First aid transfer path intelligent recommendation method and system based on AI technology

Abstract

The present application relates to the field of first-aid medical information technology, and more particularly to a first-aid transfer path intelligent recommendation method and system based on AI technology, comprising the following steps: S1, collecting patient vital sign data, vehicle operating state data and traffic condition data, performing cleaning and space-time alignment, and constructing a first-aid transfer state data set; S2, performing illness criticality assessment and hospital resource matching based on the first-aid transfer state data set, locking a target hospital through a two-way resource interaction protocol, and generating a main navigation path. In the present application, by uniformly aligning the patient vital signs, vehicle operating state and real-time road condition data in space-time, and introducing illness criticality assessment, hospital resource locking and trajectory re-planning based on medical operation stability, the first-aid transfer path can be dynamically adjusted according to the patient state and road conditions, thereby significantly improving the medical continuity and path decision accuracy of critical patients during the transfer process, and achieving the improvement of overall treatment efficiency.

Description

Technical Field

[0001] This invention relates to the field of emergency medical information technology, and in particular to an intelligent recommendation method and system for emergency transport routes based on AI technology. Background Technology

[0002] In emergency transport scenarios, patients' conditions often change rapidly during high-speed travel and complex road conditions. Medical staff need to complete critical medical operations under conditions of vehicle bumps, acceleration, or turns. Meanwhile, the command center needs to select the most suitable destination hospital and plan the optimal route for the vehicle within a limited time. Therefore, the route planning of emergency vehicles not only requires speed and accuracy, but also needs to be synchronized with the patient's vital signs, hospital resource status, and road conditions to ensure that the medical treatment process is not affected by changes in driving conditions, thereby achieving a safer and more efficient closed-loop dispatch of pre-hospital emergency care.

[0003] Most existing technologies rely solely on traditional navigation or route planning based on traffic data, lacking a comprehensive assessment of patient vital signs, the stability of medical procedures, and the real-time status of hospital resources. This makes it difficult to control vehicle acceleration and route during medical procedures, and also makes it difficult to dynamically adjust the transfer target hospital and route when hospital resources are strained or when there are sudden changes in demand. As a result, route decision-making is delayed, medical procedures are disrupted, and treatment efficiency is reduced. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides an intelligent recommendation method and system for emergency transport routes based on AI technology. It aims to improve the existing technology's insufficient comprehensive utilization of patient status, hospital resources, and vehicle driving stability during route planning and scheduling, which can easily lead to decision-making delays or medical operations being affected by driving during the transport process.

[0005] In a first aspect, the present invention provides the following technical solution: an intelligent recommendation method for emergency transport routes based on AI technology, comprising the following steps:

[0006] S1. Collect patient vital signs data, vehicle operation status data, and traffic condition data, clean and align them in time and space, and construct an emergency transport status dataset.

[0007] S2. Based on the emergency transport status dataset, assess the severity of the patient's condition and match hospital resources. Locate the target hospital through a two-way resource interaction protocol and generate the main navigation path.

[0008] S3. In response to the generation of the main navigation path, calculate the capacity coverage gap in the current area and send dispatch instructions to nearby idle vehicles;

[0009] S4. During the journey along the main navigation path, monitor the medical intervention action commands input by the vehicle terminal and analyze the stability constraint features corresponding to the action;

[0010] S5. Based on the stability constraint features and the features of the road ahead, perform local driving trajectory replanning on the main navigation path, generate the execution path, and output driving prompts.

[0011] Preferably, constructing the emergency transport status dataset specifically includes the following steps:

[0012] Acquire patients' electrocardiogram, blood pressure and blood oxygen data, acquire vehicle's three-axis acceleration and angular velocity data, and acquire road network traffic flow speed and road surface geometric feature data;

[0013] Poll the information systems of candidate hospitals within the target area to obtain data on emergency bed occupancy rate, equipment availability, and emergency triage queue time.

[0014] The collected data is filtered and interpolated.

[0015] The data is synchronized and aligned with the system time, and the road condition data is mapped with the vehicle location to generate a multi-dimensional state tensor.

[0016] Preferably, the process of locking onto the target hospital via a two-way resource interaction protocol specifically includes the following steps:

[0017] Based on the assessment results of the severity of the illness, the characteristics of medical resources, and the estimated travel time, a priority list of candidate hospitals is generated.

[0018] Send an exclusive resource lock request containing the patient ID, equipment requirements, and estimated arrival time to the first hospital in the priority list;

[0019] If a resource lock confirmation code is received from the first hospital within the preset window, then that hospital is established as the transfer destination.

[0020] If a rejection code or timeout is received within the preset window, the current hospital is removed from the list, and the next candidate hospital is selected to repeat the step of sending an exclusive resource lock request.

[0021] Preferably, calculating the current area's transportation capacity coverage gap specifically includes the following steps:

[0022] Based on historical emergency call data, extract the probability of emergency events occurring within the current time window for the grid where the current vehicle is located;

[0023] Calculate the backup response time for the nearest available emergency medical vehicle within the surrounding grid to reach the center of the grid.

[0024] Calculate the time delay increment of the substitute response time relative to the original response time of the current vehicle;

[0025] The regional emergency response risk index is calculated based on the probability of the emergency event and the time delay increment. If the regional emergency response risk index exceeds a preset threshold, it is determined that there is a capacity coverage gap.

