An artificial intelligence-based driving robot driving behavior analysis and correction system

Abstract

The application provides a driving behavior analysis and correction system based on artificial intelligence, and relates to the technical field of driving models, comprising an operation environment acquisition module, a time sequence segment generation module, a reasonable interval determination module, a behavior attribute discrimination module and a hierarchical correction output module; the system synchronously acquires student operation data and road environment data, generates a driving behavior time sequence segment set, extracts an operation change gradient and an environment risk change amount, and generates an operation rationality interval containing a deceleration amplitude range, a brake advance amount range and a steering change range according to the environment risk change amount; then the operation change gradient and the operation rationality interval are compared and matched section by section, active avoidance behaviors and non-standard operation behaviors are distinguished, and hierarchical correction instructions are output according to deviation amounts; the application can improve the accuracy of driving behavior analysis and correction under complex traffic situations.

Description

Technical Field

[0001] This invention relates to the field of driving model technology, and in particular to an artificial intelligence-based driving training robot driving behavior analysis and correction system. Background Technology

[0002] Existing technologies primarily rely on sensor- and rule-based driver training and assessment systems. These systems typically deploy cameras, steering wheel angle sensors, and accelerator / brake pedal sensors in training vehicles or simulators to collect real-time data on student actions. This data is then analyzed and scored using pre-defined scoring rules (such as lane departure thresholds, braking timing, and steering amplitude). Simultaneously, some systems incorporate simple machine learning models to categorize driving behaviors, such as identifying violations like sudden braking and lane departure, and provide corrections via voice or screen prompts, thus assisting instructors in teaching and assessment.

[0003] However, in real-world driver training scenarios, such as urban road simulation training, when trainees attempt to make left turns at complex intersections, existing systems often rely solely on steering wheel angle and vehicle speed thresholds for judgment, failing to comprehensively analyze the surrounding traffic environment (such as pedestrian dynamics and oncoming vehicle speeds). This can lead the system to misinterpret trainees' deceleration or steering adjustments to avoid pedestrians as "improper operation," issuing incorrect correction prompts, interfering with trainees' judgment, and even fostering incorrect driving habits, thus reducing training effectiveness. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based driver training robot driving behavior analysis and correction system, which aims to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: A driving behavior analysis and correction system for driver training robots based on artificial intelligence, the system comprising: The operating environment acquisition module is used to simultaneously collect student operation data and road environment data. The student operation data includes the change in steering wheel angle, the rate of change in accelerator pedal opening, and the duration of brake pedal action. The road environment data includes the distance to the target pedestrian, the speed of the target pedestrian, the relative speed of surrounding vehicles, and the boundary information of the road passage area, generating a corresponding dataset for the operating environment. The time-series segment generation module is used to synchronously segment student operation changes and environmental changes according to a preset time window based on the dataset corresponding to the operation environment, extract the operation change gradient and environmental risk change amount within each time window, and generate a set of driving behavior time-series segments. The operation change gradient includes steering change gradient, driving change gradient and braking change gradient. The reasonable interval determination module is used to determine the safe operating range corresponding to each driving behavior time sequence segment based on the environmental risk change in the driving behavior time sequence segment set, and generate a set of reasonable operating intervals including deceleration range, braking advance range and steering change range; The behavior attribute discrimination module is used to compare the operation change gradient in the time sequence of driving behavior segments with the operation rationality interval set segment by segment. When the operation change trend matches the environmental risk change trend and the operation change gradient falls into the corresponding operation rationality interval, an active avoidance behavior discrimination result is generated. When the operation change gradient exceeds the corresponding operation rationality interval, a non-standard operation behavior discrimination result is generated. The graded correction output module is used to perform corrective control based on the judgment results of active avoidance behavior and non-standard operation behavior. When generating non-standard operation behavior judgment results, it outputs graded correction instructions according to the deviation amount exceeding the corresponding operation reasonableness range. When generating active avoidance behavior judgment results, it suppresses the output of misjudgment prompts.

[0006] The above-described solution of the present invention has at least the following beneficial effects: First, by synchronously collecting student operation data and road environment data, and constructing a time-series set of driving behavior segments, the driving operation is correlated with the environmental changes within the corresponding time period. Compared with the method of judging solely based on operation thresholds, this enables driving behavior analysis to have environmental semantic support and reflect the specific background of the operation. Based on this, by extracting the changes in environmental risks and generating the reasonableness range of operations, the deceleration magnitude, braking timing and steering range are uniformly incorporated into the dynamic range constraint, so that the driving behavior assessment is transformed from fixed rules to range judgment that matches the actual traffic situation, thereby avoiding misjudging reasonable avoidance behavior as illegal operation. Furthermore, by matching the trend of operational changes with the trend of environmental risk changes, and combining this with interval constraints for joint judgment, the system can identify whether the student's operation is a response to environmental risk, thereby distinguishing between proactive avoidance behavior and non-standard operating behavior, and reducing erroneous intervention in normal driving adjustments. Finally, by outputting graded correction instructions based on deviation and suppressing prompt output when the behavior is determined to be active avoidance, the correction process is kept consistent with the driving situation. This not only provides a basis for adjustment based on actual deviations but also avoids interfering with reasonable operations, thereby improving the accuracy of behavior analysis and the effectiveness of guidance in the driver training process. Attached Figure Description

[0007] Figure 1 This is an architecture diagram of an artificial intelligence-based driver training robot driving behavior analysis and correction system provided by an embodiment of the present invention. Detailed Implementation

[0008] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0009] like Figure 1 As shown, an embodiment of the present invention proposes a driving behavior analysis and correction system for a driver training robot based on artificial intelligence, the system comprising: The operating environment acquisition module is used to simultaneously collect student operation data and road environment data. The student operation data includes the change in steering wheel angle, the rate of change in accelerator pedal opening, and the duration of brake pedal action. The road environment data includes the distance to the target pedestrian, the speed of the target pedestrian, the relative speed of surrounding vehicles, and the boundary information of the road passage area, generating a corresponding dataset for the operating environment. The time-series segment generation module is used to synchronously segment student operation changes and environmental changes according to a preset time window based on the dataset corresponding to the operation environment, extract the operation change gradient and environmental risk change amount within each time window, and generate a set of driving behavior time-series segments. The operation change gradient includes steering change gradient, driving change gradient and braking change gradient. The reasonable interval determination module is used to determine the safe operating range corresponding to each driving behavior time sequence segment based on the environmental risk change in the driving behavior time sequence segment set, and generate a set of reasonable operating intervals including deceleration range, braking advance range and steering change range; The behavior attribute discrimination module is used to compare the operation change gradient in the time sequence of driving behavior segments with the operation rationality interval set segment by segment. When the operation change trend matches the environmental risk change trend and the operation change gradient falls into the corresponding operation rationality interval, an active avoidance behavior discrimination result is generated. When the operation change gradient exceeds the corresponding operation rationality interval, a non-standard operation behavior discrimination result is generated. The graded correction output module is used to perform corrective control based on the judgment results of active avoidance behavior and non-standard operation behavior. When generating non-standard operation behavior judgment results, it outputs graded correction instructions according to the deviation amount exceeding the corresponding operation reasonableness range. When generating active avoidance behavior judgment results, it suppresses the output of misjudgment prompts.

[0010] In this embodiment of the invention, by synchronously collecting student operation data and road environment data, a dataset corresponding to the operation environment is formed. Then, a set of driving behavior time sequence segments is generated according to a preset time window. The environmental risk change and operation change gradient are extracted from the driving behavior time sequence segment set. This allows the student's steering, driving and braking operations to establish a correspondence with the road environment changes within the corresponding time period. This enables driving behavior analysis to move beyond judging a single operation value and instead identify the correlation between environmental changes and operation changes within the same time period, thereby improving the accuracy of identifying the causes of driving behavior.

[0011] By determining the operational rationality range, which includes the deceleration range, braking advance range, and steering change range, based on the changes in environmental risk, and comparing the operational change gradient with the operational rationality range segment by segment, different safe operating ranges can be formed for different driving behavior time segments. This transforms the behavior judgment process from fixed threshold judgment to range judgment corresponding to changes in the actual scenario, thereby reducing misjudgments of evasive deceleration, preventive braking, and steering adjustments constrained by road boundaries.

[0012] By generating proactive avoidance behavior judgment results when the trend of operational changes matches the trend of environmental risk changes and the operational change gradient falls within the operational rationality range, and generating non-standard operational behavior judgment results when the operational change gradient exceeds the operational rationality range, it is possible to distinguish between normal driving adjustments caused by risk changes and abnormal operations that deviate from scenario constraints. This avoids judging the safety responses made by trainees to pedestrians, surrounding vehicles, and road boundaries as incorrect actions, thereby improving the consistency between driving behavior analysis results and actual traffic situations.

[0013] By implementing corrective control based on the results of active avoidance behavior and non-standard operation behavior, and by outputting graded corrective instructions according to the deviation amount exceeding the reasonableness range when generating non-standard operation behavior results, and suppressing the output of misjudgment prompts when generating active avoidance behavior results, the corrective information can correspond to the trainee's current operation deviation, avoid outputting interfering prompts during reasonable avoidance, and provide corresponding adjustment guidance based on the degree of deviation, making it easier for trainees to complete corrections according to the current scenario.

[0014] The following explanation is based on specific scenarios: When a trainee drives a training vehicle to an intersection with a pedestrian crossing and a left-turn lane, road environment data reflects a shortening distance to the target pedestrian, changes in the target pedestrian's speed, and a narrowing of the road's traffic area boundary. The system generates corresponding environmental risk changes. During this period, the trainee reduces the accelerator pedal opening, extends the brake pedal action time, and controls the steering wheel angle change. The system compares the corresponding operational change gradient with the reasonableness range of the operation during this period. If the braking and steering adjustments fall within the deceleration range, braking advance range, and steering change range, it is judged as an active avoidance behavior, and a false judgment prompt is output. If the braking intervention is delayed or the steering change exceeds the steering change range, it is judged as an non-standard operation, and a corresponding correction instruction is output based on the deviation, helping the trainee understand the appropriate operation under the conditions of this intersection.

