Method and system for determining man-machine co-driving takeover request time of unmanned agricultural machine

Abstract

The invention discloses a method and system for determining man-machine co-driving takeover request time of an unmanned agricultural machine, and belongs to the technical field of intelligent agricultural machines and automatic driving. The method aims at solving the problem that in man-machine co-driving of the unmanned agricultural machine, an individual difference cannot be considered in a method for determining the take-over request time. The method comprises the steps that different original action sequences of drivers are collected, and the situation response delay time of each driver is collected; processing the collected data; reaction time after driver state correction is calculated, original action sequence complexity is obtained through calculation based on mapping of an original action sequence and environment characteristics, and action coupling correction time is obtained through calculation; calculating a risk advance based on the environmental risk index; calculating a safety advance based on the safety time; and calculating system request time based on the response time after driver state correction, the action coupling correction time, the risk advance and the safety advance, and calculating takeover request time in combination with API call response delay. The method is used for calculating the man-machine co-driving take-over request time.

Description

Technical Field

[0001] This invention relates to a method and system for determining the time of human-machine takeover request in unmanned agricultural machinery, belonging to the field of intelligent agricultural machinery and autonomous driving technology. Background Technology

[0002] In the human-machine co-driving scenario of unmanned agricultural machinery, existing takeover requests mostly rely on fixed thresholds or single signals, which makes it difficult to take into account individual differences of drivers, complexity of action sequences and uncertainty of the working environment. This can easily lead to problems of takeover being too early or too late, affecting safety and efficiency. Summary of the Invention

[0003] To address the issue that the method for determining the takeover request time in human-machine co-driving of unmanned agricultural machinery cannot take into account individual differences, this invention provides a method and system for determining the takeover request time in human-machine co-driving of unmanned agricultural machinery.

[0004] The present invention provides a method for determining the takeover request time for unmanned agricultural machinery, comprising:

[0005] Collect different original action sequences of the drivers, as well as the situational reaction delay time of each driver;

[0006] Each action in the original action sequence is standardized to obtain a standardized action sequence; the duration of the standardized action sequence is labeled, and the feature vector of each standardized action is recorded in combination with the contextual response delay time.

[0007] Establish a mapping between each original action sequence and environmental features, and establish a mapping between the feature vector of each standardized action and driver attribute features;

[0008] The standard deviation of the driver's steering wheel offset is calculated based on the feature vector and the mapping between the feature vector and the driver's attribute features. Combined with the fatigue correction factor of the original action sequence, the driver's reaction time after state correction is calculated.

[0009] The complexity of the original action sequence is calculated based on the mapping between each original action sequence and environmental features, and the action coupling correction time is also calculated.

[0010] Calculate risk lead time based on environmental risk indicators;

[0011] Calculate safety lead time based on safety time;

[0012] The system request time is calculated based on the driver's state correction reaction time, action coupling correction time, risk lead time, and safety lead time. The takeover request time is then calculated by combining the API call response delay.

[0013] According to the method for determining the takeover request time of unmanned agricultural machinery co-driving according to the present invention, I different original action sequences of the driver are collected:

[0014] ,

[0015] In the formula For the first A sequence of original actions, , For the first The j-th original action in a sequence of original actions; For action functions, Let J be the sensor signal intensity vector corresponding to the j-th original action. This is the set of environment variables corresponding to the j-th original action. This is the timestamp corresponding to the j-th original action; the original actions include pressing the accelerator, steering, straightening, observing the surroundings, changing gears, pressing the brake, and operating the working device;

[0016] The situational reaction delay time of the kth driver is expressed as: , , Total number of drivers;

[0017] right The method for standardization is as follows:

[0018] ,

[0019] In the formula For the first The j-th standardized action in a sequence of standardized actions Let j be the mean of the original action. Let be the standard deviation of the j-th original action.