[0026] Preferably, sending dispatch instructions to nearby idle vehicles specifically includes the following steps:

[0027] Identify the target areas where there are capacity coverage gaps, and identify the directly adjacent and next-to-nearest jurisdictions surrounding the target areas;

[0028] Select available candidate dispatch vehicles within the jurisdiction and calculate the secondary risk increment caused to the original jurisdiction by each candidate dispatch vehicle after it moves to the target filling area.

[0029] The target vehicle-location scheduling pair is solved based on the difference between the risk reduction in the target filling area and the secondary risk increment.

[0030] Based on the vehicle-location scheduling, a dynamic redeployment instruction containing the target coordinates and suggested path is generated and sent.

[0031] Preferably, the medical intervention action command input by the monitoring vehicle-mounted terminal specifically includes the following steps:

[0032] Collect voice control signals and touch screen operation signals from the medical cabin inside the vehicle;

[0033] Perform semantic recognition on the voice control signal and extract action category labels;

[0034] Detect the trigger status of touchscreen buttons and generate standard operation code;

[0035] The action category label and the standard operation code are fused to generate a verified medical intervention action instruction.

[0036] Preferably, the process of analyzing the stability constraint features corresponding to the action specifically includes the following steps:

[0037] Retrieve from a pre-set mapping library the standard operation duration, maximum permissible longitudinal acceleration threshold, and maximum permissible lateral acceleration threshold that match the verified medical intervention action command;

[0038] Calculate the dynamic stability driving window period by combining the current vehicle speed and road adhesion coefficient;

[0039] The longitudinal acceleration threshold, lateral acceleration threshold, and dynamic stability driving window period are combined into an instantaneous stability constraint feature vector.

[0040] Preferably, the local replanning of the main navigation path specifically includes the following steps:

[0041] Extract the number of lanes, radius of curvature, and lane smoothness index within a preset distance ahead of the main navigation path;

[0042] Generate candidate driving trajectories that include lane keeping, lane changing, and deceleration trajectories;

[0043] Simulate the vehicle's body posture as it travels along a candidate driving trajectory and predict the expected acceleration sequence;

[0044] The expected acceleration sequence is compared with the instantaneous stability constraint feature vector, and the trajectory that meets the constraints and has the shortest time consumption is selected as the target driving path.

[0045] Preferably, generating the execution path and outputting driving prompts specifically includes the following steps:

[0046] The target driving path is analyzed into a speed control profile and a steering guidance sequence;

[0047] A visual driving corridor covering the target driving path is projected in front of the field of vision using an in-vehicle display device;

[0048] Play voice commands to keep the vehicle in the driving corridor and a countdown timer for medical procedures;

[0049] The system collects actual vehicle acceleration feedback. If the indicator is detected to exceed the constraint characteristics, an audible and visual alarm is triggered.

[0050] Secondly, the present invention provides the following technical solution: an intelligent recommendation system for emergency transport routes based on AI technology, the system comprising:

[0051] The data acquisition module is used to collect patient vital signs data, vehicle operation status data, and traffic condition data, and to clean and align them in time and space to build an emergency transport status dataset.

[0052] The resource locking module is used to assess the severity of the patient's condition and match hospital resources based on the emergency transport status dataset, and to lock the target hospital and generate the main navigation path through a two-way resource interaction protocol.

[0053] The capacity scheduling module is used to calculate the capacity coverage gap in the current area and send scheduling instructions to nearby idle vehicles;

[0054] The motion monitoring module is used to monitor medical intervention motion commands input by the vehicle terminal and analyze the stability constraint characteristics corresponding to the motion.

[0055] The micro-execution module is used to perform local driving trajectory replanning on the main navigation path based on the stability constraint features and the features of the road segment ahead, generate the execution path, and output driving prompts.

[0056] The present invention has the following beneficial effects:

[0057] 1. In this invention, by unifying the spatiotemporal alignment of patient vital signs, vehicle operating status and real-time road condition data, and introducing the assessment of the severity of the illness, hospital resource locking and trajectory replanning based on the stability of medical operations, the emergency transport route can be dynamically adjusted according to the patient's condition and road conditions, thereby significantly improving the medical continuity and route decision accuracy of critically ill patients during the transport process, and achieving an overall improvement in treatment efficiency.

[0058] 2. In this invention, an emergency medical service capacity coverage gap determination mechanism is established by calculating the probability of emergency medical events at the city grid level and the backup response time, and dynamic dispatch instructions are automatically issued to nearby idle vehicles, so that the overall emergency medical resource layout is kept in real time and balanced, thereby effectively reducing the risk of decline in urban emergency medical response capacity caused by vehicle transfer or route deviation.

[0059] 3. In this invention, by constructing a mapping library between medical actions and vehicle acceleration thresholds and by performing acceleration constraint screening on the driving trajectory during medical operations, the vehicle can maintain a safe and controllable driving state while ensuring the stability of medical operations, thereby reducing the interference of vehicle driving dynamics on medical operations and improving the safety of the overall transportation process. Attached Figure Description

[0060] Figure 1 This is a flowchart of the intelligent recommendation method for emergency transport routes based on AI technology proposed in this invention.