[0015] In a preferred embodiment of the present invention, the operating environment acquisition module includes: Set up steering wheel angle acquisition device, accelerator pedal displacement acquisition device, brake pedal status acquisition device, forward environment perception device and road boundary recognition device on training vehicles or simulation training equipment, and configure a unified time reference for each acquisition device so that each acquisition device outputs corresponding data according to the same sampling rhythm. The continuous steering wheel position data output by the steering wheel angle acquisition device is compared with adjacent sampling times to determine the angle change between the current sampling time and the previous sampling time, and this angle change is recorded as the steering wheel angle change. The continuous accelerator pedal position data output by the accelerator pedal displacement acquisition device is read sequentially over time to determine the rate of change of the accelerator pedal opening within a unit sampling time, and this rate of change is recorded as the accelerator pedal opening change rate. The pedal trigger state and continuous state output by the brake pedal state acquisition device are continuously recorded to determine the duration of the brake pedal from the start of action to the release of action, and this duration is recorded as the brake pedal action time. The system performs target recognition processing on image information, distance information, and target motion information acquired by the forward-facing environmental perception device to identify pedestrian targets, surrounding vehicle targets, and road area targets. For identified pedestrian targets, the system reads the real-time distance between the training vehicle and the pedestrian target and determines the pedestrian's movement speed based on the position information changes over continuous sampling time. For identified surrounding vehicle targets, the system reads the relative motion state between the training vehicle and surrounding vehicles and determines the relative speed of surrounding vehicles based on the relative position and relative speed changes over continuous sampling time. For road area targets, the system identifies lane boundary lines, stop lines, turning guide lines, and road edge contours and determines the road traffic area boundary information based on the recognition results. The steering wheel angle change, accelerator pedal opening change rate, brake pedal action time, target pedestrian distance, target pedestrian movement speed, relative speed of surrounding vehicles, and road traffic area boundary information are aligned in a uniform time sequence, and data with missing or abnormal fluctuations are screened for validity. For valid data, they are written into the same data recording unit according to a preset data format, so that each record contains the student operation information and road environment information corresponding to the same sampling time, generating a dataset corresponding to the operation environment.

[0016] In a preferred embodiment of the present invention, the time segment generation module includes: The window partitioning unit is used to continuously divide the student operation data and road environment data according to the dataset corresponding to the operation environment, and generate a time window sequence by sliding time windows of fixed length. The gradient generation unit is used to generate a steering change gradient by performing continuous difference processing on the steering wheel angle change in each time window sequence, generating a drive change gradient by performing continuous difference processing on the accelerator pedal opening change rate in each time window, and generating a braking change gradient by performing cumulative change processing on the brake pedal application time in each time window. The braking change gradient is used to characterize the degree of braking response; the continuous increase, maintenance, or shortening of the brake pedal application time corresponds to the enhancement, maintenance, or weakening of the braking response, respectively. The risk component generation unit is used to correlate the change in the distance of the target pedestrian with the moving speed of the target pedestrian in each time window sequence to generate a pedestrian risk component, extract the change in the relative speed of the surrounding vehicles in each time window to generate a vehicle risk component, and identify the boundary information of the road passage area in each time window to generate a spatial constraint risk component. The risk quantity synthesis unit is used to perform weighted combination processing based on pedestrian risk components, vehicle risk components and spatial constraint risk components to generate environmental risk change quantities corresponding to each time window. The segment output unit is used to bind the steering change gradient, driving change gradient, braking change gradient and environmental risk change one by one, and arrange them in order according to the time window sequence to generate a set of driving behavior time sequence segments.

[0017] In this embodiment of the invention, student operation data and road environment data are continuously divided according to a sliding time window of fixed length. Steering change gradient, driving change gradient, and braking change gradient representing the degree of braking response are generated in each time window. At the same time, the change in target pedestrian distance, the change in relative speed of surrounding vehicles, and the boundary information of the road passage area are transformed into pedestrian risk components, vehicle risk components, and spatial constraint risk components. Furthermore, environmental risk change is generated by weighted combination. Then, each type of operation change gradient is bound to the environmental risk change to form a set of driving behavior time sequence segments. This can realize the synchronous modeling of driving behavior and environmental changes, so that the operation behavior in each time window has clear environmental semantic support. This avoids the problem of fragmentation caused by relying on only a single operation data for analysis, and provides a structured data foundation with temporal continuity for subsequent interval generation and behavior discrimination.

[0018] In a preferred embodiment of the present invention, the operational gradient generation unit includes: The steering wheel angle change data within each time window is read sequentially according to the time window sequence. The steering wheel angle change at adjacent sampling times within the same time window is compared to determine the increasing, decreasing, and holding states of the steering wheel angle change within each sampling interval. The change states of each sampling interval within the time window are then summarized sequentially to generate the steering change gradient corresponding to the time window. The accelerator pedal opening change rate data is read sequentially within each time window sequence, and the accelerator pedal opening change rate at adjacent sampling times within the time window is continuously compared to determine the decreasing trend, increasing trend, and stable trend of the accelerator pedal opening within that time window. When multiple consecutive sampling times show opening contraction, it is recorded as a driving weakening process; when multiple consecutive sampling times show opening expansion, it is recorded as a driving strengthening process. Then, the corresponding driving change gradient is generated based on the overall change state of each sampling time within the time window. The brake pedal application time data is read sequentially within each time window sequence, and the braking duration corresponding to each sampling moment within the time window is accumulated in chronological order to determine the continuous change process of the brake pedal application within that time window. When the braking duration at a subsequent sampling moment is longer than that at a previous sampling moment, it is recorded as a braking response enhancement process; when the braking duration at a subsequent sampling moment is shorter than that at a previous sampling moment, it is recorded as a braking response weakening process; when the braking duration at consecutive sampling moments remains consistent, it is recorded as a braking response maintenance process. Then, a braking change gradient characterizing the degree of braking response is generated based on the continuous change process within the time window. The steering change gradient, drive change gradient, and braking change gradient generated within the same time window are uniformly identified and bound to the start and end times of that time window for storage. This allows each type of operation gradient to reflect the changes in steering, drive, and braking response made by the student within that time window.

[0019] In a preferred embodiment of the present invention, the risk component generation unit includes: The system sequentially reads the target pedestrian distance and movement speed data within each time window. First, it determines the distance between the target pedestrian and the training vehicle within that time window: shortening, maintaining, or increasing. Then, it correlates this with the pedestrian's movement along the training vehicle's direction of travel or turning path. When the target pedestrian distance continuously shortens and the pedestrian's movement direction enters the training vehicle's designated passage area, the time window is recorded as an enhanced pedestrian risk state. When the target pedestrian distance remains stable and the pedestrian does not enter the training vehicle's designated passage area, the time window is recorded as a maintained pedestrian risk state. When the target pedestrian distance increases and the pedestrian moves away from the training vehicle's designated passage area, the time window is recorded as a reduced pedestrian risk state. Finally, it generates pedestrian risk components based on the results within each time window. The relative speed data of surrounding vehicles within each time window is read sequentially according to the time window sequence. The relative motion states of surrounding vehicles at adjacent sampling times within the same time window are compared to determine the relative approach state, relative distance state, and relative stable state of surrounding vehicles. When the relative approach speed between surrounding vehicles and training vehicles continues to increase, the time window is recorded as the vehicle risk enhancement state. When the relative position change of surrounding vehicles remains within a preset stable range, the time window is recorded as the vehicle risk maintenance state. When the relative distance between surrounding vehicles and training vehicles continues to widen, the time window is recorded as the vehicle risk reduction state. Then, the vehicle risk component is generated based on the state results within the time window. The road traffic area boundary information within each time window is read sequentially according to the time window sequence, and the lane boundary position, road edge position, stop line position, and turning traffic area range within that time window are identified. When the lateral traffic width decreases, the road turning guidance area is restricted, or the space for vehicles to avoid obstacles is reduced, the time window is recorded as a state of enhanced spatial constraints. When the road boundary is stable and the traffic range remains unchanged, the time window is recorded as a state of maintained spatial constraints. When the lateral traffic width increases or the adjustable traffic area increases, the time window is recorded as a state of weakened spatial constraints. Then, a spatial constraint risk component is generated based on the state results within the time window. The pedestrian risk component, vehicle risk component, and spatial constraint risk component generated within the same time window are uniformly bound together, and their corresponding risk source types and risk status results are marked respectively. This allows each risk component to jointly reflect the comprehensive environmental constraints on the driving behavior of the training vehicle caused by pedestrian dynamics, the movement status of surrounding vehicles, and changes in road traffic boundaries within the same time window.

[0020] In a preferred embodiment of the present invention, the reasonable interval determination module includes: The deceleration demand generation unit is used to perform hierarchical mapping processing based on the environmental risk changes in the driving behavior time sequence segment set, determine the minimum deceleration demand value for each driving behavior time sequence segment, and generate deceleration demand results. The braking advance range generation unit is used to perform pre-constraint processing on the braking intervention time point based on the changing trend of environmental risk changes, and generate the braking advance range. The turning boundary generation unit is used to limit the lateral passable range based on the spatial constraint information corresponding to the environmental risk change, and generate the turning change upper limit. The longitudinal constraint synthesis unit is used to perform linkage constraint processing based on the deceleration demand result and the braking advance range to generate the longitudinal operation constraint result. The interval output unit is used to perform fusion and limitation processing based on the longitudinal operation constraint results and the upper limit of steering change to generate the operation rationality interval corresponding to each driving behavior time sequence segment. The operation rationality interval includes the deceleration range, the braking advance range, and the steering change range.

