A method and related equipment for correcting highway radius based on cognitive delay under hypoxic conditions

Abstract

This invention discloses a method and related equipment for correcting highway radius based on cognitive delay under hypoxic conditions. This invention relates to the field of high-altitude highway design optimization technology. By constructing a dynamically updated altitude-cognitive mapping model, this invention achieves refined decomposition and dynamic prediction of driver cognitive delay time. Through the calculation of comprehensive cognitive delay time across multiple time scales, combined with the formulation of scenario-differentiated sight distance compensation strategies, it achieves refined compensation for various complex scenarios and collaborative compensation under multiple scenario superpositions. Finally, the compensated stopping sight distance is converted into the minimum radius index of horizontal curves, constructing a full-chain correction system from driver physiological and psychological load to road design parameters. Compared with existing technologies, this significantly improves the safety and economy of high-altitude highway alignment design, achieving substantial technological improvements.

Description

Technical Field

[0001] This invention relates to the field of high-altitude highway design optimization technology, specifically to a highway radius correction method and related equipment based on cognitive delay under low-oxygen conditions. Background Technology

[0002] my country's plateau regions mainly include the Qinghai-Tibet Plateau, the Inner Mongolia Plateau, the Loess Plateau, and the Yunnan-Guizhou Plateau. These regions have complex terrains and harsh climates, making highway construction extremely difficult. However, highway construction is crucial for regional economic development and improving people's livelihoods; therefore, the country is vigorously developing highway construction in plateau regions. The plateau environment is characterized by high altitude, low air pressure, strong radiation, large temperature differences, and fragile ecosystems. These natural conditions not only shape unique geographical landscapes but also pose severe challenges to human activities, biological survival, and engineering construction.

[0003] In the low-oxygen environment of high altitudes, drivers experience significant changes in their physiological and psychological state, with decreased cognitive ability and increased reaction delays, posing a serious threat to driving safety, especially on complex road sections such as curves. The design of reasonable curve radii must fully consider the physiological and psychological factors of drivers. To address this challenge, existing technologies propose three main methods: the stopping sight distance method, the curve-slope combination method, and the horizontal curve method. The stopping sight distance method breaks down the driver's braking reaction time by considering factors such as altitude, age, and lighting conditions, providing recommended sight distance values ​​for different altitude ranges. The curve-slope combination method and the horizontal curve method construct the CISE (Condition of Influence on Sight) physiological and psychological load index, attempting to incorporate the driver's physiological and psychological load into the road alignment design. These methods, to some extent, fill the gaps in driver-related factors in high-altitude highway design, but still have many technical shortcomings.

[0004] To address this design challenge, existing technologies have proposed three targeted methods: the stopping sight distance method, the curve-slope combination method, and the horizontal curve method. The stopping sight distance method incorporates factors such as altitude, age, and lighting conditions to break down and analyze driver braking reaction time, providing suggested sight distance values ​​for different altitude ranges and offering a sight distance reference for curve radius design. The curve-slope combination method and the horizontal curve method construct a Comprehensive Evaluation Index (CISE) for driver physiological and psychological load, incorporating driver physiological and psychological load into the road alignment design considerations. This aims to provide a basis for curve radius correction from a physiological and psychological perspective. These methods, to some extent, fill the gap in the design of plateau highways that neglects driver physiological and psychological factors, providing preliminary technical ideas for curve radius correction in low-oxygen environments.

[0005] However, existing technical solutions still have many prominent technical shortcomings. In terms of quantifying cognitive delay time, existing technologies suffer from a single dimension and crude breakdown. The parking sight distance method simply breaks down the driver's reaction time into braking reaction time and braking operation time, without effectively distinguishing between perception delay and judgment / decision delay. In contrast, the impact of the high-altitude, low-oxygen environment on the driver's judgment / decision delay is far greater than that on the operation stage, directly leading to a lack of specificity in the relevant correction factors. At the same time, there is a significant gap in the coverage of factors affecting cognitive delay, only including basic factors such as altitude, age, and light, while ignoring key influencing factors such as high-altitude acclimatization period, driver fatigue, vehicle type, and driving experience. Furthermore, the interaction between various factors is not considered, making it unable to address the problem of aggravated cognitive delay in scenarios such as high altitude, nighttime, and fatigue. In addition, although the curve-slope combination section and horizontal curve methods construct the CISE biopsychological load index, they fail to establish an effective quantitative correlation between this index and cognitive delay time, making the research on biopsychological load and cognitive delay isolated from each other, and unable to form a complete chain correction system of "biopsychological load - cognitive delay - road design".

