Electric wheelchair rollover risk detection and prevention system

Abstract

The invention relates to the technical field of intelligent control, and discloses an electric wheelchair rollover risk detection and prevention system which comprises a data acquisition module, a data preprocessing module, a risk detection module, a risk prevention module and an intelligent control module. The risk detection module and the risk prevention module are arranged, so that a risk identification model is constructed, a risk prediction value is acquired, risk grades are divided based on the prediction value of the risk detection module, a corresponding risk prevention scheme is acquired, and an optimal risk prevention scheme is acquired based on an improved genetic algorithm; the risk prevention scheme is formulated while the risk is detected, corresponding schemes are obtained for different risk levels, various different risk environments and risk conditions can be effectively handled, rollover risk conditions are obtained from multiple angles, the method is more comprehensive and accurate, the effect of the risk prevention scheme is evaluated through the deep learning technology, and the risk prevention efficiency is improved. Therefore, pertinence and effectiveness of rollover risk prevention of the electric wheelchair are improved. And environmental data and functional data are fully utilized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and more specifically to a rollover risk detection and prevention system for an electric wheelchair. Background Art

[0002] In recent years, with the continuous improvement of human living standards and the advancement of medical technology, wheelchair equipment has played an important role in providing convenience and support. Electric wheelchairs are also favored by more and more people due to their high intelligence and convenience.

[0003] A public document with publication number CN118259584A discloses a dynamic balance control method for an electric wheelchair based on a gyroscope, wherein the pitch angle and roll angle of the current electric wheelchair are obtained through the gyroscope, as well as the driving speed of the electric wheelchair; the pitch angle, roll angle and driving speed data of the electric wheelchair are input into an offline trained random forest model to predict the rollover of the electric wheelchair; when it is predicted that the electric wheelchair will roll over, a linear Hall sensor is used to detect the position and speed of the electric wheelchair motor, and the electric wheelchair is controlled to run at zero speed or low speed under rated load; when it is predicted that the electric wheelchair will not roll over, the electric wheelchair adjusts the lowering and retracting distance of the left and right support wheels in real time, so that the wheelchair device maintains a minimum roll angle and continues to travel; by constructing a random forest model, the risk of the electric wheelchair rolling over is monitored in real time, and dynamic balance control of the electric wheelchair is achieved.

[0004] However, since electric wheelchairs face a variety of rollover risk situations, different risk prevention strategies need to be formulated for different risk levels, so as to reduce the risk while ensuring the best user experience. The adjustment of different parameters of electric wheelchairs to the changing trend of risk levels is the key to solving the rollover risk. It is necessary to combine the functional parameters of the electric wheelchair for analysis to obtain the best adjustment form to reduce the risk, effectively prevent rollover, and ensure the user experience.

[0005] In view of this, the present application proposes an electric wheelchair rollover risk detection and prevention system to detect the rollover risk of the electric wheelchair and formulate a personalized specific risk prevention plan. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides an electric wheelchair rollover risk detection and prevention system to solve the problems existing in the above-mentioned background technology.

[0007] The present invention provides the following technical solutions: an electric wheelchair rollover risk detection and prevention system, comprising a data acquisition module, a data preprocessing module, a risk detection module, a risk prevention module and an intelligent control module; The data acquisition module is used to collect target data and transmit it to the data preprocessing module; The data preprocessing module is used to perform preprocessing operations on the data, and perform data fusion on the preprocessed data to obtain fusion target data; The risk detection module is used to input the fusion target data into the constructed risk identification model, obtain the risk prediction value, and transmit it to the risk prevention module; The risk prevention module divides the risk levels and obtains corresponding risk prevention plans based on the predicted values ​​of the risk detection module, and obtains the best risk prevention plan from the corresponding risk prevention plans based on the improved genetic algorithm; The intelligent control module is used to execute risk prevention solutions.

[0008] Preferably, the target data includes environmental data and functional data, the data acquisition module includes an environmental data acquisition unit and a functional data acquisition unit, the environmental data acquisition unit is used to collect environmental data of the target object through visual sensors and edge measurement sensors, and transmit the data to the data preprocessing module, the functional data acquisition unit is used to collect functional data of the target object through torque sensors and inertial sensors, and transmit the data to the data preprocessing module; the target object is an electric wheelchair that needs to be detected for rollover risk.