[0021] In this embodiment of the invention, the minimum deceleration requirement is determined by hierarchical mapping based on the change in environmental risk, and the braking intervention timing is pre-constrained based on the trend of environmental risk change. At the same time, the upper limit of steering change is limited by combining the spatial constraint information reflected in the change in environmental risk. Then, the longitudinal operation constraint result is generated by linking the deceleration requirement result with the braking advance range, and further fused with the upper limit of steering change to form an operational rationality range. The impact of environmental changes on longitudinal control and lateral control can be uniformly incorporated into the same constraint system, so that a correlation constraint relationship is formed between deceleration behavior, braking timing and steering adjustment. Thus, the generated operational rationality range simultaneously reflects the speed control requirements and path constraints under the current driving situation, avoiding judgment bias caused by considering only a single control dimension.

[0022] In a preferred embodiment of the present invention, the deceleration demand generation unit includes: According to the time sequence of the driving behavior time sequence set, the environmental risk change corresponding to each driving behavior time sequence segment is read sequentially, and the change state of the environmental risk change between consecutive driving behavior time sequence segments is identified to determine whether the current driving behavior time sequence segment is in a risk-increasing state, a risk-maintaining state, or a risk-decreasing state. When the environmental risk change is higher in the current driving behavior time sequence segment than in the previous driving behavior time sequence segment, the current driving behavior time sequence segment is recorded as a risk-increasing segment. When the environmental risk change is consistent with the previous driving behavior time sequence segment, the current driving behavior time sequence segment is recorded as a risk-maintaining segment. When the environmental risk change is lower than in the previous driving behavior time sequence segment, the current driving behavior time sequence segment is recorded as a risk-decreasing segment. Based on the risk status corresponding to the change in environmental risk, the deceleration demand for each driving behavior time segment is graded. When the current driving behavior time segment is identified as a risk-increasing segment, the deceleration demand corresponding to that driving behavior time segment is set to a higher level of deceleration demand. When the current driving behavior time segment is identified as a risk-maintaining segment, the deceleration demand corresponding to that driving behavior time segment is set to the current maintaining deceleration demand. When the current driving behavior time segment is identified as a risk-decreasing segment, the deceleration demand corresponding to that driving behavior time segment is set to a lower level of deceleration demand. The deceleration demand levels corresponding to each driving behavior time sequence segment are sequentially mapped so that each deceleration demand level corresponds to a minimum deceleration execution requirement; wherein, the higher the deceleration demand level, the higher the corresponding minimum deceleration execution requirement; the lower the deceleration demand level, the lower the corresponding minimum deceleration execution requirement; and the minimum deceleration execution requirement is determined as the minimum deceleration demand value of the current driving behavior time sequence segment. The minimum deceleration requirement value corresponding to each driving behavior time sequence segment is stored in chronological order and bound to the corresponding driving behavior time sequence segment to generate deceleration requirement results. This enables the subsequent longitudinal constraint establishment process to determine the deceleration amplitude constraint range based on the minimum deceleration requirement value corresponding to each driving behavior time sequence segment.

[0023] In a preferred embodiment of the present invention, the braking advance range generation unit includes: According to the time sequence of the driving behavior time sequence set, the environmental risk change corresponding to each driving behavior time sequence segment is continuously read, and the direction of change of environmental risk change between adjacent driving behavior time sequence segments is continuously compared to determine whether the environmental risk change has a continuous upward trend, a stable trend, or a continuous downward trend in the current time period. When the environmental risk change of multiple consecutive driving behavior time sequence segments increases sequentially, the consecutive time period is recorded as a period of continuous risk increase; when the environmental risk change of multiple consecutive driving behavior time sequence segments remains unchanged, the consecutive time period is recorded as a period of stable risk; when the environmental risk change of multiple consecutive driving behavior time sequence segments decreases sequentially, the consecutive time period is recorded as a period of continuous risk decrease. Based on the changing trend of environmental risk, the braking intervention time points corresponding to each driving behavior time segment are subject to pre-constraint processing. When the current driving behavior time segment is in a period of continuously increasing risk, the braking intervention time point is moved to the beginning of the current driving behavior time segment so that the braking response begins to build before the risk accumulates further. When the current driving behavior time segment is in a period of stable risk, the braking intervention time point is limited to the middle region of the current driving behavior time segment. When the current driving behavior time segment is in a period of continuously decreasing risk, the braking intervention time point is limited to the latter region of the current driving behavior time segment. The timing of braking intervention after pre-positioning is ranged to determine the earliest and latest time periods during which braking action can begin. When the rate of increase in environmental risk changes accelerates, the earliest time period is further adjusted forward, while the latest time period is simultaneously compressed, thus narrowing the range of braking advance. When the rate of increase in environmental risk changes remains stable, the earliest and latest time periods are maintained within the preset range. When the rate of increase in environmental risk changes decreases, the earliest and latest time periods are adjusted backward as a whole, thus shifting the range of braking advance. The braking advance range corresponding to each driving behavior time segment is bound and stored in chronological order with the corresponding driving behavior time segment to generate the braking advance range, so that the subsequent longitudinal constraint synthesis process can determine the allowable intervention time interval of braking action according to different risk change trends.

[0024] In a preferred embodiment of the present invention, the steering boundary generation unit includes: Spatial constraint information related to changes in road traffic area boundaries is extracted from the environmental risk changes corresponding to each driving behavior time segment. Combined with the lane boundary position, road edge position, stop line position, and turnable traffic area range within the corresponding time period, the lateral traffic range corresponding to the current driving behavior time segment is determined. When the traffic width between the lane boundary and the road edge decreases, the driving behavior time segment is recorded as a spatial contraction segment; when the traffic width between the lane boundary and the road edge remains unchanged, the driving behavior time segment is recorded as a spatial stability segment; when the traffic width between the lane boundary and the road edge increases, the driving behavior time segment is recorded as a spatial widening segment. Based on the status of the lateral passable range, the allowable steering wheel adjustment range within the current driving behavior time segment is limited. When the current driving behavior time segment is a spatial contraction segment, the allowable lateral steering adjustment boundary is reduced to keep the vehicle steering action within a narrower passable area. When the current driving behavior time segment is a spatial stability segment, the preset lateral steering adjustment boundary is maintained. When the current driving behavior time segment is a spatial expansion segment, the allowable lateral steering adjustment boundary is expanded to allow the vehicle to adjust within a larger lateral space. The upper limit of the defined lateral steering adjustment boundary is extracted to determine the maximum allowable adjustment range of the steering wheel angle within the current driving behavior time segment, and this maximum adjustment range is determined as the steering change upper limit; where the narrower the lateral passable range, the smaller the steering change upper limit; the wider the lateral passable range, the larger the steering change upper limit. The upper limit of steering change corresponding to each driving behavior time sequence segment is stored in chronological order and bound to the corresponding driving behavior time sequence segment to generate the upper limit of steering change. This enables the subsequent interval output process to limit the range of steering change in the reasonable operation interval based on the lateral boundary limit corresponding to each driving behavior time sequence segment.

[0025] In a preferred embodiment of the present invention, the behavior attribute discrimination module includes: The trend correspondence extraction unit is used to extract the correspondence between the risk change direction and the braking change direction in each driving behavior time sequence segment based on the environmental risk change amount and braking change gradient in the driving behavior time sequence segment set, and generate trend correspondence relationship. The continuous consistency verification unit is used to verify the trend continuity status between adjacent driving behavior time segments according to the trend correspondence relationship and generate a trend consistency judgment result. The interval conformity determination unit is used to compare the steering change gradient, driving change gradient and braking change gradient in the driving behavior time sequence segment set with the corresponding operation rationality intervals one by one to generate the interval conformity determination result. The avoidance result generation unit is used to make a joint judgment based on the trend consistency judgment result and the interval conformity judgment result. When the trend consistency judgment result indicates that the change in environmental risk and the braking change gradient are consistent, and the interval conformity judgment result indicates that the steering change gradient, driving change gradient and braking change gradient all fall within the reasonable range of operation, the active avoidance behavior judgment result is generated. The violation result generation unit is used to make a joint judgment based on the trend consistency judgment result and the interval conformity judgment result. When the trend consistency judgment result indicates that the change in environmental risk and the braking change gradient do not maintain a consistent response, or when the interval conformity judgment result indicates that any operation change gradient exceeds the operation rationality range, the non-standard operation behavior judgment result is generated.

[0026] In this embodiment of the invention, by extracting the direction of change of environmental risk change and braking change gradient respectively and aligning them segment by segment to form a trend correspondence, and then verifying the trend continuity between adjacent driving behavior time segments, a trend consistency judgment result is obtained. At the same time, the result of the item-by-item comparison between the operation change gradient and the operation rationality interval is used for joint judgment. The two dimensions of "change trend consistency" and "interval constraint compliance" can be introduced into the judgment process at the same time. This makes the behavior judgment not only dependent on whether the operation result falls within the reasonable range, but also consider whether the operation change forms a response relationship with the environmental change. In this way, it is possible to identify continuous braking response behavior caused by changes in environmental risk and distinguish it from abnormal operation that does not have trend consistency, thereby improving the accuracy of the judgment of the nature of driving behavior.