[0006] Regarding scenario-based compensation, existing technologies also have significant shortcomings. On the one hand, the scenario division dimensions are limited, failing to fully cover the typical complex scenarios of plateau highways. Road alignment is simply divided into ordinary sections, curved and sloping sections, and flat curve sections, without considering extreme alignment scenarios such as cliff edges, sharp bends and steep slopes, tunnel entrances, and long longitudinal slopes. In terms of environmental scenarios, special plateau climates such as blizzards, strong winds, low visibility, and freezing fog are ignored, as well as complex traffic scenarios such as mixed traffic, interference from non-motorized vehicles or livestock, and road construction. Spatiotemporal scenarios are also only divided into daytime, dawn / dusk, and nighttime based on lighting conditions, without considering scenarios that are prone to sudden cognitive delays, such as driving times and sudden changes in altitude. These scenarios significantly increase the difficulty of drivers' perception and judgment, exacerbating cognitive delays, yet they are not included in the existing compensation system. On the other hand, the compensation methods are relatively simple, failing to establish a "scenario-impact degree-compensation coefficient" system. The dynamic correlation of altitude compensation is not addressed by using fixed interval divisions and uniform compensation standards within the same interval. This ignores the linear impact of continuous altitude changes on cognitive delay and lacks quantifiable compensation coefficients for different scenarios. It also fails to provide tiered compensation based on scenario complexity and does not consider individual differences among drivers, using the average driver or the most unfavorable driver as the design benchmark. This results in either overly conservative designs that increase engineering costs or insufficient safety guarantees for specific driving groups such as novice drivers and those with short acclimatization periods at high altitudes. Furthermore, the scenario compensation standards of different methods are incompatible, failing to form a unified compensation system for high-altitude highways. The altitude division standards of the stopping sight distance method and the alignment design method are inconsistent, and illumination compensation and physiological and psychological load compensation are disconnected. Moreover, a collaborative compensation mechanism for multiple scenarios has not been established. Faced with scenarios with multiple overlapping factors such as high altitude, curves and slopes, cliffs, and nighttime, the existing single-scenario compensation methods cannot meet the actual safety design requirements.

[0007] In summary, existing technologies have significant shortcomings in terms of the precise quantification of cognitive delay time, the comprehensiveness and targeted compensation for scenario differences, the lack of synergy among different methods, and the weak data verification foundation. As a result, the correction of the curve radius on plateau highways cannot accurately match the actual driving risks, which may cause safety hazards and unnecessary engineering waste. Summary of the Invention

[0008] Based on the problems raised in the background technology, the purpose of this invention is to provide a method and related equipment for correcting the radius of a highway based on cognitive delay in a low-oxygen environment. This solves the problem that the existing technology has obvious shortcomings in terms of the fine quantification of cognitive delay time, the comprehensiveness and targeted compensation for different scenarios, the lack of synergy between different methods, and the weak data verification foundation, which leads to the problem that the correction of the radius of a curve on a plateau highway cannot accurately match the actual driving risks.

[0009] This invention is achieved through the following technical solution:

[0010] The first aspect of this invention provides a method for correcting highway radius based on cognitive delay under hypoxic conditions, comprising the following steps:

[0011] Step S1: Collect multimodal physiological data of drivers under hypoxic conditions and construct a dynamically updated altitude-cognition mapping model;

[0012] Step S2: Obtain the basic cognitive delay time based on the altitude-cognition mapping relationship model, and calculate the comprehensive cognitive delay time at multiple time scales by combining real-time cognitive related parameters;

[0013] Step S3: Based on the multi-timescale comprehensive cognitive delay time, a scene-differentiated sight distance compensation strategy is adopted to compensate for parking sight distance, and the compensated parking sight distance is obtained.

[0014] Step S4: Determine the sight distance constraint radius based on the compensated parking sight distance, and generate the minimum radius of the horizontal curve of the highway under low oxygen conditions based on the sight distance constraint radius.

[0015] In the above technical solution, multimodal physiological data that can reflect the driver's cognitive state are collected comprehensively, such as electroencephalogram (EEG, which directly reflects brain activity), eye movement data (which reflects attention allocation and perception process), and electrocardiogram. A dynamically updated altitude-cognition mapping relationship model is constructed using multimodal physiological data. This model learns the relationship between cognitive features extracted from multimodal physiological data and altitude to quantify the dynamic relationship between altitude and driver's cognitive ability.

[0016] The basic cognitive delay time is obtained based on the altitude-cognitive mapping model. This basic cognitive delay time is the delay time of the driver under standard hypoxic conditions at a specific altitude. This delay time only reflects the impact of the driver's physiology on driving state under certain hypoxic conditions, which is a static impact feedback. By combining the basic delay with real-time parameters, a multi-timescale comprehensive cognitive delay time for final design is calculated. This time can be dynamically adjusted for the most unfavorable scenario. It can dynamically predict based on the driver's current instantaneous state and the immediate environment to be faced, thus solving the problem of insufficient dynamism.

[0017] Driving safety in low-oxygen environments depends not only on the driver's physiological state but also on the driving scenario. For example, different driving scenarios, such as combined curves and slopes, or sharp curves, steep slopes, and cliff edges, require different safe braking spaces. Currently, parking sight distance compensation typically focuses on the core variable of cognitive delay time. However, this is problematic because: first, cognitive delay time is a static variable and cannot reflect the danger under extreme conditions; second, relying on cognitive delay time for parking sight distance compensation only considers the driver's physiological and psychological feedback to driving operations from a single perspective. Neither physiological nor psychological data can reflect the driving scenario as perceived by the driver's eyes. Compensating solely based on physiological and psychological data, without considering the driver's visual experience, may result in insufficient sight distance compensation, inadequate subsequent sight distance constraint radius, and ultimately, increased driving danger.

[0018] The calculated sight distance constraint radius is compared with the minimum radius of the horizontal curve determined based on vehicle dynamics and other factors. The maximum value is taken as the final design value, thereby ensuring that the designed road is both safe and in compliance with engineering specifications.

[0019] In one optional embodiment, multimodal physiological data of the driver are collected under hypoxic conditions, and a dynamically updated altitude-cognition mapping model is constructed, including the following steps:

[0020] Step S11: Clean the driver's multimodal physiological data, extract cognitive features from the cleaned driver's multimodal physiological data, and construct a structured cognitive feature dataset based on the cognitive features;

[0021] Step S12: Divide the structured cognitive feature dataset into hierarchical levels based on altitude, and map the structured cognitive feature datasets at different levels to generate an initial static mapping model;

[0022] Step S13: Deploy the initial static mapping model to the edge computing node to access the real-time collected physiological data, and use a sliding window to process the real-time collected physiological data to obtain recursive parameters;

[0023] Step S14: Update the initial static mapping model using recursive parameters to generate a dynamically updated altitude-cognition mapping relationship model.