[0009] Preferably, the constructed risk identification model is specifically an LSTM neural network model, and the construction method is specifically as follows: Initialize the basic architecture of the LSTM neural network model, which includes an input layer, two LSTM units, a Dropout layer, a fully connected layer, a learnable parameter activation function, a comprehensive loss function, an optimizer, a regularizer, and an output layer; The input layer is used to input fusion target data; Each LSTM unit consists of a memory unit and three control gates; the three control gates are input gate, forget gate and output gate; a fully connected layer is added between LSTM units; The comprehensive loss function LOSS is expressed as follows: ; Where a is the quantile, indicating the quantile level in the loss calculation; a takes a value between (0, 1); y true is the real target value, indicating the actual risk value; is the lower quantile of the predicted distribution; is the upper quantile of the predicted distribution; is a positive function; The output layer is used to output the risk prediction value; Collect the fused target data of the target object within a period of time in history, and collect the risk value corresponding to the historical fused target data to construct training data; divide the training data into a training set and a validation set; During the training process, when the function value of the comprehensive loss function does not decrease during P consecutive iterations, the training is completed and a trained LSTM neural network model is obtained; The real-time fusion target data is input into the trained LSTM neural network model to obtain the risk prediction value.

[0010] Preferably, the calculation formula of the actual risk value is expressed as: , where D is the actual risk value, HJ is the environmental data risk coefficient, XN is the performance risk coefficient of the target object, and k1 and k2 are the corresponding proportional factors; , among which, MJ i is the area of ​​the i-th type of ground obstacle, YD i is the hardness of the i-th type of ground obstacle, L is the total number of types of ground obstacles, and QT is other influencing factors; i=1, 2, 3, …, L; , where ZT j is the jth functional data fluctuation value, m is the total number of functional data, v is the real-time speed, ω is the angular speed; j=1, 2, 3, …, m; the total number of functional data is the number of types of functional data collected; , where B jt is the actual value of the jth functional data at the tth moment, B j ´ is the set value of the jth functional data at the tth moment, T is the total time, and the values ​​of each functional data at T moments are collected, t=1, 2, 3, …, T.

[0011] Preferably, the specific method of dividing the risk levels is: If the risk prediction value satisfies Y≥IU1, the risk level is high; If the risk prediction value satisfies IU2≤Y<IU1, the risk level is medium; If the risk prediction value satisfies Y<IU2, the risk level is low; The IU1 is the first risk threshold, and IU2 is the second risk threshold, satisfying IU1≥85%, 40%≤IU2<85%; When the risk level is high, obtain a high-level risk prevention plan; when the risk level is medium, obtain a medium-level risk prevention plan; when the risk level is low, obtain a low-level risk prevention plan.

[0012] Preferably, the intermediate risk prevention scheme includes multiple sub-scheme sets. , C is the intermediate risk prevention plan, ch is the hth sub-plan set, , where zm is the mth functional data, em is the adjustment value of the mth functional data, if em is a positive value, it is adjusted upward, that is, the value corresponding to the functional data is increased, if em is a negative value, it is adjusted downward, that is, the value corresponding to the functional data is reduced; There is only one solution for both the low-level risk prevention solution and the high-level risk prevention solution, so this solution is the best risk prevention solution. When the risk prevention solution obtained is a medium-level risk prevention solution, the improved genetic algorithm is used to obtain the best risk prevention solution.

[0013] Preferably, the specific method of obtaining the best risk prevention plan is: Step S01: Encode the sub-scheme set in the intermediate risk prevention scheme, obtain chromosomes, and construct an initial population; Step S02: Determine the fitness function; Step S03: Perform natural selection on chromosomes in the population; Step S04: performing crossover recombination on chromosomes in the population; Step S05: mutating the chromosomes in the population; Step S06: Obtain a new population, preset the population generation to be K, the fitness threshold to be Q, K is an integer greater than 0, and Q is a real number greater than 0; loop steps S03 to S05 until the generation corresponding to the new population is K or the fitness corresponding to the chromosome in the new population is greater than or equal to the fitness threshold Q, the loop ends, and the sub-plan set corresponding to the chromosome with the maximum fitness in the new population is obtained, which is the optimal risk prevention plan.