[0020] According to the method for determining the takeover request time of unmanned agricultural machinery co-driving according to the present invention, the duration of standardized action sequences is marked. :

[0021] ,

[0022] In the formula Let i be the duration of the i-th original action sequence. The end action label in the i-th original action sequence;

[0023] The behavioral feature vector of the j-th standardized action is represented as: :

[0024] ,

[0025] In the formula Let j be the frequency of the original action. Let j be the duration of the original action. For the j-th original action response delay, For the j-th original action operation range, Let j be the action coupling feature of the j-th original action.

[0026] According to the method for determining the takeover request time for unmanned agricultural machinery co-driving according to the present invention, a mapping between each original action sequence and environmental features is established:

[0027] ,

[0028] In the formula The r-th environmental feature includes soil condition, weather conditions, terrain, operational process, and crop condition.

[0029] Establish a mapping between the feature vector of each standardized action and the driver's attribute features:

[0030] ,

[0031] In the formula These are driver attributes, which include age, gender, experience, and reaction characteristics.

[0032] According to the method for determining the takeover request time for unmanned agricultural machinery co-driving according to the present invention, the method for calculating the standard deviation of the driver's steering wheel offset is as follows:

[0033] ,

[0034] In the formula The standard deviation of the driver's steering wheel offset. Let be the function representing the influence of driver attributes and behavioral characteristics on driver stability. The instantaneous direction angle, The average direction angle;

[0035] The method for calculating the driver's reaction time after state correction is as follows:

[0036] ,

[0037] In the formula The reaction time after driver status correction. The average reaction time, For the reaction time safety factor, The standard deviation of the reaction time. This is the operational stability coefficient. It is a fatigue correction factor;

[0038] ,

[0039] In the formula This is an empirical coefficient. It is an index of fatigue strength.

[0040] The method for determining the takeover request time for unmanned agricultural machinery according to the present invention includes the following method for calculating the action coupling correction time:

[0041] ,

[0042] In the formula This represents the number of original actions in the original action sequence. For motion coupling correction time, Let j be the weight of the original action. The complexity of the original action sequence;

[0043] ,

[0044] In the formula For the first The synchronization coupling coefficient of the j-th original action in a sequence of original actions. This is the function that describes the influence of environmental characteristics on the degree of complexity.

[0045] The method for determining the takeover request time for unmanned agricultural machinery according to the present invention, and the method for calculating the risk lead time based on environmental risk indicators, are as follows:

[0046] ,

[0047] In the formula Here, A represents the number of environmental risk indicators. Risk weighting These are environmental indicators; environmental indicators include obstacle distance, humidity, slope, and visibility.

[0048] ,

[0049] In the formula To anticipate risks, Risk sensitivity coefficient;

[0050] The method for calculating the safety lead time is as follows:

[0051] ,

[0052] In the formula To allow for a safe lead time, For safety amplification factor, Minimum safe operating time.

[0053] According to the method for determining the takeover request time for unmanned agricultural machinery co-driving according to the present invention, the method for calculating the system request time is as follows:

[0054] ,

[0055] In the formula For system request time, This is the coupling amplification factor. For environmental emergencies correction factors, For environmental feature weights, These are environmental characteristic factors.

[0056] According to the method for determining the takeover request time for unmanned agricultural machinery co-driving according to the present invention, the method for calculating the takeover request time is as follows:

[0057] ,

[0058] In the formula For the time of the takeover request, For API call response delay, This is to compensate for the time lost due to database updates;

[0059] ,

[0060] In the formula For index retrieval time, For data extraction time, Calculate the time taken for takeover.

[0061] This invention also provides a system for determining the takeover request time for unmanned agricultural machinery, comprising:

[0062] Data acquisition module: used to collect different raw action sequences of the driver, as well as the situational reaction delay time of each driver;

[0063] The data processing module is used to standardize each action in the original action sequence to obtain a standardized action sequence; it also labels the duration of the standardized action sequence and records the feature vector of each standardized action in combination with the contextual response delay time.

[0064] The reaction characteristics module is used to establish a mapping between each original action sequence and environmental features, and to establish a mapping between the feature vector of each standardized action and driver attribute features;

[0065] The time calculation module is used to calculate the standard deviation of the driver's steering wheel offset based on the feature vector and the mapping between the feature vector and the driver's attribute features. Combined with the fatigue correction factor of the original action sequence, the driver's reaction time after state correction is calculated.