[0027] In a preferred embodiment of the present invention, the interval conformity determination unit includes: According to the time sequence of the driving behavior time sequence set, the steering change gradient, driving change gradient and braking change gradient corresponding to each driving behavior time sequence segment are read sequentially, and the operation rationality interval corresponding to each driving behavior time sequence segment is read synchronously. The operation rationality interval includes the deceleration range, braking advance range and steering change range. The steering change gradient and steering change range in the current driving behavior time sequence segment are compared and matched. First, the steering wheel angle adjustment state represented by the current steering change gradient is determined. Then, it is determined whether the steering wheel angle adjustment state falls within the steering change range corresponding to the current driving behavior time sequence segment. When the steering wheel angle adjustment state is within the steering change range, the current driving behavior time sequence segment is recorded as a steering compliance segment. When the steering wheel angle adjustment state exceeds the steering change range, the current driving behavior time sequence segment is recorded as a steering over-limit segment. The driving behavior time sequence segment is compared with the driving change gradient and deceleration range. First, the throttle contraction state, throttle holding state, or throttle increase state represented by the current driving change gradient is determined. Then, it is determined whether the driving adjustment state meets the deceleration range corresponding to the current driving behavior time sequence segment. When the deceleration execution degree corresponding to the driving adjustment state falls within the deceleration range, the current driving behavior time sequence segment is recorded as a driving conformity segment. When the deceleration execution degree corresponding to the driving adjustment state does not fall within the deceleration range, the current driving behavior time sequence segment is recorded as a driving deviation segment. The braking change gradient and braking advance range in the current driving behavior time sequence are compared and matched. First, the braking intervention state, braking maintenance state, or braking release state represented by the current braking change gradient is determined. Then, it is determined whether the start time and continuous change process of the braking action are within the braking advance range corresponding to the current driving behavior time sequence. When the start time and continuous change process of the braking action fall within the braking advance range, the current driving behavior time sequence is recorded as a braking conformity segment. When the start time of the braking action is earlier than the start time of the braking advance range or later than the end time of the braking advance range, the current driving behavior time sequence is recorded as a braking deviation segment. The steering compliance segment, drive compliance segment, and braking compliance segment corresponding to the current driving behavior time sequence segment are jointly summarized. When the steering change gradient, drive change gradient, and braking change gradient all meet the corresponding operation rationality interval, the interval compliance judgment result corresponding to the driving behavior time sequence segment is generated. When any operation change gradient does not meet the corresponding operation rationality interval, the interval non-compliance judgment result corresponding to the driving behavior time sequence segment is generated, and the deviation direction and deviation degree corresponding to the non-compliance item are marked and stored.

[0028] In a preferred embodiment of the present invention, the avoidance result generation unit includes: The trend consistency judgment result and interval conformity judgment result corresponding to each driving behavior time sequence segment are read sequentially, and the two types of judgment results are time aligned to make the trend consistency state and interval conformity state within the same driving behavior time sequence segment correspond to each other. The system identifies the trend consistency judgment results corresponding to the current driving behavior time segment and determines whether the change process of the environmental risk change is consistent with the change process of the braking change gradient, which characterizes the degree of braking response. When the environmental risk change shows a continuous increase, if the braking change gradient also shows a continuous increase, the current driving behavior time segment is recorded as a risk response consistent segment. When the environmental risk change shows a stable state, if the braking change gradient also shows a stable state, the current driving behavior time segment is recorded as a risk response consistent segment. When the environmental risk change shows a continuous decrease, if the braking change gradient also shows a continuous decrease, the current driving behavior time segment is recorded as a risk response consistent segment. The system identifies the interval conformity judgment results corresponding to the current driving behavior time sequence segment and determines whether the steering change gradient, driving change gradient and braking change gradient all fall within the operation rationality interval corresponding to the current driving behavior time sequence segment. When all three types of operation change gradients meet the corresponding deceleration range, braking advance range and steering change range, the current driving behavior time sequence segment is recorded as an interval overall conformity segment. The risk response consistency segment is jointly matched with the interval overall conformity segment; when the same driving behavior time segment satisfies both risk response consistency and interval overall conformity, the steering adjustment, drive adjustment and braking response in the driving behavior time segment are determined to be active avoidance operations made in response to changes in the current environmental risk; then, an active avoidance behavior discrimination result corresponding one-to-one with the driving behavior time segment is generated, and the active avoidance behavior discrimination result is bound and stored with the current driving behavior time segment for subsequent graded correction output process to provide prompt suppression processing.

[0029] In a preferred embodiment of the present invention, the violation result generation unit includes: Read the trend consistency judgment result and interval conformity judgment result corresponding to each driving behavior time sequence segment in sequence, and check the two types of judgment results to determine whether there is a risk response inconsistency state or interval non-conformity state in the current driving behavior time sequence segment; When the trend consistency judgment result corresponding to the current driving behavior time segment, which represents the change in environmental risk, is not consistent with the braking change gradient representing the degree of braking response, the specific manifestation of the inconsistency is first identified. When the change in environmental risk continues to increase while the braking change gradient does not increase synchronously, the current driving behavior time segment is recorded as a braking response lag segment. When the change in environmental risk remains constant while the braking change gradient weakens prematurely, the current driving behavior time segment is recorded as a braking response insufficiency segment. When the change in environmental risk weakens while the braking change gradient continues to increase, the current driving behavior time segment is recorded as a braking response over-segment. The corresponding results are stored as a risk response inconsistency state. When the interval corresponding to the current driving behavior time segment meets the judgment result, indicating that any operation change gradient does not fall within the operation reasonableness interval, the specific type of the unmet item is further identified; when the steering change gradient exceeds the steering change range, the current driving behavior time segment is recorded as a steering over-limit segment; when the drive change gradient does not meet the deceleration range, the current driving behavior time segment is recorded as a drive under-adjustment segment or a drive over-adjustment segment; when the braking change gradient does not meet the braking advance range, the current driving behavior time segment is recorded as a braking intervention too early segment or a braking intervention too late segment; and the corresponding result is stored as an interval non-compliance state. The system combines inconsistent risk response states with non-compliant interval states. When an inconsistent risk response state or a non-compliant interval state exists in the current driving behavior time segment, it determines that the operation within that driving behavior time segment does not meet the safety operation requirements corresponding to the current environmental changes, and then generates a non-standard operation behavior discrimination result corresponding to that driving behavior time segment. Subsequently, the non-standard operation behavior discrimination result is bound and stored with the corresponding deviation type, deviation direction, and deviation degree for subsequent graded correction output process to generate corresponding correction instructions.

[0030] In a preferred embodiment of the present invention, the longitudinal constraint synthesis unit includes: The amplitude constraint generation subunit is used to limit the deceleration execution requirements corresponding to each driving behavior time segment based on the deceleration demand results, and generate deceleration amplitude constraint results. The timing constraint generation subunit is used to limit the braking intervention timing corresponding to each driving behavior timing segment according to the braking advance range, and generate braking timing constraint results. The vertical association sub-unit is used to perform corresponding association processing on the deceleration execution requirements and braking intervention time points in each driving behavior time segment based on the deceleration amplitude constraint results and braking timing constraint results, and generate the vertical constraint association results. The longitudinal result output subunit is used to uniformly limit the longitudinal control range in each driving behavior time segment based on the longitudinal constraint association results, and generate longitudinal operation constraint results.

[0031] In this embodiment of the invention, deceleration amplitude constraint results are generated from the deceleration demand results, and braking timing constraint results are generated from the braking advance range. Then, the deceleration execution requirements are correlated with the braking intervention time points to form a unified longitudinal operation constraint result. This allows for coordinated constraint of deceleration amplitude and braking timing, ensuring that braking behavior not only meets amplitude requirements but also timing requirements. This avoids control incoordination problems caused by limiting only deceleration amplitude or only braking timing, enabling the longitudinal control process to simultaneously reflect the matching relationship between risk level and response timing, and ensuring the continuity and consistency of deceleration behavior under different risk change conditions.

[0032] In a preferred embodiment of the present invention, the amplitude constraint generation subunit includes: According to the time sequence of the driving behavior time sequence segment set, read the deceleration demand results corresponding to each driving behavior time sequence segment in turn, and extract the minimum deceleration demand value and deceleration demand level corresponding to each driving behavior time sequence segment. The deceleration execution requirements of the current driving behavior time segment are hierarchically limited according to the deceleration demand level. When the deceleration demand level corresponding to the current driving behavior time segment is high, the deceleration execution requirements of the current driving behavior time segment are limited to rapidly contracting the drive output and simultaneously establishing a braking response. When the deceleration demand level corresponding to the current driving behavior time segment is medium, the deceleration execution requirement of the current driving behavior time segment is limited to first shrinking the drive output and gradually establishing the braking response; When the deceleration requirement level corresponding to the current driving behavior time segment is low, the deceleration execution requirement of the current driving behavior time segment is limited to slowly reducing the drive output or maintaining a light braking response. Based on the driving change gradient and the braking change gradient representing the degree of braking response within the current driving behavior time segment, the state of meeting the deceleration execution requirements is checked accordingly, and the degree of driving contraction and braking establishment that meet the current minimum deceleration requirement value are determined as the deceleration amplitude constraint result of the current driving behavior time segment. The deceleration amplitude constraint results corresponding to each driving behavior time sequence segment are stored in chronological order and bound one by one with the corresponding driving behavior time sequence segment for subsequent association processing of braking time sequence constraint results.

[0033] In a preferred embodiment of the present invention, the time-point constraint generation subunit includes: According to the time sequence of the driving behavior time sequence set, the braking advance range corresponding to each driving behavior time sequence segment is read sequentially, and the allowable braking start time and allowable braking end time corresponding to each driving behavior time sequence segment are extracted. Based on the braking change gradient representing the degree of braking response within the current driving behavior time segment, the start time, duration, and release time of the braking action are identified, and the identified start time of the braking action is compared with the allowable braking start time and allowable braking end time corresponding to the current driving behavior time segment. When the braking action begins between the permitted braking start time and the permitted braking end time, the current driving behavior time sequence segment is recorded as a time point satisfaction segment; When the braking action begins earlier than the permitted braking start time, the current driving behavior sequence segment is recorded as an early intervention segment. When the braking action begins later than the end of the permitted braking period, the current driving behavior sequence segment is recorded as a late intervention segment; Based on the comparison results, the braking intervention time point of the current driving behavior time segment is limited, and the braking intervention start position and duration position that meet the allowable time period requirements are determined as the braking timing constraint result of the current driving behavior time segment. The braking timing constraint results corresponding to each driving behavior timing segment are stored in chronological order and bound one by one with the corresponding driving behavior timing segment for use in the subsequent vertical association establishment process.