[0014] Preferably, the sub-scheme set in the intermediate risk prevention scheme is coded as g, where g is a chromosome, and U chromosomes are randomly generated to form an initial population. ; The expression of the fitness function is: , where f r is the fitness of the rth chromosome, PP r is the risk mitigation value of the rth chromosome, r=1, 2, 3, …, U; The method for obtaining the risk level mitigation value is: using the sub-scheme set corresponding to the chromosome and the risk value data as research data, inputting the research data into a trained risk mitigation prediction model, and predicting the corresponding risk level mitigation value.

[0015] Preferably, the training process of the risk mitigation prediction model is: A set of functional data adjustment values ​​and environmental data are taken as a set of analysis data, d sets of analysis data are collected in advance, d is an integer greater than 1, and the analysis data and the corresponding risk degree mitigation values ​​are converted into a corresponding set of feature vectors; the risk degree mitigation value corresponding to the analysis data can be calculated by subtracting the actual risk value before the functional data adjustment from the actual risk value after the functional data adjustment, d sets of analysis data are collected, and under the conditions of each set of analysis data, the influence of the change trend of the risk degree on the rollover of the electric wheelchair under the real-time environmental data and the functional data is comprehensively analyzed; Each set of feature vectors is used as the input of the risk mitigation prediction model. The risk mitigation prediction model uses a set of risk mitigation values ​​corresponding to each set of analysis data as output, and uses the actual risk mitigation value corresponding to each set of analysis data as the prediction target. The actual risk mitigation value is the risk mitigation value corresponding to the analysis data collected in advance; the training target is to minimize the sum of the prediction errors of all analysis data; the formula of the prediction error is expressed as: , where ε p is the prediction error, p is the group number of the eigenvector corresponding to the analyzed data, θ p is the predicted risk mitigation value corresponding to the p-th group analysis data, μ p The actual risk level mitigation value corresponding to the p-th group of analysis data is used to train the risk mitigation prediction model until the sum of the prediction errors reaches convergence and the training is stopped.

[0016] Technical effects and advantages of the present invention: The present invention is provided with a risk detection module and a risk prevention module, which is conducive to building a risk identification model, obtaining risk prediction values, and dividing risk levels and obtaining corresponding risk prevention plans based on the prediction values ​​of the risk detection module, and obtaining the best risk prevention plan based on the improved genetic algorithm; formulating risk prevention plans while detecting risks, and obtaining corresponding plans for different risk levels, which can effectively cope with a variety of different risk environments and risk conditions, and obtain rollover risk conditions from multiple angles, which is more comprehensive and accurate, and integrates the trained neural network model into the fitness calculation of the improved genetic algorithm, which can give full play to the nonlinear and multimodal data modeling capabilities of the neural network and effectively capture complex data relationships; based on the parallel computing processing mechanism, the computing efficiency is improved, and the effect of the risk prevention plan is evaluated through deep learning technology, thereby improving the pertinence and effectiveness of the prevention of electric wheelchair rollover risks; making full use of environmental data and functional data, timely detecting and preventing risks, and effectively reducing the incidence of rollover risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a structural diagram of the electric wheelchair rollover risk detection and prevention system of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are merely illustrative. The electric wheelchair rollover risk detection and prevention system involved in the present invention is not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0019] like Figure 1 As shown, the present invention provides an electric wheelchair rollover risk detection and prevention system, including a data acquisition module, a data preprocessing module, a risk detection module, a risk prevention module and an intelligent control module; The data acquisition module is used to collect target data and transmit it to the data preprocessing module; the target data includes environmental data and functional data, the data acquisition module includes an environmental data acquisition unit and a functional data acquisition unit, the environmental data acquisition unit is used to collect environmental data of the target object through a visual sensor and an edge measurement sensor, and transmit it to the data preprocessing module, the functional data acquisition unit is used to collect functional data of the target object through a torque sensor and an inertial sensor, and transmit it to the data preprocessing module; the target object is an electric wheelchair that needs to be detected for rollover risk; the environmental data is the surrounding environment of the target object, including but not limited to obstacle detection, road conditions, wheelchair position, distance between the ground and the wheelchair, and ground slope, etc. The obstacle detection situation specifically refers to whether there are ground obstacles and the type of obstacles, and the road conditions specifically refer to whether there are slopes; the functional data includes but is not limited to data describing the performance status of the electric wheelchair, such as motor power, output torque, acceleration and angular velocity of the electric wheelchair; The data preprocessing module is used to perform preprocessing operations on the data, and perform data fusion on the preprocessed data to obtain fused target data; the preprocessing operations include but are not limited to cleaning and denoising the data, removing useless data and duplicate data, and obtaining target data that can be directly used; The risk detection module is used to input the fusion target data into the constructed risk identification model, obtain the risk prediction value, and transmit it to the risk prevention module; The risk prevention module divides the risk level and obtains the corresponding risk prevention plan based on the predicted value of the risk detection module, and obtains the best risk prevention plan from the corresponding risk prevention plan based on the improved genetic algorithm; the purpose is to integrate the trained neural network model into the fitness calculation of the improved genetic algorithm by adopting the improved genetic algorithm, so as to give full play to the nonlinear and multimodal data modeling capabilities of the neural network and effectively capture complex data relationships; based on the parallel computing processing mechanism, the computing efficiency is improved, and the effect of the risk prevention plan is evaluated through deep learning technology, so as to improve the pertinence and effectiveness of the prevention of the risk of electric wheelchair rollover; make full use of environmental data and functional data, and capture complex data relationships through deep learning technology, so as to fully and accurately understand the prevention plans of different electric wheelchair risk levels, and provide customized plans for the risk prevention of electric wheelchairs, thereby effectively improving the effectiveness and accuracy of the prevention of the risk of electric wheelchair rollover; The intelligent control module is used to execute risk prevention solutions.