[0024] In one optional embodiment, the driver's multimodal physiological data includes: electroencephalogram (EEG) signals, eye movement data, and blood oxygenation data;

[0025] The extraction of cognitive features from the cleaned driver's multimodal physiological data includes the following steps:

[0026] A sliding time window is determined by combining the EEG signals, eye movement data, and blood oxygenation data, and the EEG signals, eye movement data, and blood oxygenation data are segmented according to the sliding time window;

[0027] Frequency domain decomposition technology is used to separate the EEG signals in each time window into frequency bands to obtain frequency band data. The frequency band energy of the frequency band data is calculated, and a first cognitive load index is constructed based on the frequency band energy. Feature waveforms of the frequency band data are extracted, and an information processing speed index is constructed based on the feature waveforms.

[0028] Identify pupil trajectories within each time window, classify eye movement behaviors into fixation behaviors and saccade behaviors based on the pupil trajectories, calculate the duration and frequency of fixation behaviors, and construct a visual attention index; calculate the amplitude, peak velocity, and duration of saccade behaviors, and construct a second cognitive load index.

[0029] Calculate the blood oxygen deviation and blood oxygen fluctuation amplitude within each time window, and construct a third cognitive load index using the blood oxygen deviation and blood oxygen fluctuation amplitude.

[0030] In one optional embodiment, the basic cognitive delay time is obtained based on the altitude-cognition mapping model, and the comprehensive cognitive delay time across multiple time scales is calculated by combining real-time cognitive-related parameters, including the following steps:

[0031] Step S21: Obtain the real-time altitude and map the real-time altitude to the altitude-cognition mapping relationship model to obtain the basic cognitive delay time;

[0032] Step S22: Extract cognitive state features from real-time cognitive-related parameters using a sliding window, and integrate the extracted cognitive state features to form a set of cognitive state parameters;

[0033] Step S23: Construct a time scale, use the time scale to perform scale-specific correction on the cognitive state parameter set, and adaptively fuse the corrected cognitive state parameter set to obtain a multi-time scale comprehensive cognitive delay time.

[0034] In one optional embodiment, a scene-differentiated sight distance compensation strategy is used for parking sight distance compensation, including the following steps:

[0035] Step S31: Divide the highway scene into gridded information units, and use the multi-timescale comprehensive cognitive delay time to perform cognitive load-scene complexity coupling on the gridded information units to obtain the scene hazard coefficient;

[0036] Step S32: Calculate the standard parking sight distance, decompose the standard parking sight distance into a two-dimensional parking sight distance matrix, and compensate the standard parking sight distance according to the scene topology using the scene hazard coefficient to obtain the compensated parking sight distance.

[0037] In one optional embodiment, the highway scene is divided into gridded information units, and the cognitive load-scene complexity of the gridded information units is coupled by the multi-timescale integrated cognitive delay time, including the following steps:

[0038] Step S311: Divide the highway scene into several gridded information units with the driver as the center, the driving direction as the longitudinal direction, and the perpendicular driving direction as the lateral direction, and calculate the visual information entropy of each gridded information unit.

[0039] Step S312: Map the multi-timescale integrated cognitive delay time to the gridded information unit to form a cognitive load spatial distribution;

[0040] Step S313: The spatial distribution of cognitive load and the visual information entropy are resonantly fused through a nonlinear coupling function to output a gridded scene hazard coefficient.

[0041] In one optional embodiment, the standard parking sight distance is decomposed into a two-dimensional parking sight distance matrix, and the standard parking sight distance is compensated according to the scene hazard coefficient based on the scene topology, including the following steps:

[0042] Step S321: Decompose the standard parking sight distance into a two-dimensional parking sight distance matrix with angular latitude and distance latitude; wherein, the angular latitude covers the driver's effective field of vision;

[0043] Step S322: Map the multi-timescale integrated cognitive delay time into space to generate a cognitive compensation field with the same dimension as the two-dimensional parking sight distance matrix;

[0044] Step S323: Perform feature analysis on the scene topology based on the scene hazard coefficient to obtain scene features, and determine the compensation mode based on the scene features;

[0045] Step S324: Based on the determined compensation mode, the two-dimensional parking sight distance matrix, the cognitive compensation field, and the scene hazard coefficient are fused nonlinearly at the element level to obtain the compensated parking sight distance.

[0046] A second aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a method for correcting highway radius based on cognitive delay under hypoxic conditions.

[0047] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for correcting highway radius based on cognitive delay under low-oxygen conditions.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] 1. By constructing a dynamically updated altitude-cognition mapping model, a refined breakdown and dynamic prediction of the driver's cognitive delay time can be achieved;

[0050] 2. By calculating the comprehensive cognitive delay time across multiple time scales and combining it with the formulation of scene-differentiated viewing distance compensation strategies, we can achieve refined compensation for various complex scenes and collaborative compensation under the superposition of multiple scenes.

[0051] 3. The compensated parking sight distance is converted into the minimum radius index of the horizontal curve, and a full-chain correction system is constructed from the driver's psychological load to the road design parameters. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0053] Figure 1 This is a flowchart illustrating the road radius correction method based on cognitive delay under low oxygen conditions provided in Embodiment 1 of the present invention.