[0034] In a preferred embodiment of the present invention, the vertical association establishment sub-unit includes: According to the time order of the driving behavior time sequence segment set, the deceleration amplitude constraint result and braking time sequence constraint result corresponding to each driving behavior time sequence segment are read sequentially, and the two are time aligned within the same driving behavior time sequence segment. After completing the time alignment, it is determined whether the establishment phase of the deceleration execution requirement matches the allowable phase of the braking intervention time. When the drive output contraction process and the braking response establishment process occur within the allowable braking start period and the allowable braking end period, the current driving behavior time sequence segment is recorded as a longitudinally associated satisfaction segment. When the deceleration execution requirement has been established but the braking intervention time is later than the end of the permitted braking period, the current driving behavior sequence segment is recorded as the deceleration precedence mismatch segment; When the braking intervention time falls within the allowable range but the deceleration execution requirement does not reach the corresponding minimum deceleration requirement value, the current driving behavior time sequence segment is recorded as a braking advance mismatch segment. Based on the identification results of longitudinal correlation satisfying segments, deceleration prior mismatch segments, and braking prior mismatch segments, a correspondence is established between the deceleration execution requirements and the braking intervention time points within the current driving behavior time sequence segment, and the correlation states that satisfy the same time sequence requirements and the same amplitude requirements are determined as the longitudinal constraint correlation results; The longitudinal constraint association results corresponding to each driving behavior time sequence segment are stored in chronological order and bound one by one with the corresponding driving behavior time sequence segment for use in the unified limitation of the subsequent longitudinal control range.

[0035] In a preferred embodiment of the present invention, the vertical result output subunit includes: According to the time sequence of the driving behavior time sequence fragment set, the longitudinal constraint association results corresponding to each driving behavior time sequence fragment are read sequentially, and the association status information that represents the completion of the corresponding association between the deceleration execution requirement and the braking intervention time point is extracted. Based on the associated status information, the longitudinal control process within the current driving behavior time segment is uniformly limited. When the current driving behavior time segment is identified as a longitudinally associated fulfillment segment, the drive contraction range, braking establishment range, and continuous response range corresponding to the deceleration execution requirement are integrated into the allowable longitudinal control range. When the current driving behavior sequence segment is identified as a deceleration-preceding mismatch segment or a braking-preceding mismatch segment, the drive adjustment process or braking adjustment process that exceeds the allowable correlation range will be marked as an out-of-longitudinal-limit state. The states outside the allowed longitudinal control range and the longitudinal restriction range are uniformly merged to form the longitudinal control boundary corresponding to the current driving behavior time segment, and this longitudinal control boundary is determined as the longitudinal operation constraint result of the current driving behavior time segment; The longitudinal operation constraint results corresponding to each driving behavior time sequence segment are output in chronological order and bound to the corresponding driving behavior time sequence segment for subsequent operation rationality interval generation process.

[0036] In a preferred embodiment of the present invention, the interval output unit includes: The longitudinal interval generation sub-unit is used to perform intervalization processing on the deceleration amplitude constraints and braking timing constraints in each driving behavior time sequence segment based on the longitudinal operation constraint results, and generate the longitudinal operation interval. A lateral boundary sub-unit is introduced to perform boundary processing on the lateral passage restriction range in each driving behavior time segment based on the upper limit of steering change, and generate lateral boundary results; The interval fusion constraint subunit is used to jointly constrain the longitudinal control range and the lateral control range in each driving behavior time segment based on the longitudinal operation interval and the lateral boundary results, and generate a multi-dimensional constraint interval. The reasonable interval output sub-unit is used to perform unified boundary convergence processing on the deceleration amplitude range, braking advance range, and steering change range in each driving behavior time sequence segment based on the multi-dimensional constraint interval, and generate the operation reasonableness interval.

[0037] In this embodiment of the invention, by transforming the longitudinal operation constraint result into a longitudinal operation interval and introducing a steering change upper limit to form a lateral boundary result, and then jointly limiting the longitudinal control range and the lateral control range to form a multi-dimensional constraint interval, and finally performing unified boundary convergence on the deceleration range, braking advance range and steering change range, the longitudinal control and lateral control can be integrated into the same constraint space, so that the operation rationality interval has both speed control and path control constraints, thereby avoiding the conflict problems caused by setting the longitudinal and lateral rules separately, and making the generated operation rationality interval have uniformity and integrity within the same time window.

[0038] In a preferred embodiment of the present invention, the longitudinal interval generation sub-unit includes: According to the time sequence of the driving behavior time sequence fragment set, the longitudinal operation constraint results corresponding to each driving behavior time sequence fragment are read sequentially, and the allowable deceleration range corresponding to the deceleration execution requirement, the allowable start time corresponding to the braking intervention time point, and the allowable maintenance time corresponding to the braking continuous process are extracted from the longitudinal operation constraint results. The permissible deceleration range within the current driving behavior time segment is divided into several continuous deceleration levels in order from weak to strong. The braking intervention start time and braking maintenance time are divided into continuous time segment boundaries according to the time sequence, so that the deceleration execution requirements and braking intervention requirements form a corresponding relationship within the same driving behavior time segment. After the corresponding relationship is established, the deceleration levels and time period boundaries are uniformly merged to determine the starting interval, continuation interval and ending interval of the deceleration action allowed within the current driving behavior time segment. The starting interval, continuation interval and ending interval are jointly determined as the longitudinal operation interval corresponding to the current driving behavior time segment. The vertical operation intervals corresponding to each driving behavior time sequence segment are stored in chronological order and bound one by one to the corresponding driving behavior time sequence segment for subsequent horizontal boundary introduction process calls.

[0039] In a preferred embodiment of the present invention, the lateral boundary introducing sub-units includes: According to the time sequence of the driving behavior time sequence segment set, read the steering change limit corresponding to each driving behavior time sequence segment in turn, and extract the maximum steering adjustment range allowed in the current driving behavior time sequence segment; Based on the road traffic direction, lane boundary position, and lateral passable range corresponding to the current driving behavior time segment, the maximum allowable steering adjustment range is divided by direction to determine the allowable boundaries when the vehicle adjusts to the left and when it adjusts to the right. When the lateral passable range corresponding to the current driving behavior time segment is narrow, the allowable boundary will be limited to a small range close to the current driving trajectory; When the lateral passable range corresponding to the current driving behavior time segment remains stable, the allowed boundary will be limited to the preset passable range; When the lateral passable range corresponding to the current driving behavior time segment expands, the boundary will be allowed to extend to the outside of the passable area; The left and right permissible boundaries after division are uniformly organized to form the lateral traffic restriction boundary corresponding to the current driving behavior time sequence segment, and this lateral traffic restriction boundary is determined as the lateral boundary result; The lateral boundary results corresponding to each driving behavior time sequence segment are stored in chronological order and bound one by one to the corresponding driving behavior time sequence segment for subsequent interval fusion constraint process.

[0040] In a preferred embodiment of the present invention, the interval fusion limiting subunit includes: According to the time order of the driving behavior time sequence segment set, the vertical operation interval and horizontal boundary result corresponding to each driving behavior time sequence segment are read sequentially, and the two are time aligned within the same driving behavior time sequence segment. After completing the time alignment, the deceleration start interval, deceleration continuation interval and deceleration end interval in the longitudinal operation interval are combined with the left allowable boundary and right allowable boundary in the lateral boundary result to form a synchronous constraint relationship between the longitudinal control requirements and the lateral control requirements in the same driving behavior time segment. When the deceleration start interval and the lateral permissible boundary within the current driving behavior time segment simultaneously meet the passage requirements, the combined state is recorded as the joint satisfaction state. When the longitudinal deceleration range meets the requirements but the lateral allowable boundary is exceeded, the combined state is recorded as the lateral boundary violation state. When the lateral allowable boundary meets the requirements but the longitudinal deceleration range does not, the combined state is recorded as the longitudinal mismatch state. The joint satisfaction state, lateral boundary crossing state and longitudinal mismatch state are merged and organized to determine the control boundary of the current driving behavior time sequence segment that is constrained in both the time and space dimensions, and this control boundary is defined as a multi-dimensional constraint interval. The multidimensional constraint intervals corresponding to each driving behavior time sequence segment are stored in chronological order and bound one by one to the corresponding driving behavior time sequence segment for subsequent reasonable interval output process.

[0041] In a preferred embodiment of the present invention, the reasonable interval output subunit includes: According to the time sequence of the driving behavior time sequence fragment set, the multidimensional constraint interval corresponding to each driving behavior time sequence fragment is read sequentially, and the constraint content representing the longitudinal control boundary and the lateral control boundary is extracted. Based on the longitudinal control boundary, determine the allowable deceleration range and braking advance range within the current driving behavior time segment; Based on the lateral control boundary, determine the permissible range of steering changes within the current driving behavior time segment; After determining the deceleration range, braking advance range, and steering change range, the boundary positions of the three are uniformly converged to ensure that the three have consistent start and end times and consistent control boundaries within the same driving behavior time segment. When the boundaries of the three elements intersect or conflict, the common overlapping area that simultaneously satisfies the longitudinal control requirements and the lateral control requirements shall be taken as the final allowable range. When the boundaries of the three are consistent, the consistent boundary is directly used as the final allowable interval. The deceleration range, braking advance range and steering change range after the unified boundary convergence are combined and output to generate an operation rationality interval corresponding to the current driving behavior time sequence segment. The operation rationality intervals corresponding to each driving behavior time sequence segment are stored in chronological order for subsequent interval compliance judgment process.