[0020] In this embodiment, it should be specifically explained that the torque sensor is installed on the driving shaft of the electric wheelchair and is closely connected to the motor of the electric wheelchair to monitor the torque of the electric wheelchair in real time. The torque sensor can be detected by strain measurement, current measurement, etc. It is very important to judge whether the electric wheelchair is working normally and whether it is disturbed by external forces. The inertial sensor, or IMU, is a sensor component that integrates an accelerometer and a gyroscope. It is installed on the main part of the electric wheelchair and can sense the tilt, acceleration, and deceleration of the electric wheelchair, providing important dynamic data to help determine the stability and motion state of the electric wheelchair.

[0021] In this embodiment, it should be specifically explained that the constructed risk identification model is specifically an LSTM neural network model, and the construction method is specifically as follows: Initialize the basic architecture of the LSTM neural network model, which includes an input layer, two LSTM units, a Dropout layer, a fully connected layer, a learnable parameter activation function, a comprehensive loss function, an optimizer, a regularizer, and an output layer; the output layer can have n nodes, representing n moments in the future; The input layer is used to input fusion target data; Each LSTM unit consists of a memory unit and three control gates; the three control gates are input gate, forget gate and output gate; a fully connected layer is added between LSTM units; The Dropout layer randomly discards a certain proportion of nodes from the LSTM neural network model temporarily, which can effectively prevent overfitting; The learnable parameter activation function is: ; Where f is the input data; p1 is a learnable slope parameter that controls the slope of the linear transformation of f. The value of p1 affects the gradient of the Sigmoid function in the input space, thereby affecting the response speed of the Sigmoid function to the input; learning p1 can make the model adapt to different ranges of activation values ​​in the data; p2 is a learnable offset parameter that controls the offset of the Sigmoid function; enabling the model to adapt to different baselines or offsets in the input data; learning p2 can enable the model to better model the overall pattern of the data; The specific values ​​of P1 and p2 are obtained through model training; in deep learning, training is performed through the back-propagation algorithm and optimizer, which adjusts the parameters in the network so that the model achieves better performance on the training data; Uncertainty modeling is introduced into the loss function to obtain the comprehensive loss function LOSS, which is expressed as follows: ; where a is the quantile, indicating the quantile level in the loss calculation; a takes a value between (0, 1), which determines whether the loss is calculated for the lower quantile or the upper quantile; y true is the real target value, indicating the actual risk value; is the lower quantile of the predicted distribution; is the upper quantile of the predicted distribution; is the positive function; it is used to ensure that only when the difference between the predicted value and the true value is positive will it contribute to the loss; The optimizer adopts an adaptive optimizer that adaptively adjusts the learning rate; The optimization formula of the adaptive optimizer is: ;in, is the parameter of the bth layer of the LSTM neural network model when the number of training iterations is λ+1; is the parameter of the bth layer of the LSTM neural network model when the number of training iterations is λ; η b is the learning rate of the bth layer of the LSTM neural network model, and different learning rates are selected according to different layers; v λ is the exponential moving average of the squared gradient for training iterations λ; is a small constant that prevents the denominator from being zero; m λ,b is the parameter gradient when the number of training iterations is λ; like , then there is the formula: ; Among them LOSS λ is the value of the comprehensive loss function when the number of training iterations is λ; LOSS λ-1 is the value of the comprehensive loss function when the number of training iterations is λ-1; A is the reduction parameter; The regularizer uses L2 norm regularization to control the complexity of the LSTM neural network model and prevent overfitting; The output layer is used to output the risk prediction value; Collect the fusion target data of the target object within a period of history, and collect the risk value corresponding to the historical fusion target data to construct training data; divide the training data into a training set and a validation set; the specific division ratio can be determined according to the actual situation. For example, 80% of the training data is used as the training set and 20% of the training data is used as the validation set; During the training process, when the function value of the comprehensive loss function does not decrease during P consecutive iterations, the training is completed and a trained LSTM neural network model is obtained; The real-time fusion target data is input into the trained LSTM neural network model to obtain the risk prediction value.