[0054] Figure 2 This is a schematic diagram of the structure of an electronic device provided in Embodiment 2 of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0056] This embodiment 1 provides a method for correcting highway radius based on cognitive delay under hypoxic conditions, such as... Figure 1 As shown, the highway radius correction method based on cognitive delay under hypoxic conditions includes the following steps:

[0057] Step S1: Collect multimodal physiological data of drivers under hypoxic conditions and construct a dynamically updated altitude-cognition mapping model;

[0058] Step S2: Obtain the basic cognitive delay time based on the altitude-cognition mapping relationship model, and calculate the comprehensive cognitive delay time at multiple time scales by combining real-time cognitive related parameters;

[0059] Step S3: Based on the multi-timescale comprehensive cognitive delay time, a scene-differentiated sight distance compensation strategy is adopted to compensate for parking sight distance, and the compensated parking sight distance is obtained.

[0060] Step S4: Determine the sight distance constraint radius based on the compensated parking sight distance, and generate the minimum radius of the horizontal curve of the highway under low oxygen conditions based on the sight distance constraint radius.

[0061] It should be noted that a comprehensive collection of multimodal physiological data that can reflect the driver's cognitive state, such as electroencephalograms (which directly reflect brain activity), eye-tracking data (which reflect attention allocation and perception processes), and electrocardiograms, is used to construct a dynamically updated altitude-cognitive mapping model. This model learns the relationship between cognitive features extracted from multimodal physiological data and altitude to quantify the dynamic relationship between altitude and the driver's cognitive ability.

[0062] The basic cognitive delay time is obtained based on the altitude-cognitive mapping model. This basic cognitive delay time is the delay time of the driver under standard hypoxic conditions at a specific altitude. This delay time only reflects the impact of the driver's physiology on driving state under certain hypoxic conditions, which is a static impact feedback. By combining the basic delay with real-time parameters, a multi-timescale comprehensive cognitive delay time for final design is calculated. This time can be dynamically adjusted for the most unfavorable scenario. It can dynamically predict based on the driver's current instantaneous state and the immediate environment to be faced, thus solving the problem of insufficient dynamism.

[0063] Driving safety in low-oxygen environments depends not only on the driver's physiological state but also on the driving scenario. For example, different driving scenarios, such as combined curves and slopes, or sharp curves, steep slopes, and cliff edges, require different safe braking spaces. Currently, parking sight distance compensation typically focuses on the core variable of cognitive delay time. However, this is problematic because: first, cognitive delay time is a static variable and cannot reflect the danger under extreme conditions; second, relying on cognitive delay time for parking sight distance compensation only considers the driver's physiological and psychological feedback to driving operations from a single perspective. Neither physiological nor psychological data can reflect the driving scenario as perceived by the driver's eyes. Compensating solely based on physiological and psychological data, without considering the driver's visual experience, may result in insufficient sight distance compensation, inadequate subsequent sight distance constraint radius, and ultimately, increased driving danger.

[0064] The calculated sight distance constraint radius is compared with the minimum radius of the horizontal curve determined based on vehicle dynamics and other factors. The maximum value is taken as the final design value, thereby ensuring that the designed road is both safe and in compliance with engineering specifications.

[0065] From collecting multimodal physiological data to establishing an altitude-cognitive mapping model (quantifying driver physiological data), to dynamically calculating the comprehensive cognitive delay time across multiple time scales, to differentiated sight distance compensation (matching the driver's environment), and finally to generating the minimum horizontal curve radius, a complete chain correction system of "physiological and psychological load - cognitive delay - road design" is formed.

[0066] In one optional embodiment, multimodal physiological data of the driver are collected under hypoxic conditions, and a dynamically updated altitude-cognition mapping model is constructed, including the following steps:

[0067] Step S11: Clean the driver's multimodal physiological data, extract cognitive features from the cleaned driver's multimodal physiological data, and construct a structured cognitive feature dataset based on the cognitive features;

[0068] Step S12: Divide the structured cognitive feature dataset into hierarchical levels based on altitude, and map the structured cognitive feature datasets at different levels to generate an initial static mapping model;

[0069] Step S13: Deploy the initial static mapping model to the edge computing node to access the real-time collected physiological data, and use a sliding window to process the real-time collected physiological data to obtain recursive parameters;

[0070] Step S14: Update the initial static mapping model using recursive parameters to generate a dynamically updated altitude-cognition mapping relationship model.

[0071] It should be noted that traditional methods for determining the minimum safe radius of horizontal curves on highways typically directly correlate physiological and psychological factors with the minimum safe radius to ensure that it meets the driving needs of drivers in the low-oxygen environment of high-altitude areas. However, this method only partially fills the gap in the design of high-altitude highways that neglects the physiological and psychological factors of drivers, providing a preliminary technical approach for curve radius correction in low-oxygen environments. The impact of the delay in the driver's judgment and decision-making process in the low-oxygen environment of high altitudes is far greater than that in the operational process, directly leading to a lack of specificity in the relevant correction factors, and a significant deficiency in the coverage of factors affecting cognitive delay.

[0072] Therefore, this embodiment constructs a "physiological and psychological load-cognitive delay-road design" model. First, physiological data is quantified into cognitive data, and then the cognitive data is quantified into a comprehensive cognitive delay time with multiple time scales related to braking reaction time and braking operation time. In this way, the comprehensive cognitive delay time is used as a quantitative correlation bridge between the driver's physiological and psychological load and the road alignment design, realizing the accurate mapping between physiological and psychological load and road design indicators, and effectively improving the accuracy and adaptability of road alignment design in high-altitude and low-oxygen environments.