[0042] In a preferred embodiment of the present invention, the trend-corresponding extraction unit includes: The risk direction generation subunit is used to extract the direction of risk change status within each driving behavior time segment based on the environmental risk change amount in the driving behavior time segment set, and generate a risk change direction sequence. The braking direction generation subunit is used to extract the braking response state within each driving behavior time segment based on the braking change gradient in the driving behavior time segment set, and generate a braking response direction sequence. The direction alignment processing subunit is used to perform segment-by-segment alignment processing of the risk change direction and the braking response direction within each driving behavior time segment based on the risk change direction sequence and the braking response direction sequence, and generate direction alignment results. The correspondence output sub-unit is used to determine whether the risk change direction and braking response direction in each driving behavior time segment form a preset correspondence based on the direction alignment result, and to generate a trend correspondence.

[0043] In this embodiment of the invention, by extracting the risk change direction sequence corresponding to the environmental risk change amount and the braking response direction sequence corresponding to the braking change gradient, and aligning the two segments one by one, and then determining whether a preset correspondence is formed based on the alignment result, the relationship between environmental change and driving operation can be transformed into a directional correspondence. This transforms driving behavior analysis from numerical comparison to trend matching, thereby enabling the identification of whether driving operation responds and adjusts with changes in environmental risk. This avoids the misjudgment problem caused by judging only by numerical range, and makes the behavior discrimination process closer to the dynamic response characteristics in actual driving.

[0044] In a preferred embodiment of the present invention, the risk direction generation subunit includes: According to the time sequence of the driving behavior time sequence, the environmental risk change corresponding to each driving behavior time sequence is read sequentially, and the environmental risk change of the current driving behavior time sequence is compared with the environmental risk change of the previous driving behavior time sequence to determine whether the environmental risk in the current driving behavior time sequence is in an enhanced state, a maintained state, or a weakened state. When the change in environmental risk of the current driving behavior time segment is higher than that of the previous driving behavior time segment, the current driving behavior time segment is recorded as a risk-enhanced segment. When the change in environmental risk of the current driving behavior time segment is consistent with that of the previous driving behavior time segment, the current driving behavior time segment is recorded as a risk maintenance segment. When the change in environmental risk of the current driving behavior time segment is lower than that of the previous driving behavior time segment, the current driving behavior time segment is recorded as a risk-reduced segment. Arrange the risk enhancement segment, risk maintenance segment, and risk reduction segment corresponding to each driving behavior time sequence segment in chronological order, and write the corresponding risk direction identifier for each driving behavior time sequence segment to generate a risk change direction sequence. The risk change direction sequence is bound and stored one by one with the corresponding driving behavior time sequence segment for subsequent braking response direction extraction and direction alignment processing.

[0045] In a preferred embodiment of the present invention, the braking direction generation subunit includes: According to the time sequence of the driving behavior time sequence set, the braking change gradient corresponding to each driving behavior time sequence segment, which represents the degree of braking response, is read sequentially. The braking change gradient of the current driving behavior time sequence segment is compared with the braking change gradient of the previous driving behavior time sequence segment to determine whether the braking response in the current driving behavior time sequence segment is in an enhanced state, a maintained state, or a weakened state. When the braking change gradient of the current driving behavior time segment is higher than that of the previous driving behavior time segment, the current driving behavior time segment is recorded as a braking enhancement segment. When the braking change gradient of the current driving behavior time segment is consistent with that of the previous driving behavior time segment, the current driving behavior time segment is recorded as a braking maintenance segment. When the braking change gradient of the current driving behavior time segment is lower than that of the previous driving behavior time segment, the current driving behavior time segment is recorded as a braking reduction segment. Arrange the braking enhancement segment, braking maintenance segment, and braking reduction segment corresponding to each driving behavior time sequence segment in chronological order, and write the corresponding braking direction identifier for each driving behavior time sequence segment to generate a braking response direction sequence. The braking response direction sequence is bound and stored one by one with the corresponding driving behavior time sequence segment for subsequent direction alignment processing.

[0046] In a preferred embodiment of the present invention, the orientation alignment processing subunit includes: According to the time sequence of the driving behavior time sequence fragment set, the risk change direction sequence item and braking response direction sequence item corresponding to each driving behavior time sequence fragment are read sequentially, and the same driving behavior time sequence fragment is used as the alignment benchmark to establish the corresponding relationship. The risk direction indicator and braking direction indicator in the current driving behavior time segment are compared synchronously. When the risk direction indicator is risk enhancement and the braking direction indicator is braking enhancement, the current driving behavior time segment is recorded as a synchronous enhancement aligned segment. When the risk direction is identified as risk maintenance and the braking direction is identified as braking maintenance, the current driving behavior time sequence segment is recorded as a synchronization maintenance alignment segment. When the risk direction is marked as risk reduction and the braking direction is marked as braking reduction, the current driving behavior time sequence segment is recorded as a synchronized reduction alignment segment. When the risk direction indicator and the braking direction indicator are not in the same state of change, the current driving behavior time sequence segment is recorded as a direction mismatch segment; The synchronous enhancement alignment segment, synchronous maintenance alignment segment, synchronous weakening alignment segment, and orientation mismatch segment are merged and organized in chronological order to generate the orientation alignment result. The direction alignment results are bound and stored one by one with the corresponding driving behavior time sequence segments for subsequent correspondence output and retrieval.

[0047] In a preferred embodiment of the present invention, the correspondence output subunit includes: According to the time order of the driving behavior time sequence segment set, read the direction alignment results corresponding to each driving behavior time sequence segment in sequence, and extract the alignment state type in the current driving behavior time sequence segment; When the current driving behavior time segment is identified as a synchronous enhancement alignment segment, synchronous maintenance alignment segment, or synchronous weakening alignment segment, the current driving behavior time segment is recorded as a segment that satisfies the preset correspondence relationship. When the current driving behavior time segment is identified as a direction mismatch segment, the current driving behavior time segment is recorded as a segment that does not meet the preset correspondence relationship; Segments that satisfy the preset correspondence and those that do not are uniformly identified and written into the trend correspondence status results respectively; Output the trend-corresponding state results in chronological order to generate trend correspondences; The trend correspondence is bound and stored one by one with the corresponding time sequence of driving behavior, so that it can be called in the subsequent continuous consistency verification process.

[0048] In a preferred embodiment of the present invention, the continuous consistency verification unit includes: The continuous segment construction subunit is used to continuously connect temporal segments of driving behavior that are adjacent in time according to the trend correspondence, and generate a continuous segment sequence. The consistent state extraction subunit is used to extract the trend correspondence in each continuous segment sequence according to the continuous segment sequence to generate a consistent state sequence. The interrupted segment filtering subunit is used to remove driving behavior time sequence segments whose trend correspondence does not meet the consistent continuation condition based on the consistent state sequence, and generate a continuous segment sequence after filtering. The consistency result output subunit is used to perform unified verification processing on the trend correspondence in the remaining driving behavior time sequence segments based on the filtered continuous segment sequence, and generate trend consistency judgment results.

[0049] In this embodiment of the invention, by continuously connecting sequentially adjacent driving behavior segments to form a continuous segment sequence, and extracting the trend correspondence in the continuous segment sequence segment by segment, segments that do not meet the consistent continuity condition are screened out. Finally, the screened continuous segment sequence is uniformly verified to generate a trend consistency judgment result. This introduces a continuity constraint in the behavior judgment process, ensuring that the trend correspondence is not only valid within a single time window, but also consistent across multiple consecutive time windows. This avoids misjudgment problems caused by short-term fluctuations or occasional operations, enabling the judgment result to reflect the stability of driving behavior over a continuous period of time and improving the ability to identify continuous features of driving behavior.

[0050] In a preferred embodiment of the present invention, the interruption fragment filtering subunit includes: According to the order of the consistent state sequence, read the consistent state results corresponding to each consecutive segment in sequence, and extract the consecutive segments recorded as mixed state segments and consistent mismatch segments. The trend-corresponding state of each driving behavior time sequence segment within the mixed state segment is re-examined to identify the specific driving behavior time sequence segment that causes the state interruption, and the specific driving behavior time sequence segment is recorded as the interruption segment. The system identifies all mismatched segments and records all driving behavior time sequences as mismatched segments. Interrupted segments and mismatched segments are removed in chronological order so that they no longer participate in subsequent continuous consistency determination. After the removal is completed, the remaining driving behavior time sequence segments are reconnected in chronological order so that the remaining driving behavior time sequence segments that are adjacent in time and have corresponding trends and consistent states form new continuous segments. The start time, end time, and range of driving behavior time segments contained in the reconnected continuous segments are re-recorded to generate a filtered continuous segment sequence. The filtered continuous segment sequences are bound one-to-one with the trend-corresponding state results of the remaining driving behavior time series segments for subsequent consistent result output processing.

[0051] In a preferred embodiment of the present invention, the consensus result output subunit includes: According to the order of the filtered continuous segment sequence, read the driving behavior time sequence segment and its corresponding trend state result contained in each filtered continuous segment in sequence; Each filtered continuous segment is uniformly verified to confirm whether the driving behavior time sequence segments within the filtered continuous segment continuously maintain the preset correspondence state. When all driving behavior time segments within the filtered continuous segments maintain a preset correspondence state, the filtered continuous segments are recorded as trend-consistent segments. If a driving behavior time sequence segment that does not meet the preset correspondence state exists in the filtered continuous segment, the filtered continuous segment is recorded as a trend inconsistent segment. For each driving behavior time sequence segment contained in a continuous segment recorded as a trend-consistent segment, a trend-consistent identifier is uniformly written; for each driving behavior time sequence segment contained in a continuous segment recorded as a trend-inconsistent segment, a trend-inconsistent identifier is uniformly written. The trend consistency and trend inconsistency indicators corresponding to each driving behavior time sequence segment are output in chronological order to generate the trend consistency judgment result. The trend consistency determination results are bound and stored one by one with the corresponding driving behavior time sequence segments for subsequent avoidance result generation and violation result generation processes.