[0022] In this embodiment, it should be specifically explained that the calculation formula of the actual risk value is expressed as: , where D is the actual risk value, HJ is the environmental data risk coefficient, XN is the performance risk coefficient of the target object, k1 and k2 are the corresponding proportional factors, both of which are greater than 0 and less than 1. The specific values ​​are set by those skilled in the art. The risk value corresponding to the historical fusion target data can also be calculated and obtained in this way; , among which, MJ i is the area of ​​the i-th type of ground obstacle, YD i is the hardness of the i-th type of ground obstacle, L is the total number of types of ground obstacles, and QT is other influencing factors; i=1, 2, 3, ..., L; the hardness of the ground obstacle can be obtained through a hardness table, the area of ​​the ground obstacle is the floor area, there are many types of ground obstacles, including roadblocks, stones, garbage, and trees, etc., and the risks of different obstacles are different. The higher the hardness of the obstacle, the more severe the bumps caused when passing through the obstacle, and the larger the area of ​​the obstacle, the longer the duration of the bumps when passing through the obstacle, so the risk will be higher; the other influencing factors are other influencing factors that may cause the risk of wheelchair rollover, which may be weather influencing factors or slope influencing factors. Other influencing factors, other influencing factors are taken as (0, 2]. When other influencing factors are weather influencing factors, the better the weather, the smaller the value, and the worse the weather, the larger the value. The weather can be determined by the wind speed of the day as the main factor, and the others as secondary factors, because the wind speed has the greatest impact on the wheelchair rollover. If the wind speed is higher, the weather is worse, the value is larger, and the risk of wheelchair rollover is higher. When the wind speed is zero, QT can be taken as 1; when other influencing factors are slope influencing factors, the steeper the slope, the larger the value, and the gentler the slope, the smaller the value. When the slope is 45°, the value can be 1; the values ​​of other influencing factors can be specifically set by technicians in this field in combination with actual conditions and professional knowledge; , where ZT j is the jth functional data fluctuation value, m is the total number of functional data, v is the real-time speed, and ω is the angular velocity; j=1, 2, 3, …, m; the total number of functional data is the number of types of collected functional data. For example, if the collected functional data is motor power and output torque, m is 2; if the collected functional data is motor power, output torque, electric wheelchair acceleration and angular velocity, m is 4; , where B jt is the actual value of the jth functional data at the tth moment, B j ´ is the set value of the jth functional data at the tth moment, T is the total time, and the values ​​of each functional data at T moments are collected, t=1, 2, 3, …, T.

[0023] In this embodiment, it should be specifically explained that the acquisition of fusion target data can be performed by data fusion through Kalman filtering, extended Kalman filtering, particle filtering and other technologies. If the target data is represented by , where M is the preprocessed target data, P is the preprocessed environment data, and G is the preprocessed function data. The fused target data can be expressed as: , where R is the fusion target data, and F is the fusion function. The selection of the fusion function is consistent with the technical means used for data fusion. If the particle filter technology is used for data fusion, F represents the particle filter algorithm function.