[0066] The complexity of the original action sequence is calculated based on the mapping between each original action sequence and environmental features, and the action coupling correction time is also calculated.

[0067] Calculate risk lead time based on environmental risk indicators;

[0068] Calculate safety lead time based on safety time;

[0069] The system request time is calculated based on the driver's state correction reaction time, action coupling correction time, risk lead time, and safety lead time. The takeover request time is then calculated by combining the API call response delay.

[0070] The beneficial effects of this invention are as follows: This invention covers driver action sequence modeling, sensor signal standardization and time series labeling, behavioral feature extraction, environmental risk assessment, and multi-factor fusion to calculate takeover time.

[0071] To improve the accuracy of takeover time determination, this invention starts from multi-source data: during the operation, an action sequence matrix is ​​constructed that includes action units such as throttle, steering, straightening, observation, gear shifting, braking, and operation of the work device. The original signals are standardized and time-series labeled to form a unified behavioral input. Furthermore, behavioral feature vectors such as operation frequency, duration, reaction delay, operation amplitude, and action coupling are extracted, and a data and knowledge storage / retrieval structure is constructed using scene-action sequence mapping, driver-feature index, and risk level mapping function. At the same time, the individual state is quantified by combining driver stability index and fatigue correction factor, and environmental risk indicators such as obstacle distance, humidity, slope, and visibility are combined with weights to transform the coupling relationship of "human-machine-environment" into a calculable basis for takeover time.

[0072] This invention performs individualized correction of the average reaction time based on driver stability index and fatigue correction factor, calculates coupling correction time according to action sequence complexity and action coupling characteristics, and introduces multi-scenario risk advance and safety advance to achieve risk feedforward control. Finally, it gives the takeover request time, so that the takeover timing can be adaptively adjusted according to the dynamic changes of people, machines and environment, thereby reducing false triggering and missed triggering from the mechanism.

[0073] This invention further provides an online learning update and interface-level latency compensation mechanism: by incrementally updating the feature vectors of newly added samples, the model can continuously adapt to seasonal, plot, and population differences; on the engineering deployment side, the latency of API calls, index retrieval, data extraction and calculation are simultaneously taken into account, and the real-time dynamic takeover time is output to ensure that it has an interpretable, implementable and scalable engineering effect in actual operation. Attached Figure Description

[0074] Figure 1This is a flowchart of the method for determining the takeover request time for unmanned agricultural machinery as described in this invention;

[0075] Figure 2 This is a structural block diagram of the system for determining the takeover request time for unmanned agricultural machinery as described in this invention. Detailed Implementation

[0076] The technical solutions of 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.

[0077] Specific Implementation Method 1: Combination Figure 1 As shown, the first aspect of the present invention provides a method for determining the time of human-machine takeover request for unmanned agricultural machinery, including,

[0078] Collect different original action sequences of the drivers, as well as the situational reaction delay time of each driver;

[0079] Each action in the original action sequence is standardized to obtain a standardized action sequence; the duration of the standardized action sequence is labeled, and the feature vector of each standardized action is recorded in combination with the contextual response delay time.

[0080] Establish a mapping between each original action sequence and environmental features, and establish a mapping between the feature vector of each standardized action and driver attribute features;

[0081] The standard deviation of the driver's steering wheel offset is calculated based on the feature vector and the mapping between the feature vector and the driver's attribute features. Combined with the fatigue correction factor of the original action sequence, the driver's reaction time after state correction is calculated.

[0082] The complexity of the original action sequence is calculated based on the mapping between each original action sequence and environmental features, and the action coupling correction time is also calculated.

[0083] Calculate risk lead time based on environmental risk indicators;

[0084] Calculate safety lead time based on safety time;

[0085] The system request time is calculated based on the driver's state correction reaction time, action coupling correction time, risk lead time, and safety lead time. The takeover request time is then calculated by combining the API call response delay.

[0086] The main goal of the data acquisition phase is to build an action sequence matrix for subsequent time analysis and feature extraction.