[0052] In a preferred embodiment of the present invention, the risk quantity synthesis unit includes: The component association establishment sub-unit is used to perform corresponding association processing on pedestrian risk components, vehicle risk components and spatial constraint risk components according to the time window sequence, and generate risk component combination results; The weight allocation generation sub-unit is used to allocate the influence of pedestrian risk components, vehicle risk components and spatial constraint risk components in each time window according to the risk component combination results, and generate component weight results. The weighted fusion generation sub-unit is used to perform weighted superposition processing on each risk component based on the component weight results to generate a basic risk value sequence; The change generation subunit is used to extract the differences between basic risk values ​​in adjacent time windows based on the basic risk value sequence, and generate the environmental risk change.

[0053] In this embodiment of the invention, pedestrian risk components, vehicle risk components, and spatial constraint risk components are correlated and weighted according to the degree of influence of each risk component within a time window. Then, the risk components are weighted and superimposed to generate a basic risk value sequence. Furthermore, the differences between the basic risk values ​​of adjacent time windows are extracted to form an environmental risk change quantity. This can unify multi-source environmental information into a single risk change indicator, enabling different types of environmental factors to be comprehensively expressed under the same evaluation system. This avoids the problem of biased judgment caused by a single risk source, and allows the environmental risk change quantity to reflect the changing trend of the comprehensive traffic situation, providing a unified input basis for subsequent interval generation and behavior discrimination.

[0054] In a preferred embodiment of the present invention, the weight allocation generation subunit includes: According to the order of the time window sequence, read the risk component combination results corresponding to each time window in sequence, and extract the pedestrian risk component status, vehicle risk component status and spatial constraint risk component status in the current time window. Based on the risk status represented by the three risk components within the current time window, the degree of influence of the three risk components is ranked in order. When the pedestrian risk component represents the target pedestrian entering the predetermined passage area of ​​the training vehicle and the distance of the target pedestrian continues to shorten, the pedestrian risk component is recorded as the priority influence component. When the vehicle risk component represents the increasing relative proximity of surrounding vehicles, the vehicle risk component is recorded as a highly correlated influence component. When the spatial constraint risk component represents the continuous shrinking of the lateral passable range, the spatial constraint risk component is recorded as the boundary restriction impact component. The priority impact components, highly correlated impact components, and boundary constraint impact components are sequentially assigned to ensure that risk components whose risk status directly affects the current passage safety receive higher assignment results, while risk components whose risk status has a weaker impact on the current passage safety receive lower assignment results. When all three risk components are in a state of heightened risk within the current time window, the risk component with a higher degree of conflict with the current driving path of the training vehicle will be identified as the dominant component and given a higher priority in the allocation process. Write the corresponding component weight tag to the three types of risk components after the allocation process is completed, and generate the component weight result; bind the component weight result with the corresponding risk component combination result and store them one by one for subsequent weighted fusion generation process.

[0055] In a preferred embodiment of the present invention, the weighted fusion generation subunit includes: According to the order of the time window sequence, read the component weight results corresponding to each time window in sequence, and extract the component weight labels corresponding to the pedestrian risk component, vehicle risk component and spatial constraint risk component in the current time window respectively. Based on the component weight label corresponding to each risk component, the three types of risk components are arranged in a fusion order so that the risk components with higher component weight labels participate in the basic risk value generation process first, and the risk components with lower component weight labels participate in the basic risk value generation process later. First, register the risk component with the highest weight as a basic risk. Then, add the risk component with the second highest weight to the registered basic risk. Finally, add the remaining risk components to the above-mentioned overlay result so that the three types of risk components in the current time window can form a unified basic risk expression result. When a certain risk component is in a state of enhanced risk within the current time window and its corresponding component weight label is higher than that of other risk components, the dominant role of this risk component in the basic risk expression result is maintained during the superposition process. When all three risk components are in a state of maintenance or weakening within the current time window, the three risk components are sequentially merged according to their respective component weights, so that the basic risk expression result reflects the overall state of environmental risk within the current time window. After the fusion processing of each time window is completed, the basic risk expression results corresponding to each time window are recorded in chronological order to generate a basic risk value sequence. The basic risk value sequence is bound and stored one by one with the corresponding time window sequence for subsequent change generation process.

[0056] In a preferred embodiment of the present invention, the change generation subunit includes: According to the time sequence of the basic risk value sequence, the basic risk expression results of the current time window and the basic risk expression results of the previous time window are read sequentially, and the two are compared before and after. When the basic risk expression result of the current time window is higher than the basic risk expression result of the previous time window, the current time window is recorded as the risk increase window, and the degree of difference corresponding to this increase is recorded as the risk enhancement change amount. When the basic risk expression result of the current time window is consistent with the basic risk expression result of the previous time window, the current time window is recorded as the risk maintenance window, and the degree of difference corresponding to the maintenance state is recorded as the risk maintenance change. When the basic risk expression result of the current time window is lower than the basic risk expression result of the previous time window, the current time window is recorded as the risk decline window, and the degree of difference corresponding to this decline is recorded as the risk reduction change. The changes in risk enhancement, risk maintenance, and risk reduction corresponding to each time window are organized in a unified format so that each time window corresponds to a unique record of environmental risk changes. The compiled environmental risk change records for each time window are output in chronological order to generate the environmental risk change volume. The changes in environmental risks are bound to the corresponding time window sequences and stored one by one, so that they can be called in the subsequent process of constructing the time sequence set of driving behavior and determining the reasonable interval.

[0057] In a preferred embodiment of the present invention, the graded correction output module includes: The deviation extraction unit is used to extract the deviation between the operation change gradient and the operation rationality interval in each driving behavior time segment based on the interval conformity judgment result, and generate the deviation result. The grade mapping generation unit is used to perform grade mapping processing on the degree of deviation corresponding to the deviation amount based on the deviation amount result, and generate the correction grade result; The instruction construction unit is used to match the direction and magnitude of operation adjustment in each driving behavior time sequence segment according to the correction level result, and generate a set of correction instructions; The output control unit is used to perform output control processing on the set of correction instructions based on the judgment results of active avoidance behavior and non-standard operation behavior. When a judgment result of non-standard operation behavior is generated, the corresponding correction instruction is output, and when a judgment result of active avoidance behavior is generated, the correction instruction is suppressed.

[0058] In this embodiment of the invention, deviation results are formed by extracting the deviation between the gradient of operation change and the reasonableness range of operation. Correction levels are generated by hierarchical mapping based on the deviation results. Then, the operation adjustment direction and adjustment range are matched according to the correction level to form a set of correction instructions. Finally, the set of correction instructions is output and controlled according to the behavior discrimination result. This can transform the driving behavior deviation from the identification result into specific adjustment guidance. The correction process can not only point out the deviation, but also provide corresponding adjustment methods. At the same time, by suppressing the output of correction instructions under the active avoidance behavior discrimination result, interference with reasonable operation can be avoided. This realizes the linkage between behavior judgment and correction output, and makes the correction process consistent with the driving situation.

[0059] In a preferred embodiment of the present invention, the level mapping generation unit includes: According to the time sequence of the driving behavior time sequence segment set, the deviation results corresponding to each driving behavior time sequence segment are read sequentially, and the steering deviation, driving deviation and braking deviation are extracted. Based on the magnitude and direction of the deviation, each deviation is classified into a level. When the deviation is within the preset minimum deviation range, it is recorded as a low-level deviation. When the deviation exceeds the minimum deviation range but does not reach the critical deviation range, it is recorded as a medium-level deviation; when the deviation exceeds the critical deviation range, it is recorded as a high-level deviation. The steering deviation, driving deviation and braking deviation within the same driving behavior time segment are comprehensively evaluated and processed. When any high-level deviation exists, the current driving behavior time segment is recorded as a high-level correction segment. When there is no high-level deviation but there is a medium-level deviation, the current driving behavior time sequence segment is recorded as a medium-level correction segment; When only low-level deviations exist, the current driving behavior time sequence segment is recorded as a low-level correction segment; The grading results corresponding to each driving behavior time sequence segment are uniformly identified, a correction grade result is generated, and each result is bound and stored with the corresponding driving behavior time sequence segment.

[0060] In a preferred embodiment of the present invention, the instruction construction unit includes: According to the time sequence of the driving behavior time sequence set, the correction level results corresponding to each driving behavior time sequence segment are read sequentially, and the specific deviation type is extracted by combining the deviation amount results corresponding to the driving behavior time sequence segment. The direction of operation adjustment is determined based on the type of deviation. When the steering deviation is detected to be an excessive deviation, the adjustment direction is set to reduce the steering amplitude. When the detected drive deviation is insufficient deceleration, the adjustment direction will be set to further reduce the drive output; When the detected braking deviation is due to intervention lag, the adjustment direction will be set to establish braking response in advance; The adjustment range is determined based on the correction level results. When the current driving behavior time segment corresponds to a high-level correction segment, the adjustment range is set to the high-level adjustment range. When the corresponding segment is a medium-level correction segment, the adjustment range is set to the medium-level adjustment range; When the corresponding segment is a low-level correction segment, the adjustment range is set to the low-level adjustment range; The direction and magnitude of the operation adjustment are combined to form a correction instruction for the current driving behavior time segment, and the correction instructions corresponding to multiple driving behavior time segments are summarized to generate a set of correction instructions. The set of correction instructions is bound and stored one by one with the corresponding driving behavior time sequence segments for subsequent output control processing.