[0061] Figure 2 This is an architecture diagram of the AI-based intelligent recommendation system for emergency transport routes proposed in this invention. Detailed Implementation

[0062] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1:

[0064] In the first embodiment of the present invention, the present invention provides an intelligent recommendation method for emergency transport routes based on AI technology, such as... Figure 1 As shown, it includes the following steps:

[0065] S1. Collect patient vital signs data, vehicle operation status data, and traffic condition data, clean and align them in time and space, and construct an emergency transport status dataset.

[0066] Furthermore, constructing the emergency transport status dataset specifically includes the following steps:

[0067] Acquire patients' electrocardiogram, blood pressure and blood oxygen data, acquire vehicle's three-axis acceleration and angular velocity data, and acquire road network traffic flow speed and road surface geometric feature data;

[0068] Poll the information systems of candidate hospitals within the target area to obtain data on emergency bed occupancy rate, equipment availability, and emergency triage queue time.

[0069] The collected data is filtered and interpolated.

[0070] The data is synchronized and aligned with the system time, and the road condition data is mapped with the vehicle location to generate a multi-dimensional state tensor.

[0071] Specifically, a unified master clock is used to continuously collect multi-source data in the emergency environment. Patient vital signs include instantaneous voltage values ​​of electrocardiogram signals, systolic and diastolic blood pressure values ​​output by monitoring equipment, and blood oxygen saturation values. Vehicle operating status is obtained through onboard inertial measurement units, which acquire triaxial acceleration and triaxial angular velocity values. Traffic conditions are provided by the road network monitoring platform, which provides traffic flow speed values ​​and corresponding geometric feature parameters of road segments. The geometric features include the radius of curvature and longitudinal slope angle of the road segment. All sensors and information systems generate their own original timestamps during sampling.

[0072] The acquired signals enter the cleaning module. Noise suppression of the ECG signal is achieved using a sliding window averaging filter. The filtered ECG voltage is defined as follows:

[0073] ;

[0074] in For a moment ECG filter voltage, This is the original ECG voltage. The time interval for ECG sampling. Let be the window length, and i be the index of each sampling point within the window, ranging from zero to N. 1. The blood pressure and blood oxygen values ​​are linearly interpolated based on adjacent sampling points to obtain a uniformly sampled isochronous sequence. The vehicle's three-axis acceleration and angular velocity signals are generated into effective motion observation values ​​through a low-pass digital filter. Traffic flow velocity data is filtered by median filtering to remove abrupt values. The curvature radius and slope angle of the road segment are directly obtained from the stable data provided by the road network platform.

[0075] The cleaned data is synchronized with time, and the system master clock defines a unified reference time series. any physical quantity The original sampling time is If its two most recent sampling times are and Then in the reference time The synchronization value at that location is defined as:

[0076] ;

[0077] in For time-aligned physical quantities, and To correspond to the original values, after synchronization processing, ECG voltage, blood pressure, blood oxygen, triaxial acceleration, triaxial angular velocity, traffic flow velocity, and road segment geometric characteristics all have a consistent time reference.

[0078] The vehicle's location information is spatially mapped to the road network. The vehicle's position is determined by the latitude and longitude values ​​provided by the positioning module, and the road segment number is determined by map matching methods. At each reference time The synchronized traffic flow speed is compared with the radius of curvature and slope angle of the corresponding road segment. This association aligns vehicle status with road condition status in space.

[0079] An emergency transport status dataset is constructed based on all synchronized information, with the state tensor at each reference time. Defined as

[0080] ;

[0081] in This is the filtered ECG voltage. For systolic pressure, For diastolic blood pressure, Blood oxygen saturation It is a triaxial acceleration. For triaxial angular velocity, For the road segment where the vehicle is located at the time Traffic flow speed, This is the geometric feature vector of the road segment, composed of the radius of curvature and the slope angle. It is obtained by analyzing all... The state tensor is continuously recorded to form a multidimensional state dataset covering the entire emergency transport process, which serves as the basic data input for subsequent assessment of the severity of the patient's condition, matching hospital resources, and generating the main navigation path.

[0082] S2. Based on the emergency transport status dataset, assess the severity of the patient's condition and match hospital resources. Locate the target hospital through a two-way resource interaction protocol and generate the main navigation path.

[0083] Furthermore, identifying the target hospital through a two-way resource interaction protocol specifically includes the following steps:

[0084] Based on the assessment results of the severity of the illness, the characteristics of medical resources, and the estimated travel time, a priority list of candidate hospitals is generated.

[0085] Send an exclusive resource lock request containing the patient ID, equipment requirement, and estimated arrival time to the top hospital in the priority list;

[0086] If a resource lock confirmation code is received from the first hospital within the preset window, that hospital will be designated as the transfer destination.

[0087] If a rejection code or timeout is received within the preset window, the current hospital is removed from the list, and the next candidate hospital is selected to repeat the step of sending an exclusive resource lock request.

[0088] Specifically, after obtaining the continuous emergency transport status tensor, the patient's immediate criticality score is calculated using a criticality assessment model. The model's input consists of vital sign variables from a reference time series, including filtered ECG values, systolic blood pressure values, diastolic blood pressure values, and blood oxygen saturation values. The output of the criticality scoring function is defined as the patient's current criticality index, and a linear weighted form is used to describe the basic implementation. The criticality index is defined as follows:

[0089] ;

[0090] in For a moment The severity index of the condition. The parameter weights are obtained by training based on historical labeled samples. These represent the synchronous vital sign values. A higher severity index indicates a greater degree of abnormality in vital signs. To enhance the model's adaptability, a nonlinear model structure can also be used to obtain equivalent values ​​from the same input source. The output will not affect the overall process of this method.