[0087] Furthermore, I different original action sequences of the driver were collected:

[0088] ,

[0089] In the formula For the first A sequence of original actions, , For the first The j-th original action in a sequence of original actions; For action functions, Let J be the sensor signal intensity vector corresponding to the j-th original action. This is the set of environment variables corresponding to the j-th original action. This is the timestamp corresponding to the j-th original action; the original actions include pressing the accelerator, steering, straightening, observing the surroundings, changing gears, pressing the brake, and operating the working device;

[0090] The original action sequence consists of combinations of different actions as needed.

[0091] The situational reaction delay time of the kth driver is expressed as: , , Total number of drivers;

[0092] right The method for standardization is as follows:

[0093] ,

[0094] In the formula For the first The j-th standardized action in a sequence of standardized actions Let j be the mean of the original action. Let be the standard deviation of the j-th original action.

[0095] The set of environmental variables includes slope, humidity, and crop height.

[0096] The original acquired signals are standardized and feature extracted in order to obtain a unified input vector.

[0097] To accurately depict the driver's operational actions, the original signal is time-series labeled.

[0098] Duration of standardized action sequences :

[0099] ,

[0100] In the formula The duration of the i-th original action sequence is used to characterize the location of the driver's operation on the time axis; This is the ending action label in the i-th original action sequence; it indicates the stage of operation the driver is in at that moment.

[0101] The behavioral feature vector of the j-th standardized action is represented as: :

[0102] ,

[0103] In the formula Let j be the frequency of the original action. Let j be the duration of the original action. For the j-th original action response delay, For the j-th original action operation range, Let j be the action coupling feature of the j-th original action.

[0104] In practical use, the logical structure of the model database is centered on index mapping and scene association.

[0105] Establish a mapping between each original action sequence and environmental features:

[0106] ,

[0107] In the formula The r-th environmental feature includes soil condition, weather conditions, terrain, operational process, and crop condition.

[0108] Establish a mapping between the feature vector of each standardized action and the driver's attribute features:

[0109] ,

[0110] In the formula These are driver attributes, which include age, gender, experience, and reaction characteristics.

[0111] The method for calculating the standard deviation of driver's steering wheel offset is as follows:

[0112] ,

[0113] In the formula The standard deviation of the driver's steering wheel offset. Let be the function representing the influence of driver attributes and behavioral characteristics on driver stability. The instantaneous direction angle, The average direction angle;

[0114] The method for calculating the driver's reaction time after state correction is as follows:

[0115] ,

[0116] In the formula The reaction time after driver status correction. The average reaction time, For the reaction time safety factor, The standard deviation of the reaction time. This is the operational stability coefficient. It is a fatigue correction factor;

[0117] ,

[0118] In the formula This is an empirical coefficient. It is an index of fatigue strength.

[0119] Furthermore, the method for calculating the action coupling correction time is as follows:

[0120] ,

[0121] In the formula This represents the number of original actions in the original action sequence. For motion coupling correction time, Let j be the weight of the original action. The complexity of the original action sequence;

[0122] A mapping is established between operation sequences and reaction delay characteristics in different scenarios.

[0123] ,

[0124] In the formula For the first The synchronization coupling coefficient of the j-th original action in a sequence of original actions. This is the function that describes the influence of environmental characteristics on the degree of complexity.

[0125] The following calculation of the takeover request time is based on a combination of characteristics and risk indicators:

[0126] The method for calculating risk lead time based on environmental risk indicators is as follows:

[0127] ,

[0128] In the formula Here, A represents the number of environmental risk indicators. Risk weighting These are environmental indicators; environmental indicators include obstacle distance, humidity, slope, and visibility.

[0129] ,

[0130] In the formula To anticipate risks, Risk sensitivity coefficient;

[0131] The method for calculating the safety lead time is as follows:

[0132] ,

[0133] In the formula To allow for a safe lead time, For safety amplification factor, Minimum safe operating time.

[0134] The method for calculating system request time is as follows:

[0135] ,

[0136] In the formula For system request time, This is the coupling amplification factor. This is a correction factor for sudden environmental events (such as sensor malfunctions). For environmental feature weights, These are environmental characteristic factors.