[0073] Specifically, this step constructs a hierarchical feature engineering system from raw physiological signals to structured cognitive representations. For hypoxic driving scenarios, multimodal physiological data of drivers are collected simultaneously, and cognitive features are extracted from the cleaned multimodal physiological data. A structured cognitive feature dataset is then constructed, overcoming the limitations of isolated extraction of single-modal features and establishing cross-modal consistency constraints for cognitive load.

[0074] In this embodiment, the altitude is divided into four levels:

[0075] Low altitude: 0-1000m, cognitive baseline;

[0076] Mid-altitude: 1000-2500m, mild hypoxia;

[0077] High altitude: 2500-4000m, significantly low oxygen levels;

[0078] Extremely high altitude: >4000m, severe hypoxia.

[0079] Dividing altitude into several discrete intervals means that, theoretically, a driver's cognitive abilities (such as reaction time) should change continuously and slowly as they travel from 3500m to 3501m. However, existing discrete models may produce sudden jumps in evaluation values ​​when entering the next interval. Such jumps do not conform to physiological reality and can lead to unreasonable fluctuations in safety designs (such as sight distance calculations) at this boundary.

[0080] The Wasserstein distance is used to map and align the feature distributions of adjacent altitude levels. By solving the optimal transmission scheme from low altitude distribution to high altitude distribution, a mathematical mapping relationship is established to characterize the cognitive feature drift caused by altitude changes. Based on this, an initial static mapping model is generated. This model uses discrete altitude levels as nodes and stores the mapping parameters between levels, providing a foundation for subsequent smooth cognitive assessment of continuous altitude changes.

[0081] By calculating the Wasserstein distance between function families of adjacent levels, cognitive appraisal jumps at level boundaries are eliminated, achieving smooth cognitive appraisal transitions when altitude changes continuously, thus overcoming the drift problem of cognitive feature distribution between adjacent altitude levels.

[0082] A multi-timescale sliding window is used for streaming feature extraction. An adaptive Kalman filter is constructed, using the static mapping output as the predicted value and the window features as the observed value. The latent variables of the cognitive state are estimated in real time and the Kalman gain is dynamically adjusted. The output includes recursive parameters such as the state estimate, error covariance, information sequence, and adaptive forgetting factor. Using the recursive parameters, Bayesian online learning is used to update the static mapping hyperparameters as priors and sufficient statistics as likelihoods through natural gradient descent. At the same time, active cognitive probe events are triggered based on prediction uncertainty to achieve closed-loop active sampling. Finally, a dynamically updated altitude-cognitive mapping relationship model that integrates the individual's recent cognitive history correction term is generated.

[0083] In one optional embodiment, the driver's multimodal physiological data includes: electroencephalogram (EEG) signals, eye movement data, and blood oxygenation data;

[0084] The extraction of cognitive features from the cleaned driver's multimodal physiological data includes the following steps:

[0085] A sliding time window is determined by combining the EEG signals, eye movement data, and blood oxygenation data, and the EEG signals, eye movement data, and blood oxygenation data are segmented according to the sliding time window;

[0086] Frequency domain decomposition technology is used to separate the EEG signals in each time window into frequency bands to obtain frequency band data. The frequency band energy of the frequency band data is calculated, and a first cognitive load index is constructed based on the frequency band energy. Feature waveforms of the frequency band data are extracted, and an information processing speed index is constructed based on the feature waveforms.

[0087] Identify pupil trajectories within each time window, classify eye movement behaviors into fixation behaviors and saccade behaviors based on the pupil trajectories, calculate the duration and frequency of fixation behaviors, and construct a visual attention index; calculate the amplitude, peak velocity, and duration of saccade behaviors, and construct a second cognitive load index.

[0088] Calculate the blood oxygen deviation and blood oxygen fluctuation amplitude within each time window, and construct a third cognitive load index using the blood oxygen deviation and blood oxygen fluctuation amplitude.

[0089] It should be noted that the cleaned driver's multimodal physiological data was acquired, including EEG signals, eye movement data, and blood oxygenation data. To simultaneously analyze the multimodal data, a unified sliding time window was determined by integrating the timestamps of the EEG signals, eye movement data, and blood oxygenation data. Specifically, the non-stationarity index of the EEG signals was calculated, the fixation-saccade transition frequency of the eye movement data was extracted, and the rate of blood oxygenation decline in the blood oxygenation data was monitored. A multimodal window length voting device was constructed, and fuzzy logic reasoning was used to synthesize the suggestions from the three modalities to determine the most suitable time window.

[0090] For electroencephalogram (EEG) signals, wavelet packet decomposition is used to decompose the EEG signals into frequency bands. (1 to 4 Hz) (4 to 8 Hz) (8 to 13 Hz) (13 to 30 Hz) and γ (30 to 100 Hz) are used to achieve time-frequency localization analysis of non-stationary EEG signals. The power spectral density of each frequency band is integrated to obtain the frequency band energy. The logarithmic energy density and energy percentage of each frequency band are calculated as the primary cognitive load indicator.

[0091] For event-related potential components, the matching pursuit algorithm was used to extract the P300 component (time from stimulus onset to peak). This data represents the speed of attentional resource allocation and decision-making updates, quantifying the brain's processing speed of external information, i.e., the driver's reaction speed in a hypoxic environment. An information processing speed index was constructed based on the P300 latency.

[0092] ;

[0093] In the above formula, As an indicator of information processing speed, The peak time of the P300 waveform. The base time for the P300 waveform. This is an adjustment factor.

[0094] Identify the pupil trajectory within each time window, and classify eye movement behavior into fixation behavior and saccade behavior based on the motion state of the pupil trajectory. Fixation behavior is defined as pupil velocity below 30° / s and duration > 100ms, while saccade behavior is defined as pupil velocity above 100° / s and duration between 20-200ms.