[0091] After obtaining the severity index of the patient's condition, the system continuously polls the medical institution systems within the target area. The hospital information systems provide values ​​for emergency bed occupancy rates, the availability of emergency equipment, and current emergency triage queue times. Let the set of hospitals be denoted as . The acceptability of each hospital's treatment capacity at any given moment is represented by the hospital resource availability index, which is defined as:

[0092] ;

[0093] in For the hospital Resource availability index, The percentage of available emergency beds provided to this hospital. This represents the idle status value of emergency medical equipment. This is the reciprocal of the emergency triage waiting time. This is a weighting coefficient determined by historical emergency records; the higher the value, the more suitable the hospital is for immediately admitting patients.

[0094] When identifying candidate hospitals, the system also needs to estimate the travel time from the vehicle's current location to each hospital. This is done by calculating the estimated travel time using real-time traffic flow speed data and road segment geometric feature data, thus defining the travel time from the vehicle's current location to the hospital. The estimated travel time is The overall priority of each hospital is determined by the urgency of the patient's condition, the availability of hospital resources, and the estimated travel time, which can be used to construct a comprehensive scoring function:

[0095] ;

[0096] in For the hospital Overall priority score;

[0097] For the system at reference time The severity index of the illness is calculated based on vital signs;

[0098] For the hospital The resource availability index at the current moment is calculated from information such as the availability ratio of emergency beds, the idle status of emergency equipment, and the waiting time for emergency triage at the hospital.

[0099] From the current location of the ambulance to the hospital The estimated travel time is calculated from real-time traffic flow speed and road geometry.

[0100] , , These are weighting coefficients determined through training with historical emergency medical data, used to balance the impact of patient condition, resources, and travel time on priority.

[0101] The system follows The values ​​are ranked from high to low to form a priority list of candidate hospitals.

[0102] The system constructs a resource locking request message based on the first hospital in the sorted list. The request message includes the patient identifier value, the required medical equipment value corresponding to the current criticality index, and the estimated arrival time value for that hospital. After the message is sent to the hospital information system, it enters a preset confirmation time window. Upon receiving the request, the hospital generates a resource locking confirmation code or rejection code according to its own resource allocation strategy and responds to the system. The system monitors the hospital's response within the preset time window. When a confirmation code is received, the hospital is designated as the target hospital. If a rejection code is received or no response is received within the preset time, the system automatically removes the hospital from the candidate list, selects the next hospital in the list, resends the resource locking request, and repeats the above process.

[0103] After successfully locking onto the target hospital, the system calculates the shortest travel time main navigation path based on real-time road network status data. The path calculation uses traffic flow velocity and road geometric feature values ​​from the state tensor as constraints. Assuming the path consists of continuous road segment nodes, the total travel cost of the path is defined by a cost function, with time as the sole evaluation metric. The expression for the total travel cost is:

[0104] ;

[0105] in For the cost of total travel time, For the first on the path The length of each road segment The system minimizes the traffic flow speed value of this road segment under real-time monitoring. The main navigation path is calculated and transmitted to the subsequent scheduling module and vehicle-side execution module. The generation of the main navigation path completes the entire processing flow.

[0106] S3. In response to the generation of the main navigation path, calculate the capacity coverage gap in the current area and send dispatch instructions to nearby idle vehicles;

[0107] Furthermore, calculating the current capacity coverage gap in the region specifically includes the following steps:

[0108] Based on historical emergency call data, extract the probability of emergency events occurring within the current time window for the grid where the current vehicle is located;

[0109] Calculate the backup response time for the nearest available emergency medical vehicle within the surrounding grid to reach the center of the grid.

[0110] Calculate the time delay increment of the backup response time relative to the original response time of the current vehicle;

[0111] The regional emergency response risk index is calculated based on the probability of emergency events and the time delay increment. If the regional emergency response risk index exceeds the preset threshold, it is determined that there is a gap in transportation capacity coverage.

[0112] Furthermore, sending dispatch instructions to nearby idle vehicles specifically includes the following steps:

[0113] Identify the target areas where there are capacity coverage gaps, and identify the directly adjacent and next-to-nearest jurisdictions surrounding the target areas;

[0114] Screen available candidate vehicles within the jurisdiction and calculate the secondary risk increment caused to the original jurisdiction by each candidate vehicle after it moves to the target filling area;

[0115] Based on the difference between the risk reduction and the secondary risk increment in the target filling area, the target vehicle-location scheduling pair is solved;

[0116] Based on vehicle-location scheduling, a dynamic redeployment command containing the target coordinates and suggested path is generated and sent.

[0117] Specifically, after generating the main navigation path, the system calculates the probability of emergency events occurring in the current area based on the vehicle's current location and historical emergency event statistics of the grid area it is in. The system records the set of grids after the city area is divided, and each grid contains a statistical sequence of the number of historical emergency events changing over time. Let's assume the current vehicle is located in grid... The current time window is Grid in historical data In the time window The average number of emergency incidents is denoted as The system uses a Poisson process model to describe the short-term occurrence of emergency events in the area and calculates the time window for the area. The probability of at least one emergency medical event occurring within the premises:

[0118] ;

[0119] in This represents the probability of an emergency response event occurring in the grid where the vehicle is currently located. This represents the average number of times an event occurs, based on historical statistics.