[0137] The method for calculating the takeover request time is as follows:

[0138] ,

[0139] In the formula For the time of the takeover request, For API call response delay, This is to compensate for the time lost due to database updates;

[0140] ,

[0141] In the formula For index retrieval time, For data extraction time, Calculate the time taken for takeover.

[0142] The database expansion and update mechanism in this implementation method is as follows:

[0143] Online learning update plan:

[0144] ,

[0145] in This is the feature vector of the current model. For learning rate, This is the newly acquired feature vector.

[0146] Specific Implementation Method Two: Combination Figure 2 As shown, another aspect of the present invention provides a system for determining the takeover request time for unmanned agricultural machinery, comprising,

[0147] Data acquisition module: used to collect different raw action sequences of the driver, as well as the situational reaction delay time of each driver;

[0148] The data processing module is used to standardize each action in the original action sequence to obtain a standardized action sequence; it also labels the duration of the standardized action sequence and records the feature vector of each standardized action in combination with the contextual response delay time.

[0149] The reaction characteristics module is used to establish a mapping between each original action sequence and environmental features, and to establish a mapping between the feature vector of each standardized action and driver attribute features;

[0150] The time calculation module is used to calculate the standard deviation of the driver's steering wheel offset based on the feature vector and the mapping between the feature vector and the driver's attribute features. Combined with the fatigue correction factor of the original action sequence, the driver's reaction time after state correction is calculated.

[0151] The complexity of the original action sequence is calculated based on the mapping between each original action sequence and environmental features, and the action coupling correction time is also calculated.

[0152] Calculate risk lead time based on environmental risk indicators;

[0153] Calculate safety lead time based on safety time;

[0154] The system request time is calculated based on the driver's state correction reaction time, action coupling correction time, risk lead time, and safety lead time. The takeover request time is then calculated by combining the API call response delay.

[0155] The main goal of the data acquisition phase is to build an action sequence matrix for subsequent time analysis and feature extraction.

[0156] Furthermore, I different original action sequences of the driver were collected:

[0157] ,

[0158] In the formula For the first A sequence of original actions, , For the first The j-th original action in a sequence of original actions; For action functions, Let J be the sensor signal intensity vector corresponding to the j-th original action. This is the set of environment variables corresponding to the j-th original action. This is the timestamp corresponding to the j-th original action; the original actions include pressing the accelerator, steering, straightening, observing the surroundings, changing gears, pressing the brake, and operating the working device;

[0159] The original action sequence consists of combinations of different actions as needed.

[0160] The situational reaction delay time of the kth driver is expressed as: , , Total number of drivers;

[0161] right The method for standardization is as follows:

[0162] ,

[0163] In the formula For the first The j-th standardized action in a sequence of standardized actions Let j be the mean of the original action. Let be the standard deviation of the j-th original action.

[0164] The set of environmental variables includes slope, humidity, and crop height.

[0165] The original acquired signals are standardized and feature extracted in order to obtain a unified input vector.

[0166] To accurately depict the driver's operational actions, the original signal is time-series labeled.

[0167] Duration of standardized action sequences :

[0168] ,

[0169] In the formula The duration of the i-th original action sequence is used to characterize the location of the driver's operation on the time axis; This is the ending action label in the i-th original action sequence; it indicates the stage of operation the driver is in at that moment.

[0170] The behavioral feature vector of the j-th standardized action is represented as: :

[0171] ,

[0172] In the formula Let j be the frequency of the original action. Let j be the duration of the original action. For the j-th original action response delay, For the j-th original action operation range, Let j be the action coupling feature of the j-th original action.

[0173] In practical use, the logical structure of the model database is centered on index mapping and scene association.

[0174] Establish a mapping between each original action sequence and environmental features:

[0175] ,

[0176] In the formula The r-th environmental feature includes soil condition, weather conditions, terrain, operational process, and crop condition.

[0177] Establish a mapping between the feature vector of each standardized action and the driver's attribute features:

[0178] ,

[0179] In the formula These are driver attributes, which include age, gender, experience, and reaction characteristics.