[0024] In this embodiment, it should be specifically explained that the specific method of dividing the risk levels is: If the risk prediction value satisfies Y≥IU1, the risk level is high; If the risk prediction value satisfies IU2≤Y<IU1, the risk level is medium; If the risk prediction value satisfies Y<IU2, the risk level is low; IU1 is the first risk threshold, and IU2 is the second risk threshold, which satisfies IU1≥85%, and 40%≤IU2<85%. The specific values ​​of IU1 and IU2 are set by those skilled in the art according to actual conditions, and this embodiment does not set these values ​​specifically. When the risk level is high, obtain the high-level risk prevention plan; when the risk level is medium, obtain the medium-level risk prevention plan; when the risk level is low, obtain the low-level risk prevention plan; The low-level risk prevention solution is to issue a warning to the user through sound, vibration or display screen, reminding the user that there is a risk of rollover and that the user needs to pay attention to the road conditions and speed in real time; the intermediate risk prevention solution is to adjust the motor power, output torque and electric wheelchair acceleration and other functional data to gradually reduce the risk of the electric wheelchair until the risk is minimized or there is no risk; the advanced risk prevention solution is to slowly stop the electric wheelchair from moving; There are multiple sub-scheme sets in the intermediate risk prevention scheme. , C is the intermediate risk prevention plan, ch is the hth sub-plan set, , where zm is the mth functional data, em is the adjustment value of the mth functional data, if em is a positive value, it is adjusted upward, that is, the value corresponding to the functional data is increased, if em is a negative value, it is adjusted downward, that is, the value corresponding to the functional data is reduced; There is only one solution for both the low-level risk prevention solution and the high-level risk prevention solution, so this solution is the best risk prevention solution. When the risk prevention solution obtained is a medium-level risk prevention solution, the improved genetic algorithm is used to obtain the best risk prevention solution.

[0025] In this embodiment, it should be specifically explained that the specific method of obtaining the best risk prevention plan is: Step S01: Encode the sub-scheme set in the intermediate risk prevention scheme, obtain chromosomes, and construct an initial population; Step S02: Determine the fitness function; Step S03: Perform natural selection on chromosomes in the population; Step S04: performing crossover recombination on chromosomes in the population; Step S05: mutating the chromosomes in the population; Step S06: Obtain a new population, and preset the population generation number to be K, the fitness threshold to be Q, K is an integer greater than 0, and Q is a real number greater than 0; loop steps S03 to S05 until the generation number corresponding to the new population is K or the fitness corresponding to the chromosome in the new population is greater than or equal to the fitness threshold Q, the loop ends, and the sub-scheme set corresponding to the chromosome with the maximum fitness in the new population is obtained, which is the optimal risk prevention scheme; illustratively, if the preset population generation number is 1, the chromosomes in the initial population are subjected to natural selection, crossover recombination, and mutation to obtain a new population. At this time, the generation number corresponding to the new population is 1, so the loop ends.

[0026] In this embodiment, it should be specifically explained that the sub-scheme set in the intermediate risk prevention scheme is coded as g, where g is a chromosome, and U chromosomes are randomly generated to form an initial population. ; The expression of the fitness function is: , where f r is the fitness of the rth chromosome, PP r is the risk mitigation value of the rth chromosome, r=1, 2, 3, …, U; The method for obtaining the risk degree mitigation value is as follows: using the sub-scheme set corresponding to the chromosome and the risk value data as research data, inputting the research data into a trained risk mitigation prediction model, and predicting the corresponding risk degree mitigation value; the training process of the risk mitigation prediction model is as follows: A set of functional data adjustment values ​​and environmental data are taken as a set of analysis data, d sets of analysis data are collected in advance, d is an integer greater than 1, and the analysis data and the corresponding risk degree mitigation values ​​are converted into a corresponding set of feature vectors; the risk degree mitigation value corresponding to the analysis data can be calculated by subtracting the actual risk value before the functional data adjustment from the actual risk value after the functional data adjustment, d sets of analysis data are collected, and under the conditions of each set of analysis data, the influence of the change trend of the risk degree on the rollover of the electric wheelchair under the real-time environmental data and the functional data is comprehensively analyzed; Each set of feature vectors is used as the input of the risk mitigation prediction model. The risk mitigation prediction model uses a set of risk mitigation values ​​corresponding to each set of analysis data as output, and uses the actual risk mitigation value corresponding to each set of analysis data as the prediction target. The actual risk mitigation value is the risk mitigation value corresponding to the analysis data collected in advance; the training target is to minimize the sum of the prediction errors of all analysis data; the formula of the prediction error is expressed as: , where ε p is the prediction error, p is the group number of the eigenvector corresponding to the analyzed data, θ p is the predicted risk mitigation value corresponding to the p-th group analysis data, μ p The actual risk level mitigation value corresponding to the p-th group of analysis data is used to train the risk mitigation prediction model until the sum of the prediction errors reaches convergence and the training is stopped; The risk mitigation prediction model is specifically a deep neural network model; it includes an input layer, a hidden layer and an output layer; each hidden layer includes multiple neurons, each neuron is connected to the neurons in the next layer, and the connection contains weights to determine the importance and influence of data transmission in the neural network; each neuron between the hidden layer and the output layer is applied with an activation function, and the activation function is mapped into nonlinearity, allowing the network to learn more complex patterns and features; The natural selection is carried out by combining the elite method and the rotation method, and the crossover recombination is carried out by the PMX method.