[0120] The system calculates the backup response time based on the location data of available emergency vehicles within each grid. It is assumed that an available vehicle is located in a grid... The vehicle arrived at the grid. The driving distance from the center point is The traffic flow speed of the section of road in question is provided by a real-time monitoring system and is denoted as [missing information]. The estimated response time for a replacement vehicle to move from its current location to the center of the grid is defined as:

[0121] ;

[0122] in For idle vehicles The replacement response time to the target grid center is recorded as the original response time of the currently executing emergency vehicle before it was redeployed. The system obtains the time delay increment by the difference between the replacement response time and the original response time. The time delay increment represents the additional response time required by the replacement solution. Its expression is:

[0123] ;

[0124] The system calculates a regional emergency response risk index using the probability of an emergency event and the time delay increment. A higher risk index indicates a higher risk that the area cannot be quickly covered by other vehicles at the current moment. The risk index is defined as follows:

[0125] ;

[0126] in For grid The system calculates the emergency response risk index and compares the risk index of the current area with a set threshold. When the risk index exceeds the threshold, it determines that there is a gap in capacity coverage.

[0127] Upon detecting a coverage gap, the system immediately generates a dispatch instruction. The dispatch instruction is a dynamic notification that the dispatch center needs to push to nearby available emergency vehicles. This notification includes the target grid coordinates, the optimal arrival route, and the time requirement for dispatch. The system selects the optimal available vehicle based on the principle of minimizing the backup response time and sends a dynamic dispatch instruction to the selected vehicle. The dispatch instruction contains the latitude and longitude values ​​of the target location coordinates and the standby strategy information to be executed after switching to the target area, so that the vehicle can immediately enter standby mode after arriving in the new area.

[0128] This allows for real-time assessment of the city's grid-based emergency coverage as emergency vehicles travel along their main navigation routes. When a decline in emergency response capabilities is detected in other areas, appropriate dispatch instructions are automatically sent to nearby available vehicles, thereby maintaining a dynamic balance of overall emergency resources and improving the overall response capability of the city's emergency medical system.

[0129] S4. While driving along the main navigation path, monitor the medical intervention action commands input by the vehicle terminal and analyze the stability constraint characteristics corresponding to the action.

[0130] Furthermore, monitoring the medical intervention commands input by the vehicle-mounted terminal specifically includes the following steps:

[0131] Collect voice control signals and touch screen operation signals from the medical cabin inside the vehicle;

[0132] Perform semantic recognition on voice control signals and extract action category labels;

[0133] Detect the trigger status of touchscreen buttons and generate standard operation code;

[0134] The action category labels and standard operation codes are merged to generate verified medical intervention action instructions.

[0135] Furthermore, analyzing the stability constraint features corresponding to this action specifically includes the following steps:

[0136] Retrieve from a pre-defined mapping library the standard operation duration, maximum permissible longitudinal acceleration threshold, and maximum permissible lateral acceleration threshold that match the verified medical intervention action command;

[0137] Calculate the dynamic stability driving window period by combining the current vehicle speed and road adhesion coefficient;

[0138] The longitudinal acceleration threshold, lateral acceleration threshold, and dynamic stability driving window period are combined into an instantaneous stability constraint feature vector.

[0139] Specifically, during the execution of the main navigation path, the system continuously receives voice signals and touch events input from the in-vehicle terminal. The voice signals are collected by the in-vehicle microphone to form a continuous digital voice sequence, which is synchronized with the system's master clock as timestamped voice data. Touch events are generated by the touchscreen, recording the start time, end time, button identifier, and press intensity. All data is consistent with the system's reference time sequence. Record.

[0140] The system first performs speech recognition on the speech signal, converts it into text information, and then performs semantic classification. The semantic classification model identifies the semantic intent within a predefined set of medical operation categories and outputs medical operation category labels, using symbols. This indicates that the reference time is... The system identifies the category of the recognized speech action and outputs the confidence level of that label, expressed as a symbol. This indicates that the reliability of the recognition results is represented by a mapping table, where touch data is converted into standard operation codes using symbols. This indicates that the reference time is... The corresponding touch operation code is used, and a fixed confidence level is assigned to the touch operation. Its value is determined based on the statistical analysis of the probability of accidental touches in human-computer interaction in the interface design.

[0141] The system performs consistent fusion of voice tags and touch codes within a short time window. If the voice tag is within that window... With touch code If a valid correspondence exists as specified in the mapping table, the standard action identifier corresponding to the operation can be directly retrieved using the symbol. This means that if a conflict arises between the two, the system makes a decision based on the confidence level, and the decision function is:

[0142] ;

[0143] in The finalized medical intervention instructions;

[0144] These are the standard action identifiers corresponding to voice tags;

[0145] The standard action identifier corresponding to the touch code;

[0146] For speech recognition confidence;

[0147] To ensure the confidence level of touch operation, this solution guarantees that operation commands can still be stably received even in noisy in-vehicle environments.