[0180] The method for calculating the standard deviation of driver's steering wheel offset is as follows:

[0181] ,

[0182] In the formula The standard deviation of the driver's steering wheel offset. Let be the function representing the influence of driver attributes and behavioral characteristics on driver stability. The instantaneous direction angle, The average direction angle;

[0183] The method for calculating the driver's reaction time after state correction is as follows:

[0184] ,

[0185] In the formula The reaction time after driver status correction. The average reaction time, For the reaction time safety factor, The standard deviation of the reaction time. This is the operational stability coefficient. It is a fatigue correction factor;

[0186] ,

[0187] In the formula This is an empirical coefficient. It is an index of fatigue strength.

[0188] Furthermore, the method for calculating the action coupling correction time is as follows:

[0189] ,

[0190] In the formula This represents the number of original actions in the original action sequence. For motion coupling correction time, Let j be the weight of the original action. The complexity of the original action sequence;

[0191] A mapping is established between operation sequences and reaction delay characteristics in different scenarios.

[0192] ,

[0193] In the formula For the first The synchronization coupling coefficient of the j-th original action in a sequence of original actions. This is the function that describes the influence of environmental characteristics on the degree of complexity.

[0194] The following calculation of the takeover request time is based on a combination of characteristics and risk indicators:

[0195] The method for calculating risk lead time based on environmental risk indicators is as follows:

[0196] ,

[0197] In the formula Here, A represents the number of environmental risk indicators. Risk weighting These are environmental indicators; environmental indicators include obstacle distance, humidity, slope, and visibility.

[0198] ,

[0199] In the formula To anticipate risks, Risk sensitivity coefficient;

[0200] The method for calculating the safety lead time is as follows:

[0201] ,

[0202] In the formula To allow for a safe lead time, For safety amplification factor, Minimum safe operating time.

[0203] The method for calculating system request time is as follows:

[0204] ,

[0205] In the formula For system request time, This is the coupling amplification factor. This is a correction factor for sudden environmental events (such as sensor malfunctions). For environmental feature weights, These are environmental characteristic factors.

[0206] The method for calculating the takeover request time is as follows:

[0207] ,

[0208] In the formula For the time of the takeover request, For API call response delay, This is to compensate for the time lost due to database updates;

[0209] ,

[0210] In the formula For index retrieval time, For data extraction time, Calculate the time taken for takeover.

[0211] The database expansion and update mechanism in this implementation method is as follows:

[0212] Online learning update plan:

[0213] ,

[0214] in This is the feature vector of the current model. For learning rate, This is the newly acquired feature vector.

[0215] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for determining a time for an unmanned agricultural vehicle to request a human driver to take over, characterized in that comprise, collecting different original action sequences of the driver, and collecting situational reaction delay time of each driver; standardizing each action in the original action sequence to obtain a standardized action sequence; and labeling the duration of the standardized action sequence, and recording a feature vector of each standardized action in combination with the situational reaction delay time; mapping each original action sequence to environmental features, and mapping the feature vector of each standardized action to driver attribute features; calculating a driver steering wheel offset standard deviation based on the feature vector and the mapping of the feature vector to the driver attribute features, and calculating a driver state corrected reaction time in combination with a fatigue correction factor of the original action sequence; calculating an original action sequence complexity based on the mapping of each original action sequence to environmental features, and calculating an action coupling correction time; calculating a risk advance based on an environmental risk index; calculating a safety advance based on a safety time; calculating a system request time based on the driver state corrected reaction time, the action coupling correction time, the risk advance, and the safety advance, and calculating a takeover request time in combination with an API call response delay.

2. The method of claim 1, wherein I different original action sequences (1) of the driver are collected: ， In the formula is the jth original action sequence, , is the jth original action sequence, is the jth original action in the jth original action sequence; is the action function, is the sensor signal strength vector corresponding to the jth original action, is the set of environmental variables corresponding to the jth original action, is the timestamp corresponding to the jth original action; the original actions include stepping on the accelerator, steering, returning to the straight line, looking around, changing gears, stepping on the brake, and operating the working device.​ The situation reaction delay time of the kth driver is expressed as , , is the total number of drivers; right The method for standardization is as follows: ， wherein is the jth standardized action in the ith normalized action sequence, is the jth standardized action, is the jth raw action mean, is the jth raw action standard deviation.