[0095] Calculate the duration and frequency of fixation behavior, and construct visual attention metrics:

[0096] ;

[0097] In the above formula, As a visual attention indicator, For fixation frequency, The average duration of fixation behavior. For the visual point in the pupil trajectory to fall into the first The probability of a grid. Here, the grid refers to the scene area within the driver's field of vision.

[0098] Calculate the amplitude, peak velocity, and duration of saccades, and construct a second cognitive load indicator:

[0099] ;

[0100] In the above formula, As the second cognitive load indicator, The main sequence scan amplitude, This represents the actual scanning range. For the scanning range, The average duration of saccades, The base duration for the scanning continuation.

[0101] A third cognitive workload indicator was constructed using blood oxygen deviation and blood oxygen fluctuation amplitude.

[0102] ;

[0103] In the above formula, As the third cognitive load indicator, , , As weight, Blood oxygen deviation, This refers to the amplitude of blood oxygen fluctuations. This refers to the rate of change in blood oxygen saturation.

[0104] Furthermore, the structured cognitive feature dataset is hierarchically divided based on altitude, and the structured cognitive feature datasets at different levels are mapped, including:

[0105] Univariate marginal distribution estimation was performed on the cognitive load index, information processing speed index, and visual attention index at each level, and a multivariate joint distribution was constructed using a Gaussian function to finally obtain the conditional distribution probability.

[0106] Obtain the conditional distribution probabilities of adjacent levels and calculate their Wasserstein distance, which quantifies the transport cost of cognitive feature distributions between levels.

[0107] For each self-proclaimed independent multi-output Gaussian process regression model, Wasserstein distance is used as a constraint at the physical level boundary to achieve a smooth transition.

[0108] In one optional embodiment, the basic cognitive delay time is obtained based on the altitude-cognition mapping model, and the comprehensive cognitive delay time across multiple time scales is calculated by combining real-time cognitive-related parameters, including the following steps:

[0109] Step S21: Obtain the real-time altitude and map the real-time altitude to the altitude-cognition mapping relationship model to obtain the basic cognitive delay time;

[0110] Step S22: Extract cognitive state features from real-time cognitive-related parameters using a sliding window, and integrate the extracted cognitive state features to form a set of cognitive state parameters;

[0111] Step S23: Construct a time scale, use the time scale to perform scale-specific correction on the cognitive state parameter set, and adaptively fuse the corrected cognitive state parameter set to obtain a multi-time scale comprehensive cognitive delay time.

[0112] It should be noted that Gaussian process regression is used to predict the expected distribution of cognitive features at the current altitude, and the basic cognitive delay time is mapped through the physiological-time transformation function.

[0113] In this embodiment, the physiological-time conversion function is:

[0114] ;

[0115] In the above formula, Basic cognitive delay time, , , , , The coefficient is determined through preliminary driving simulation tests. , , , This represents the expected distribution of cognitive features at the corresponding level.

[0116] For three types of driving scenarios—emergency, routine, and complex—specific correction functions are constructed for instantaneous, short-term, and long-term time scales, respectively. These functions integrate basic delay and real-time cognitive parameters, introduce adaptive weight configuration driven by scenario hazard level, and output multi-time-scale cognitive delay time with three-scale components and comprehensive uncertainty through weighted fusion and cognitive uncertainty buffering calculation. This enables accurate quantification of cognitive delay from static altitude mapping to dynamic cognitive state and from single time scale to multi-scale collaboration.

[0117] In one optional embodiment, a scene-differentiated sight distance compensation strategy is used for parking sight distance compensation, including the following steps:

[0118] Step S31: Divide the highway scene into gridded information units, and use the multi-timescale comprehensive cognitive delay time to perform cognitive load-scene complexity coupling on the gridded information units to obtain the scene hazard coefficient;

[0119] Step S32: Calculate the standard parking sight distance, decompose the standard parking sight distance into a two-dimensional parking sight distance matrix, and compensate the standard parking sight distance according to the scene topology using the scene hazard coefficient to obtain the compensated parking sight distance.

[0120] It should be noted that, driven by the comprehensive cognitive delay time at multiple time scales, a polar coordinate dynamic grid system is established that moves with the vehicle. The instantaneous, short-term, and long-term cognitive delay components are mapped to the spatial influence range of the near, mid, and far fields. Through the calculation of scene complexity by gridded cognitive load distribution and fusion of multi-source perception data, the hyperbolic tangent resonance function is used to achieve nonlinear coupling between cognitive load and scene complexity, generating a gridded scene hazard coefficient field. After spatial aggregation, the instantaneous, cumulative, and direction-specific multi-scale hazard coefficients and dominant hazard locations are output, providing spatially accurate hazard distribution input for line-of-sight compensation.

[0121] The standard parking sight distance is decomposed into a two-dimensional angle-distance distribution matrix. The cognitive delay distance is distributed non-uniformly according to the cognitive load of the grid. Based on the connected domain topology analysis of the scene hazard coefficient, three compensation modes are adaptively selected: linear, non-linear amplification, or exponential conservative. The standard sight distance, cognitive compensation, and scene hazard coefficient are integrated through element-level fusion operation. The cognitive uncertainty is dynamically buffered and then constrained. After post-processing, a compensated two-dimensional sight distance matrix and a direction-specific sight distance curve are generated, realizing the dynamic generation of personalized safe sight distance from "vehicle adapting to road" to "road adapting to human cognitive state".