[0148] Receive medical action instructions Next, the system enters the stability constraint feature analysis stage. For this purpose, the system pre-constructs a mapping library from medical actions to vehicle stability parameters. During the construction of the mapping library, the emergency vehicle is equipped with a medical cabin in a closed test range. Multiple medical personnel perform various medical operations multiple times under different vehicle speeds and acceleration conditions, and the longitudinal acceleration sequence of the vehicle is collected simultaneously. With lateral acceleration sequence subscript Number the sampling points and record the start time of each medical procedure. and end time subscript Each operation was assigned a number, and clinical experts evaluated the stability of each operation, recording whether it was a stable operation. All experimental samples that were determined to be stable constituted a stable operation sample set.

[0149] For a certain medical action symbol The system will perform statistical operations on a stable sample set for a duration of time, including the first... The duration of this operation is denoted as Let there be a total of stable samples. If there are 1, then the standard operation duration is defined as:

[0150] ;

[0151] in For medical actions The standard operating duration, in seconds.

[0152] The system's absolute longitudinal acceleration sequence for all stable samples ( To stabilize the sample, the samples (the total number of all sampling points in the sample) are sorted in ascending order, denoted as:

[0153] ;

[0154] Select the first Using each element as the 95th percentile, the maximum permissible longitudinal acceleration threshold is defined as:

[0155] ;

[0156] Similarly, sorting the sequence of absolute values ​​of lateral acceleration yields:

[0157] ;

[0158] These two values ​​represent the maximum permissible acceleration thresholds in the longitudinal and lateral directions, respectively, in m / s².

[0159] During the test, the vehicle maintained a reference speed, which was recorded as... Unit: m / s. The test road surface was a dry, high-friction surface. The adhesion coefficient was obtained through calibration testing and is denoted as [missing information]. The above parameters and action identifiers These parameters are stored together to form the basic set of stability parameters for this action.

[0160] In actual operation, the system reads the vehicle's current speed and uses a symbol... This indicates that the current road surface adhesion coefficient, estimated by the vehicle stability control system, is read and represented by the symbol... This indicates that the threshold conversion from reference conditions to the current operating condition is performed using linear scaling to obtain the allowable longitudinal acceleration value under the current operating condition:

[0161] ;

[0162] Permissible lateral acceleration:

[0163] ;

[0164] Both of these are acceleration constraints under the current road adhesion conditions, with units of m / s².

[0165] The system defines a stable operating window based on the standard operation duration. If the medical operation occurs at a certain time... If identified, the window period ends at... During this window period, vehicles must meet the aforementioned acceleration thresholds.

[0166] The system combines the allowable longitudinal acceleration, allowable lateral acceleration, and operation duration to form a stability constraint feature vector:

[0167] ;

[0168] in The three components represent the upper limit of longitudinal acceleration, the upper limit of lateral acceleration, and the expected duration of medical operation under the current road attachment conditions. This feature vector serves as the input to the subsequent trajectory replanning module and becomes the core constraint limiting the change in vehicle acceleration.

[0169] This allows the specific operational needs of medical personnel to be mapped into quantifiable constraints at the vehicle dynamics level, enabling the vehicle to maintain a safe and stable acceleration range during medical operations. This provides a structured stability constraint basis for subsequent local trajectory replanning, thereby ensuring the smooth conduct of medical operations and maintaining the stability of vehicle driving.

[0170] S5. Based on stability constraint features and road segment features ahead, perform local driving trajectory replanning on the main navigation path, generate the execution path, and output driving prompts.

[0171] Furthermore, the local replanning of the driving trajectory on the main navigation path specifically includes the following steps:

[0172] Extract the number of lanes, radius of curvature, and lane smoothness index within a preset distance ahead of the main navigation path;

[0173] Generate candidate driving trajectories that include lane keeping, lane changing, and deceleration trajectories;

[0174] Simulate the vehicle's body posture as it travels along a candidate driving trajectory and predict the expected acceleration sequence;

[0175] The expected acceleration sequence is compared with the instantaneous stability constraint feature vector, and the trajectory that meets the constraints and has the least time consumption is selected as the target driving path.

[0176] Furthermore, generating the execution path and outputting driving prompts specifically includes the following steps:

[0177] The target driving path is analyzed into a speed control profile and a steering guidance sequence;

[0178] A visual driving corridor covering the target driving path is projected in front of the field of vision using in-vehicle display equipment;

[0179] Play voice commands to keep the vehicle in the driving corridor and a countdown timer for medical procedures;

[0180] The system collects actual vehicle acceleration feedback. If the indicator is detected to exceed the constraint characteristics, an audible and visual alarm is triggered.

[0181] Specifically, after obtaining the stability constraint feature vector corresponding to the medical operation, the system extracts road structure information from the main navigation path within a certain distance range from the current position to the front. The road structure information is jointly provided by high-precision maps and traffic data, including the number of lanes in each road segment ahead. Road section curvature radius and lane smoothness index ,in The cumulative spatial distance along the road centerline, in meters. The system discretizes all road structure parameters within this spatial segment to obtain equally spaced sampling points so that subsequent trajectory generation can be performed at a fixed spatial step size.

[0182] Within this spatial segment, the system generates multiple candidate driving trajectories. Each trajectory defines three basic lateral actions at each discrete position, such as lane keeping, left-side merging, and right-side merging, as well as various longitudinal deceleration control strategies, forming a combined trajectory of lateral actions and longitudinal speed control. The control variable corresponding to the discrete point of the trajectory is the longitudinal speed. and lateral offset ,in For the first A discrete spatial location.