[0027] In the present embodiment, it should be specifically explained that the difference between the present embodiment and the prior art lies mainly in that the present embodiment has a risk detection module and a risk prevention module, constructs a risk identification model, obtains risk prediction values, and divides risk levels and obtains corresponding risk prevention plans based on the prediction values ​​of the risk detection module, and obtains the best risk prevention plan based on the improved genetic algorithm; formulates risk prevention plans while detecting risks, obtains corresponding plans for different risk levels, can effectively respond to a variety of different risk environments and risk conditions, and obtains rollover risk conditions from multiple angles, which is more comprehensive and accurate, and integrates the trained neural network model into the fitness calculation of the improved genetic algorithm, which can give full play to the nonlinear and multimodal data modeling capabilities of the neural network and effectively capture complex data relationships; improves computing efficiency based on the parallel computing processing mechanism, and evaluates the effectiveness of the risk prevention plan through deep learning technology, thereby improving the pertinence and effectiveness of the prevention of electric wheelchair rollover risks; and makes full use of environmental data and functional data.

[0028] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0029] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An electric wheelchair rollover risk detection and prevention system, characterized in that: It includes data acquisition module, data preprocessing module, risk detection module, risk prevention module and intelligent control module; The data acquisition module is used to collect target data and transmit it to the data preprocessing module; The data preprocessing module is used to perform preprocessing operations on the data, and perform data fusion on the preprocessed data to obtain fusion target data; The risk detection module is used to input the fusion target data into the constructed risk identification model, obtain the risk prediction value, and transmit it to the risk prevention module; The risk prevention module divides the risk levels and obtains corresponding risk prevention plans based on the predicted values ​​of the risk detection module, and obtains the best risk prevention plan from the corresponding risk prevention plans based on the improved genetic algorithm; The intelligent control module is used to execute risk prevention solutions.

2. The electric wheelchair rollover risk detection and prevention system according to claim 1, characterized in that: The target data includes environmental data and functional data. The data acquisition module includes an environmental data acquisition unit and a functional data acquisition unit. The environmental data acquisition unit is used to collect environmental data of the target object through a visual sensor and an edge measurement sensor, and transmit the data to the data preprocessing module. The functional data acquisition unit is used to collect functional data of the target object through a torque sensor and an inertial sensor, and transmit the data to the data preprocessing module. The target object is an electric wheelchair that needs to be detected for rollover risk.

3. The electric wheelchair rollover risk detection and prevention system according to claim 2, characterized in that: The constructed risk identification model is specifically an LSTM neural network model, and the construction method is specifically as follows: Initialize the basic architecture of the LSTM neural network model, which includes an input layer, two LSTM units, a Dropout layer, a fully connected layer, a learnable parameter activation function, a comprehensive loss function, an optimizer, a regularizer, and an output layer; The input layer is used to input fusion target data; Each LSTM unit consists of a memory unit and three control gates; the three control gates are input gate, forget gate and output gate; a fully connected layer is added between LSTM units; The comprehensive loss function LOSS is expressed as follows: Among them, a is the quantile, which represents the quantile level in the loss calculation; a takes a value between (0, 1); y true is the real target value, indicating the actual risk value; is the lower quantile of the predicted distribution; is the upper quantile of the predicted distribution; is a positive function; The output layer is used to output the risk prediction value; Collect the fused target data of the target object within a period of time in history, and collect the risk value corresponding to the historical fused target data to construct training data; divide the training data into a training set and a validation set; During the training process, when the function value of the comprehensive loss function does not decrease during P consecutive iterations, the training is completed and a trained LSTM neural network model is obtained; The real-time fusion target data is input into the trained LSTM neural network model to obtain the risk prediction value.