[0122] In one optional embodiment, the highway scene is divided into gridded information units, and the cognitive load-scene complexity of the gridded information units is coupled by the multi-timescale integrated cognitive delay time, including the following steps:

[0123] Step S311: Divide the highway scene into several gridded information units with the driver as the center, the driving direction as the longitudinal direction, and the perpendicular driving direction as the lateral direction, and calculate the visual information entropy of each gridded information unit.

[0124] Step S312: Map the multi-timescale integrated cognitive delay time to the gridded information unit to form a cognitive load spatial distribution;

[0125] Step S313: The spatial distribution of cognitive load and the visual information entropy are resonantly fused through a nonlinear coupling function to output a gridded scene hazard coefficient.

[0126] It should be noted that the vertical direction aims to focus on the safe attention distance determined by cognitive delay, while the horizontal direction aims to focus on the expansion of the visual field when attention is distracted. For the image region corresponding to each grid cell projection, the visual information entropy is calculated, including:

[0127] Scene texture complexity Extract the local binary pattern histogram of the highway scene and calculate its Shannon entropy;

[0128] Color variation : Convert the highway scene to the Lab color space and calculate the two-dimensional entropy of the color distribution in the ab plane;

[0129] Edge density : Calculate the edge direction histogram entropy after using Canny edge detection;

[0130] Combining the above three factors, we obtain the visual information entropy:

[0131] ;

[0132] In the above formula, For visual information entropy, , , Weights corresponding to texture, color, and edge. This represents the maximum complexity of the scene texture. This represents the maximum value of color variation. This represents the maximum edge density.

[0133] The time-space transformation is achieved through mapping. After the transformation, cognitive load is allocated to each grid cell according to the longitudinal distance. The longitudinal distance is divided into near-field grid (close to the driver), mid-field grid, and far-field grid (far from the driver). The near-field grid is dominated by acute cognitive load, such as hypoxia stimulation and information processing efficiency, and the load is allocated using relevant factors. The mid-field grid is dominated by factors such as attentional distraction. The far-field grid is dominated by long-term cognitive factors such as blood oxygen accumulation.

[0134] Anisotropic diffusion filtering is used to eliminate discrete noise in the grid. The process is iterated until convergence, and the output is a continuous cognitive load distribution.

[0135] Spatial correlation calculation is performed on the spatial distribution of cognitive load and visual information entropy, and the hyperbolic tangent resonance kernel is used for calculation. In low cognitive load or low complexity scenarios, the tan function input is small and the output is in the linear region, and the risk coefficient increases moderately. In high cognitive load and high complexity scenarios, the tan function saturates to 1, and the superposition resonance amplifies the risk coefficient, thus avoiding the misjudgment of "medium + medium = high risk" caused by simple linear superposition.

[0136] In one optional embodiment, the standard parking sight distance is decomposed into a two-dimensional parking sight distance matrix, and the standard parking sight distance is compensated according to the scene hazard coefficient based on the scene topology, including the following steps:

[0137] Step S321: Decompose the standard parking sight distance into a two-dimensional parking sight distance matrix with angular latitude and distance latitude; wherein, the angular latitude covers the driver's effective field of vision;

[0138] Step S322: Map the multi-timescale integrated cognitive delay time into space to generate a cognitive compensation field with the same dimension as the two-dimensional parking sight distance matrix;

[0139] Step S323: Perform feature analysis on the scene topology based on the scene hazard coefficient to obtain scene features, and determine the compensation mode based on the scene features;

[0140] Step S324: Based on the determined compensation mode, the two-dimensional parking sight distance matrix, the cognitive compensation field, and the scene hazard coefficient are fused nonlinearly at the element level to obtain the compensated parking sight distance.

[0141] In this embodiment, the formula for the minimum radius of a circular curve based on line-of-sight is established as a formula for the minimum radius of a circular curve applicable to the high-altitude environment with low oxygen levels:

[0142]

[0143] in, The unit for compensating for the rear parking sight distance is meters. The inner lateral clearance represents the line-of-sight offset, in meters (m).

[0144] Embodiment 2 of the present invention provides an electronic device, such as... Figure 2 As shown, the electronic device includes a processor 21, a memory 22, an input device 23, and an output device 24; the number of processors 21 in the computer device can be one or more. Figure 2 Taking a processor 21 as an example; the processor 21, memory 22, input device 23, and output device 24 in an electronic device can be connected via a bus or other means. Figure 2 Taking the example of a connection between China and Israel via a bus.

[0145] The memory 22, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules. The processor 21 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 22, thereby implementing the road radius correction method based on cognitive delay in a low-oxygen environment as described in Embodiment 1.

[0146] The memory 22 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 22 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory 22 may further include memory remotely located relative to the processor 21, which can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0147] Input device 23 can be used to receive user input such as ID and password. Output device 24 is used to output the network configuration page.

[0148] Embodiment 3 of the present invention also provides a computer-readable storage medium, wherein the computer-executable instructions, when executed by a computer processor, are used to implement the road radius correction method based on cognitive delay under low oxygen conditions as provided in Embodiment 1.

[0149] The storage medium containing computer-executable instructions provided in the embodiments of the present invention is not limited to the method operation provided in Embodiment 1, but can also perform related operations in the road radius correction method based on cognitive delay under low oxygen environment provided in any embodiment of the present invention.