[0061] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A driving behavior analysis and correction system for a driver training robot based on artificial intelligence, characterized in that, The system includes: The operating environment acquisition module is used to simultaneously collect student operation data and road environment data. The student operation data includes the change in steering wheel angle, the rate of change in accelerator pedal opening, and the duration of brake pedal action. The road environment data includes the distance to the target pedestrian, the speed of the target pedestrian, the relative speed of surrounding vehicles, and the boundary information of the road passage area, generating a corresponding dataset for the operating environment. The time-series segment generation module is used to synchronously segment student operation changes and environmental changes according to a preset time window based on the dataset corresponding to the operation environment, extract the operation change gradient and environmental risk change amount within each time window, and generate a set of driving behavior time-series segments. The operation change gradient includes steering change gradient, driving change gradient and braking change gradient. The reasonable interval determination module is used to determine the safe operating range corresponding to each driving behavior time sequence segment based on the environmental risk change in the driving behavior time sequence segment set, and generate a set of reasonable operating intervals including deceleration range, braking advance range and steering change range; The behavior attribute discrimination module is used to compare the operation change gradient in the time sequence of driving behavior segments with the operation rationality interval set segment by segment. When the operation change trend matches the environmental risk change trend and the operation change gradient falls into the corresponding operation rationality interval, an active avoidance behavior discrimination result is generated. When the operation change gradient exceeds the corresponding operation rationality interval, a non-standard operation behavior discrimination result is generated. The graded correction output module is used to perform corrective control based on the judgment results of active avoidance behavior and non-standard operation behavior. When generating non-standard operation behavior judgment results, it outputs graded correction instructions according to the deviation amount exceeding the corresponding operation reasonableness range. When generating active avoidance behavior judgment results, it suppresses the output of misjudgment prompts. The reasonable interval determination module includes: The deceleration demand generation unit is used to perform hierarchical mapping processing based on the environmental risk changes in the driving behavior time sequence segment set, determine the minimum deceleration demand value for each driving behavior time sequence segment, and generate deceleration demand results. The braking advance range generation unit is used to perform pre-constraint processing on the braking intervention time point based on the changing trend of environmental risk changes, and generate the braking advance range. The turning boundary generation unit is used to limit the lateral passable range based on the spatial constraint information corresponding to the environmental risk change, and generate the turning change upper limit. The longitudinal constraint synthesis unit is used to perform linkage constraint processing based on the deceleration demand result and the braking advance range to generate the longitudinal operation constraint result. The interval output unit is used to perform fusion and limitation processing based on the longitudinal operation constraint results and the upper limit of steering change to generate the operation rationality interval corresponding to each driving behavior time sequence segment. The operation rationality interval includes the deceleration range, the braking advance range, and the steering change range.

2. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 1, characterized in that, The time segment generation module includes: The window partitioning unit is used to continuously divide the student operation data and road environment data according to the dataset corresponding to the operation environment, and generate a time window sequence by sliding time windows of fixed length. The operation gradient generation unit is used to perform continuous difference processing on the steering wheel angle change in each time window sequence to generate steering change gradient, perform continuous difference processing on the accelerator pedal opening change rate in each time window to generate drive change gradient, and perform cumulative change processing on the brake pedal application time in each time window to generate brake change gradient. The risk component generation unit is used to correlate the change in the distance of the target pedestrian with the moving speed of the target pedestrian in each time window sequence to generate a pedestrian risk component, extract the change in the relative speed of the surrounding vehicles in each time window to generate a vehicle risk component, and identify the boundary information of the road passage area in each time window to generate a spatial constraint risk component. The risk quantity synthesis unit is used to perform weighted combination processing based on pedestrian risk components, vehicle risk components and spatial constraint risk components to generate environmental risk change quantities corresponding to each time window. The segment output unit is used to bind the steering change gradient, driving change gradient, braking change gradient and environmental risk change one by one, and arrange them in order according to the time window sequence to generate a set of driving behavior time sequence segments.

3. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 1, characterized in that, The behavior attribute discrimination module includes: The trend correspondence extraction unit is used to extract the correspondence between the risk change direction and the braking change direction in each driving behavior time sequence segment based on the environmental risk change amount and braking change gradient in the driving behavior time sequence segment set, and generate trend correspondence relationship. The continuous consistency verification unit is used to verify the trend continuity status between adjacent driving behavior time segments according to the trend correspondence relationship and generate a trend consistency judgment result. The interval conformity determination unit is used to compare the steering change gradient, driving change gradient and braking change gradient in the driving behavior time sequence segment set with the corresponding operation rationality intervals one by one to generate the interval conformity determination result. The avoidance result generation unit is used to make a joint judgment based on the trend consistency judgment result and the interval conformity judgment result. When the trend consistency judgment result indicates that the change in environmental risk and the braking change gradient are consistent, and the interval conformity judgment result indicates that the steering change gradient, driving change gradient and braking change gradient all fall within the reasonable range of operation, the active avoidance behavior judgment result is generated. The violation result generation unit is used to make a joint judgment based on the trend consistency judgment result and the interval conformity judgment result. When the trend consistency judgment result indicates that the change in environmental risk and the braking change gradient do not maintain a consistent response, or when the interval conformity judgment result indicates that any operation change gradient exceeds the operation rationality range, the non-standard operation behavior judgment result is generated.

4. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 1, characterized in that, The longitudinal constraint synthesis unit includes: The amplitude constraint generation subunit is used to limit the deceleration execution requirements corresponding to each driving behavior time segment based on the deceleration demand results, and generate deceleration amplitude constraint results. The timing constraint generation subunit is used to limit the braking intervention timing corresponding to each driving behavior timing segment according to the braking advance range, and generate braking timing constraint results. The vertical association sub-unit is used to perform corresponding association processing on the deceleration execution requirements and braking intervention time points in each driving behavior time segment based on the deceleration amplitude constraint results and braking timing constraint results, and generate the vertical constraint association results. The longitudinal result output subunit is used to uniformly limit the longitudinal control range in each driving behavior time segment based on the longitudinal constraint association results, and generate longitudinal operation constraint results.

5. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 1, characterized in that, The interval output unit includes: The longitudinal interval generation sub-unit is used to perform intervalization processing on the deceleration amplitude constraints and braking timing constraints in each driving behavior time sequence segment based on the longitudinal operation constraint results, and generate the longitudinal operation interval. A lateral boundary sub-unit is introduced to perform boundary processing on the lateral passage restriction range in each driving behavior time segment based on the upper limit of steering change, and generate lateral boundary results; The interval fusion constraint subunit is used to jointly constrain the longitudinal control range and the lateral control range in each driving behavior time segment based on the longitudinal operation interval and the lateral boundary results, and generate a multi-dimensional constraint interval. The reasonable interval output sub-unit is used to perform unified boundary convergence processing on the deceleration amplitude range, braking advance range, and steering change range in each driving behavior time sequence segment based on the multi-dimensional constraint interval, and generate the operation reasonableness interval.

6. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 3, characterized in that, The trend extraction unit includes: The risk direction generation subunit is used to extract the direction of risk change status within each driving behavior time segment based on the environmental risk change amount in the driving behavior time segment set, and generate a risk change direction sequence. The braking direction generation subunit is used to extract the braking response state within each driving behavior time segment based on the braking change gradient in the driving behavior time segment set, and generate a braking response direction sequence. The direction alignment processing subunit is used to perform segment-by-segment alignment processing of the risk change direction and the braking response direction within each driving behavior time segment based on the risk change direction sequence and the braking response direction sequence, and generate direction alignment results. The correspondence output sub-unit is used to determine whether the risk change direction and braking response direction in each driving behavior time segment form a preset correspondence based on the direction alignment result, and to generate a trend correspondence.

7. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 3, characterized in that, The continuous consistency verification unit includes: The continuous segment construction subunit is used to continuously connect temporal segments of driving behavior that are adjacent in time according to the trend correspondence, and generate a continuous segment sequence. The consistent state extraction subunit is used to extract the trend correspondence in each continuous segment sequence according to the continuous segment sequence to generate a consistent state sequence. The interrupted segment filtering subunit is used to remove driving behavior time sequence segments whose trend correspondence does not meet the consistent continuation condition based on the consistent state sequence, and generate a continuous segment sequence after filtering. The consistency result output subunit is used to perform unified verification processing on the trend correspondence in the remaining driving behavior time sequence segments based on the filtered continuous segment sequence, and generate trend consistency judgment results.

8. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 2, characterized in that, The risk quantity synthesis unit includes: The component association establishment sub-unit is used to perform corresponding association processing on pedestrian risk components, vehicle risk components and spatial constraint risk components according to the time window sequence, and generate risk component combination results; The weight allocation generation sub-unit is used to allocate the influence of pedestrian risk components, vehicle risk components and spatial constraint risk components in each time window according to the risk component combination results, and generate component weight results. The weighted fusion generation sub-unit is used to perform weighted superposition processing on each risk component based on the component weight results to generate a basic risk value sequence; The change generation subunit is used to extract the differences between basic risk values ​​in adjacent time windows based on the basic risk value sequence, and generate the environmental risk change.

9. The driving behavior analysis and correction system for a driver training robot based on artificial intelligence according to claim 1, characterized in that, The graded correction output module includes: The deviation extraction unit is used to extract the deviation between the operation change gradient and the operation rationality interval in each driving behavior time segment based on the interval conformity judgment result, and generate the deviation result. The grade mapping generation unit is used to perform grade mapping processing on the degree of deviation corresponding to the deviation amount based on the deviation amount result, and generate the correction grade result; The instruction construction unit is used to match the direction and magnitude of operation adjustment in each driving behavior time sequence segment according to the correction level result, and generate a set of correction instructions; The output control unit is used to perform output control processing on the set of correction instructions based on the judgment results of active avoidance behavior and non-standard operation behavior. When a judgment result of non-standard operation behavior is generated, the corresponding correction instruction is output, and when a judgment result of active avoidance behavior is generated, the correction instruction is suppressed.

Images