[0183] The system uses a vehicle dynamics model to calculate the expected acceleration sequence for each candidate trajectory, with the longitudinal acceleration expressed as a discrete velocity difference.

[0184] ;

[0185] Lateral acceleration is based on the relationship of turning motion:

[0186] ;

[0187] in The distance between adjacent discrete positions. is the radius of curvature at the corresponding position.

[0188] Subsequently, the system compares the expected acceleration sequence node-by-node with the previously obtained stability constraint feature vector, which contains the upper limit of the allowed longitudinal acceleration. Permissible upper limit of lateral acceleration and duration of medical procedures Within the time window corresponding to the operation duration, the system maps it to a spatial range and checks each trajectory node:

[0189] ;

[0190] If any node violates the above constraints, the trajectory is removed.

[0191] For the remaining candidate trajectories that satisfy the acceleration constraints, the system calculates their expected travel time:

[0192] ;

[0193] in For trajectory The total travel time, in seconds. The system selects the trajectory with the shortest travel time from all trajectories that meet the constraints as the target trajectory. It is then converted into a path structure that can be executed by the vehicle, including a sequence of node spatial positions and a corresponding desired speed control sequence, for subsequent generation of driving instructions.

[0194] Under the premise of meeting the stability requirements of specific medical operations, the driving trajectory that meets both acceleration safety constraints and has the shortest travel time is selected from multiple candidate trajectories that conform to road structure constraints. This enables the emergency vehicle to maintain a stable driving state during the medical operation and minimize the overall travel time, thereby improving the safety and efficiency of the transfer process.

[0195] Example 2:

[0196] Insufficient comprehensive utilization of patient condition, hospital resources, and vehicle stability during route planning and scheduling can easily lead to decision-making delays or medical procedures being affected by travel delays during transport. To address these issues, this invention provides an AI-based intelligent recommendation system for emergency transport routes, the structure of which is as follows: Figure 2 As shown, the system includes:

[0197] The data acquisition module is used to collect patient vital signs data, vehicle operation status data, and traffic condition data, and to clean and align them in time and space to build an emergency transport status dataset.

[0198] The resource locking module is used to assess the severity of the patient's condition and match hospital resources based on the emergency transport status dataset, and to lock the target hospital and generate the main navigation path through a two-way resource interaction protocol.

[0199] The capacity scheduling module is used to calculate the capacity coverage gap in the current area and send scheduling instructions to nearby idle vehicles;

[0200] The motion monitoring module is used to monitor medical intervention motion commands input by the vehicle terminal and analyze the stability constraint characteristics corresponding to the motion.

[0201] The micro-execution module is used to perform local driving trajectory replanning on the main navigation path based on the stability constraint features and the features of the road segment ahead, generate the execution path, and output driving prompts.

[0202] Specifically, the data acquisition module continuously acquires patient vital signs, vehicle acceleration and angular velocity, and road condition information such as traffic speed and geometric features. It then performs cleaning and spatiotemporal alignment through filtering, interpolation, and unified timestamps to construct a multi-dimensional emergency transport status dataset. The resource locking module calculates the severity of the patient's condition based on this dataset and generates a candidate hospital ranking by combining hospital bed occupancy rates, equipment availability, and triage queue times. It then locks in the target hospital's resources through a two-way resource interaction protocol and calculates the main navigation path accordingly. The capacity scheduling module uses historical emergency event probabilities and backup vehicles... The vehicle response time is calculated to identify gaps in regional capacity coverage. When a gap appears, a dynamic dispatch command is sent to the most suitable nearby available vehicle. The action monitoring module generates verified medical intervention action commands through voice recognition and touch parsing, and queries the mapping library to obtain stability constraint features consisting of operation duration and allowable acceleration threshold. The micro-execution module generates multiple candidate trajectories based on these stability constraint features and the curvature and lane structure of the road ahead, calculates their expected acceleration, and selects the target driving trajectory that meets the acceleration limit and has the shortest time. Finally, the execution path including speed prompts and a visualized driving corridor is output to assist driving.

[0203] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent recommendation method for emergency transport routes based on AI technology, characterized in that, Includes the following steps: S1. Collect patient vital signs data, vehicle operation status data, and traffic condition data, clean and align them in time and space, and construct an emergency transport status dataset. S2. Based on the emergency transport status dataset, assess the severity of the patient's condition and match hospital resources. Locate the target hospital through a two-way resource interaction protocol and generate the main navigation path. S3. In response to the generation of the main navigation path, calculate the capacity coverage gap in the current area and send dispatch instructions to nearby idle vehicles; S4. During the driving along the main navigation path, monitor the medical intervention action commands input by the vehicle terminal and analyze the instantaneous stability constraint characteristics corresponding to the action. S5. Based on the instantaneous stability constraint features and the features of the road segment ahead, perform local driving trajectory replanning on the main navigation path to generate a target driving path, and generate an execution path based on the target driving path and output driving prompts. The medical intervention action commands input by the monitoring vehicle-mounted terminal specifically include the following steps: Collect voice control signals and touch screen operation signals from the medical cabin inside the vehicle; Perform semantic recognition on the voice control signal and extract action category labels; Detect the trigger status of touchscreen buttons and generate standard operation code; The action category label and the standard operation code are fused to generate a verified medical intervention action instruction; The process of analyzing the instantaneous stability constraint features corresponding to this action specifically includes the following steps: Retrieve from a pre-set mapping library the standard operation duration, maximum permissible longitudinal acceleration threshold, and maximum permissible lateral acceleration threshold that match the verified medical intervention action command; Calculate the dynamic stability driving window period by combining the current vehicle speed and road adhesion coefficient; The longitudinal acceleration threshold, the lateral acceleration threshold, and the dynamic stability driving window period are combined into an instantaneous stability constraint feature vector. The process of performing local replanning of the main navigation path specifically includes the following steps: Extract the number of lanes, radius of curvature, and lane smoothness index within a preset distance ahead of the main navigation path; Generate candidate driving trajectories that include lane keeping, lane changing, and deceleration trajectories; Simulate the vehicle's body posture as it travels along a candidate driving trajectory and predict the expected acceleration sequence; The expected acceleration sequence is compared with the instantaneous stability constraint feature vector, and the trajectory that meets the constraints and has the shortest time consumption is selected as the target driving path.