[0150] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for highway radius correction based on cognitive delay in hypoxic environment, characterized by, Step S1: Collect multimodal physiological data of drivers under hypoxic conditions and construct a dynamically updated altitude-cognition mapping model; Step S2: Obtain the basic cognitive delay time based on the altitude-cognition mapping relationship model, and calculate the comprehensive cognitive delay time at multiple time scales by combining real-time cognitive related parameters; Step S3: Based on the multi-timescale comprehensive cognitive delay time, a scene-differentiated sight distance compensation strategy is adopted to compensate for parking sight distance, and the compensated parking sight distance is obtained. Step S4: Determine the sight distance constraint radius based on the compensated parking sight distance, and generate the minimum radius of the horizontal curve of the highway under low oxygen conditions based on the sight distance constraint radius; Multimodal physiological data of drivers were collected under hypoxic conditions, and a dynamically updated altitude-cognition mapping model was constructed, including the following steps: Step S11: Clean the driver's multimodal physiological data, extract cognitive features from the cleaned driver's multimodal physiological data, and construct a structured cognitive feature dataset based on the cognitive features; Step S12: Divide the structured cognitive feature dataset into hierarchical levels based on altitude, and map the structured cognitive feature datasets at different levels to generate an initial static mapping model; Step S13: Deploy the initial static mapping model to the edge computing node to access the real-time collected physiological data, and use a sliding window to process the real-time collected physiological data to obtain recursive parameters; Step S14: Update the initial static mapping model using recursive parameters to generate a dynamically updated altitude-cognitive mapping relationship model; A scenario-differentiated sight distance compensation strategy is adopted for parking sight distance compensation, including the following steps: Step S31: Divide the highway scene into gridded information units, and use the multi-timescale comprehensive cognitive delay time to perform cognitive load-scene complexity coupling on the gridded information units to obtain the scene hazard coefficient; Step S32: Calculate the standard parking sight distance, decompose the standard parking sight distance into a two-dimensional parking sight distance matrix, and compensate the standard parking sight distance according to the scene topology using the scene hazard coefficient to obtain the compensated parking sight distance. The highway scene is divided into gridded information units, and the cognitive load-scene complexity of the gridded information units is coupled by the multi-timescale comprehensive cognitive delay time, including the following steps: Step S311: Divide the highway scene into several gridded information units with the driver as the center, the driving direction as the longitudinal direction, and the perpendicular driving direction as the lateral direction, and calculate the visual information entropy of each gridded information unit. Step S312: Map the multi-timescale integrated cognitive delay time to the gridded information unit to form a cognitive load spatial distribution; Step S313: The spatial distribution of cognitive load and the visual information entropy are resonantly fused through a nonlinear coupling function to output a gridded scene hazard coefficient.

2. The method of claim 1, wherein the method is based on a cognitive delay in a hypoxic environment. The driver's multimodal physiological data includes: electroencephalogram (EEG) signals, eye movement data, and blood oxygen saturation data; The extraction of cognitive features from the cleaned driver's multimodal physiological data includes the following steps: A sliding time window is determined by combining the EEG signals, eye movement data, and blood oxygenation data, and the EEG signals, eye movement data, and blood oxygenation data are segmented according to the sliding time window; Frequency domain decomposition technology is used to separate the EEG signals in each time window into frequency bands to obtain frequency band data. The frequency band energy of the frequency band data is calculated, and a first cognitive load index is constructed based on the frequency band energy. Feature waveforms of the frequency band data are extracted, and an information processing speed index is constructed based on the feature waveforms. Identify pupil trajectories within each time window, classify eye movement behaviors into fixation behaviors and saccade behaviors based on the pupil trajectories, calculate the duration and frequency of fixation behaviors, and construct a visual attention index; calculate the amplitude, peak velocity, and duration of saccade behaviors, and construct a second cognitive load index. Calculate the blood oxygen deviation and blood oxygen fluctuation amplitude within each time window, and construct a third cognitive load index using the blood oxygen deviation and blood oxygen fluctuation amplitude.

3. The method of claim 1, wherein the method is based on a cognitive delay in a hypoxic environment. Based on the altitude-cognition mapping model, the basic cognitive delay time is obtained, and the comprehensive cognitive delay time across multiple time scales is calculated by combining real-time cognitive-related parameters, including the following steps: Step S21: Obtain the real-time altitude and map the real-time altitude to the altitude-cognition mapping relationship model to obtain the basic cognitive delay time; Step S22: Extract cognitive state features from real-time cognitive-related parameters using a sliding window, and integrate the extracted cognitive state features to form a set of cognitive state parameters; Step S23: Construct a time scale, use the time scale to perform scale-specific correction on the cognitive state parameter set, and adaptively fuse the corrected cognitive state parameter set to obtain a multi-time scale comprehensive cognitive delay time.

4. The method of claim 1, wherein the method is based on a cognitive delay in a hypoxic environment. The standard parking sight distance is decomposed into a two-dimensional parking sight distance matrix, and the standard parking sight distance is compensated according to the scene topology using the scene hazard coefficient, including the following steps: Step S321: Decompose the standard parking sight distance into a two-dimensional parking sight distance matrix with angular latitude and distance latitude; wherein, the angular latitude covers the driver's effective field of vision; Step S322: Map the multi-timescale integrated cognitive delay time into space to generate a cognitive compensation field with the same dimension as the two-dimensional parking sight distance matrix; Step S323: Perform feature analysis on the scene topology based on the scene hazard coefficient to obtain scene features, and determine the compensation mode based on the scene features; Step S324: Based on the determined compensation mode, the two-dimensional parking sight distance matrix, the cognitive compensation field, and the scene hazard coefficient are fused nonlinearly at the element level to obtain the compensated parking sight distance.

5. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the road radius correction method based on cognitive delay under hypoxic conditions as described in any one of claims 1 to 4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for correcting highway radius based on cognitive delay under hypoxic conditions as described in any one of claims 1 to 4.

Images