3. The method of claim 2, wherein Standardized motion sequence labeling duration : ， wherein is the duration of the i-th original action sequence, is the end action tag in the i-th original action sequence; The behavior feature vector of the jth normalized action is represented as : ， In the formula is the jth original action operation frequency, is the jth original action duration, is the jth original action reaction delay, is the jth original action operation amplitude, is the jth original action motion coupling characteristic.

4. The method of claim 3, wherein the mapping of each original action sequence to environmental features is established: ， In the formula is the rth environmental feature, and the environmental features include soil state, weather condition, terrain, work link, and crop state; and a mapping of a feature vector of each standardized action and a driver attribute feature is established. ， In the formula are driver attribute features, the driver attribute features including age, gender, experience, and reaction characteristics.

5. The method of claim 4, wherein the method for calculating the driver steering wheel offset standard deviation is: ， In the formula is the standard deviation of the driver's steering wheel offset, is the function of the influence of the driver's attribute characteristics and behavior characteristics on the driver's stability, is the instantaneous direction angle, is the average direction angle; the method for calculating the driver state corrected reaction time is: ， wherein is the driver state corrected reaction time, is the average reaction time, is the reaction time safety factor, is the reaction time standard deviation, is the operating stability factor, is the fatigue correction factor; ， wherein is an empirical coefficient, is a fatigue strength index.

6. The method of claim 5, wherein the method for calculating the action coupling correction time is: ， wherein is the number of original actions in the original action sequence, is the action coupling correction time, is the jth original action weight, is the original action sequence complexity; ， In the formula is the jth original action in the ith original action sequence, is the synchronization coupling coefficient of the jth original action in the ith original action sequence, is an environmental feature influence function on the complexity.

7. The method of claim 6, wherein the method for calculating the risk advance based on the environmental risk index is: ， wherein is an environmental risk indicator, A is a number of environmental indicators, is a risk weight, is an environmental indicator; the environmental indicators include obstacle distance, humidity, slope, and visibility; ， In the formula is the risk advance, is the risk sensitivity coefficient; the method for calculating the safety advance is: ， In the formula is a safety advance, is a safety amplification factor, is a minimum safety operation time.

8. The method of claim 7, wherein the method for calculating the system request time is: ， In the formula is a system request time, is a coupling amplification coefficient, is an environmental burst correction coefficient, is an environmental characteristic weight, is an environmental characteristic factor.

9. The method of claim 8, wherein the method for calculating the takeover request time is: ， In the formula is the time to take over the request, is the API call response delay, is the time compensation due to database updates; ， In the formula is the index search time, is the data extraction time, is the takeover time calculation time.

10. An unmanned agricultural machine man-machine co-driving interface request time determination system, characterized in that comprise, a data collection module for collecting different original action sequences of the driver, and collecting situational reaction delay time of each driver; a data processing module for standardizing each action in the original action sequence to obtain a standardized action sequence; and labeling the duration of the standardized action sequence, and recording a feature vector of each standardized action in combination with the situational reaction delay time; a reaction characteristic module, configured to establish a mapping between each original action sequence and an environmental characteristic, and establish a mapping between a feature vector of each standardized action and a driver attribute characteristic; a time calculation module, configured to calculate a driver steering wheel offset standard deviation based on the feature vector and the mapping between the feature vector and the driver attribute characteristic, and calculate a driver state corrected reaction time by combining a fatigue correction factor of the original action sequence; calculate an original action sequence complexity based on the mapping between each original action sequence and the environmental characteristic, and calculate an action coupling correction time; calculate a risk advance based on the environmental risk index; calculate a safety advance based on the safety time; calculate a system request time based on the driver state corrected reaction time, the action coupling correction time, the risk advance, and the safety advance, and calculate a takeover request time by combining an API call response delay.

Images