2. The intelligent recommendation method for emergency transport routes based on AI technology according to claim 1, characterized in that, The construction of the emergency transport status dataset specifically includes the following steps: Acquire patients' electrocardiogram, blood pressure and blood oxygen data, acquire vehicle's three-axis acceleration and angular velocity data, and acquire road network traffic flow speed and road surface geometric feature data; Poll the information systems of candidate hospitals within the target area to obtain data on emergency bed occupancy rate, equipment availability, and emergency triage queue time. The collected data is filtered and interpolated. The data is synchronized and aligned with the system time, and the road condition data is mapped with the vehicle location to generate a multi-dimensional state tensor.

3. The intelligent recommendation method for emergency transport routes based on AI technology according to claim 1, characterized in that, The process of locating the target hospital through a two-way resource interaction protocol includes the following steps: Based on the assessment results of the severity of the illness, the characteristics of medical resources, and the estimated travel time, a priority list of candidate hospitals is generated. Send an exclusive resource lock request containing the patient ID, equipment requirements, and estimated arrival time to the first hospital in the priority list; If a resource lock confirmation code is received from the first hospital within the preset window, then that hospital is established as the transfer destination. If a rejection code or timeout is received within the preset window, the current hospital is removed from the list, and the next candidate hospital is selected to repeat the step of sending an exclusive resource lock request.

4. The intelligent recommendation method for emergency transport routes based on AI technology according to claim 1, characterized in that, The calculation of the current transportation capacity coverage gap in the region specifically includes the following steps: Based on historical emergency call data, extract the probability of emergency events occurring within the current time window for the grid where the current vehicle is located; Calculate the backup response time for the nearest available emergency medical vehicle within the surrounding grid to reach the center of the grid. Calculate the time delay increment of the substitute response time relative to the original response time of the current vehicle; The regional emergency response risk index is calculated based on the probability of the emergency event and the time delay increment. If the index exceeds a preset threshold, it is determined that there is a capacity coverage gap.

5. The intelligent recommendation method for emergency transport routes based on AI technology according to claim 1, characterized in that, Sending dispatch instructions to nearby idle vehicles specifically includes the following steps: Identify target areas where there are capacity coverage gaps, and identify the directly adjacent and next-to-nearest jurisdictions surrounding these areas; Select available candidate dispatch vehicles within the jurisdiction and calculate the secondary risk increment caused to the original jurisdiction by each candidate dispatch vehicle after it moves to the target filling area. The target vehicle-location scheduling pair is solved based on the difference between the risk reduction in the target filling area and the secondary risk increment. Based on the vehicle-location scheduling, a dynamic redeployment instruction containing the target coordinates and suggested path is generated and sent.

6. The intelligent recommendation method for emergency transport routes based on AI technology according to claim 1, characterized in that, The process of generating the execution path and outputting driving prompts specifically includes the following steps: The target driving path is analyzed into a speed control profile and a steering guidance sequence; A visual driving corridor covering the target driving path is projected in front of the field of vision using an in-vehicle display device; Play voice commands to keep the vehicle in the driving corridor and a countdown timer for medical procedures; The system collects actual vehicle acceleration feedback. If the indicator is detected to exceed the constraint characteristics, an audible and visual alarm is triggered.

7. An AI-based intelligent recommendation system for emergency transport routes, characterized in that: The system for the AI-based intelligent recommendation method for emergency transport routes according to any one of claims 1-6, the system comprising: The data acquisition module is used to collect patient vital signs data, vehicle operation status data, and traffic condition data, and to clean and align them in time and space to build an emergency transport status dataset. The resource locking module is used to assess the severity of the patient's condition and match hospital resources based on the emergency transport status dataset, and to lock the target hospital and generate the main navigation path through a two-way resource interaction protocol. The capacity scheduling module is used to calculate the capacity coverage gap in the current area and send scheduling instructions to nearby idle vehicles; The motion monitoring module is used to monitor medical intervention motion commands input by the vehicle terminal and analyze the stability constraint characteristics corresponding to the motion. The micro-execution module is used to perform local driving trajectory replanning on the main navigation path based on the stability constraint features and the features of the road segment ahead, generate the execution path, and output driving prompts.

Images