortening the training cycle. Thirdly, it enables personalized teaching for each student. Based on the adaptive logic of student ability profiles, it allows students with outstanding comprehension to skip the basic level, while forcing students with weak foundations to advance to the next level after meeting the standards at the corresponding level. This avoids high-achieving students wasting time "running alongside" and prevents students with weak foundations from being "forced to grow too fast," thus optimizing the efficiency of teaching resource allocation.

[0043] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An adaptive promotion robot driving training method based on ability grading and task decomposition, characterized in that: The method is executed based on a three-layer teaching control system, which includes a basic skills training layer, a task decomposition and reinforcement layer, and a comprehensive business logic layer arranged sequentially. The method includes the following steps: S1 controls the trainee to perform vehicle control ability training that is decoupled from specific exam subjects at the basic ability training layer. S2, in the task decomposition and reinforcement layer, divides the target comprehensive driving subject into multiple independent atomic operation segments according to the operation nodes, and controls the trainee to perform cyclic reinforcement training on the selected atomic operation segments. S3 controls the trainee to complete the training process of the target comprehensive driving subject at the integrated business logic layer. During the execution of steps S1 to S3 above, the student's operation data is collected in real time, and the student's ability profile, which includes multiple quantitative ability indicators, is dynamically calculated and updated based on the operation data. Based on the quantitative ability indicators in the trainee's ability profile, the system automatically determines whether the trainee meets the conditions for advancing from the current training level to the next training level; when the determination result is met, the system automatically controls the trainee to advance to the next training level.

2. The method of claim 1, wherein: The training items in the basic ability training layer include at least one of starting and stopping practice, straight driving practice, and reversing direction adjustment practice; The starting and stopping exercise is configured to involve multiple vehicle starts and stops within a short distance to train the trainee's basic vehicle control skills. The straight-line driving exercise is configured to train trainees to control the vehicle onto any target straight line using a deduced four-step steering wheel operation method.

3. The method of claim 1, wherein: The target comprehensive driving subject is divided into multiple independent atomic operation segments according to the operation nodes. Specifically, the right reverse parking subject is broken down into multiple atomic operation segments corresponding to the timing of turning the steering wheel to full, the timing of straightening the steering wheel, and the timing of stopping.

4. The method according to claim 1 or 3, characterized in that: The selected atomic operation segments are subjected to iterative reinforcement training, which specifically includes: constructing an independent state machine for each atomic operation segment, wherein the state machine sequentially includes an initialization state, an operation execution state, an operation evaluation state, and a system reset state, and running the state machine to achieve iterative training for a single skill point.

5. The method of claim 1, wherein: The quantitative capability indicators include vehicle control stability score, spatial perception score, and operational accuracy score. The vehicle control stability score is calculated based on the standard deviation of vehicle speed and the smoothness of steering wheel angular velocity during the training process; the spatial perception score is calculated based on the vehicle's centering rate within the lane. The operation accuracy score is calculated based on the average error between the actual position and the preset position of the key point during the training of the atomic operation segment, and the key point includes the parking point.

6. The method according to claim 5, characterized in that: The system calculates the student's overall ability score through weighted summation. The calculation formula is as follows: in, , , The preset weighting coefficients, The vehicle control stability score is given. The spatial perception score is given. The accuracy of the operation is scored.

7. The method according to claim 1 or 6, characterized in that: The advancement condition is configured as follows: when a trainee's quantitative ability index score in the basic ability training layer meets the first threshold condition, it is determined that the trainee meets the condition to advance from the basic ability training layer to the task decomposition enhancement layer. When a trainee's pass rate for all atomic operation segments included in the target comprehensive driving subject in the task decomposition enhancement layer meets the second threshold condition, it is determined that the trainee meets the condition for promotion from the task decomposition enhancement layer to the comprehensive business logic layer.

8. The method according to claim 7, characterized in that: The first threshold condition is represented by the comprehensive ability score. The second threshold condition is expressed as a pass rate greater than 90%; where This is the preset threshold for advancing to the first level.

9. The method according to claim 1 or 3, characterized in that: During the training of atomic operation segments, trainees can independently mark personalized visual reference points. The system records these personalized visual reference points and associates them with the trainee's ability profile. In subsequent training, the system adapts these personalized visual reference points to provide trainees with operation guidance.

10. The method according to claim 1, characterized in that: The adaptive adjustment strategy for loading training content includes: For training content corresponding to indicators whose scores in the trainee's ability profile are below the preset threshold, increase the training duration and number of cycles. For training content corresponding to indicators with scores higher than a preset threshold, reduce the number of repeated training sessions. For trainees with outstanding basic abilities, they can skip the basic ability training layer directly and have the system unlock the task decomposition enhancement layer and begin training.

Images