4. The electric wheelchair rollover risk detection and prevention system according to claim 3, characterized in that: The calculation formula of the actual risk value is expressed as: , where D is the actual risk value, HJ is the environmental data risk coefficient, XN is the performance risk coefficient of the target object, and k1 and k2 are the corresponding proportional factors; , among which, MJ i is the area of ​​the i-th type of ground obstacle, YD i is the hardness of the i-th type of ground obstacle, L is the total number of types of ground obstacles, and QT is other influencing factors; i=1, 2, 3, …, L; , where ZT j is the jth functional data fluctuation value, m is the total number of functional data, v is the real-time speed, ω is the angular speed; j=1, 2, 3, …, m; the total number of functional data is the number of types of collected functional data; , where B jt is the actual value of the jth functional data at the tth moment, B j ´ is the set value of the jth functional data at the tth moment, T is the total time, and the values ​​of each functional data at T moments are collected, t=1, 2, 3, …, T.

5. The electric wheelchair rollover risk detection and prevention system according to claim 4, characterized in that: The specific method of dividing the risk levels is as follows: If the risk prediction value satisfies Y≥IU1, the risk level is high; If the risk prediction value satisfies IU2≤Y<IU1, the risk level is medium; If the risk prediction value satisfies Y<IU2, the risk level is low; The IU1 is the first risk threshold, and IU2 is the second risk threshold, satisfying IU1≥85%, 40%≤IU2<85%; When the risk level is high, obtain a high-level risk prevention plan; when the risk level is medium, obtain a medium-level risk prevention plan; when the risk level is low, obtain a low-level risk prevention plan.

6. The electric wheelchair rollover risk detection and prevention system according to claim 5, characterized in that: There are multiple sub-scheme sets in the intermediate risk prevention scheme. , C is the intermediate risk prevention plan, ch is the hth sub-plan set, , where zm is the mth functional data, em is the adjustment value of the mth functional data, if em is a positive value, it is adjusted upward, that is, the value corresponding to the functional data is increased, if em is a negative value, it is adjusted downward, that is, the value corresponding to the functional data is reduced; There is only one solution for both the low-level risk prevention solution and the high-level risk prevention solution, so this solution is the best risk prevention solution. When the risk prevention solution obtained is a medium-level risk prevention solution, the improved genetic algorithm is used to obtain the best risk prevention solution.

7. The electric wheelchair rollover risk detection and prevention system according to claim 6, characterized in that: The specific method of obtaining the best risk prevention plan is: Step S01: Encode the sub-scheme set in the intermediate risk prevention scheme, obtain chromosomes, and construct an initial population; Step S02: Determine the fitness function; Step S03: Perform natural selection on chromosomes in the population; Step S04: performing crossover recombination on chromosomes in the population; Step S05: mutating the chromosomes in the population; Step S06: Obtain a new population, preset the population generation to be K, the fitness threshold to be Q, K is an integer greater than 0, and Q is a real number greater than 0; loop steps S03 to S05 until the generation corresponding to the new population is K or the fitness corresponding to the chromosome in the new population is greater than or equal to the fitness threshold Q, the loop ends, and the sub-plan set corresponding to the chromosome with the maximum fitness in the new population is obtained, which is the optimal risk prevention plan.

8. The electric wheelchair rollover risk detection and prevention system according to claim 7, characterized in that: The sub-scheme set in the intermediate risk prevention scheme is coded as g, where g is a chromosome. U chromosomes are randomly generated to form the initial population. ; The expression of the fitness function is: , where f r is the fitness of the rth chromosome, PP r is the risk mitigation value of the rth chromosome, r=1, 2, 3, …, U; The method for obtaining the risk level mitigation value is: using the sub-scheme set corresponding to the chromosome and the risk value data as research data, inputting the research data into a trained risk mitigation prediction model, and predicting the corresponding risk level mitigation value.

9. The electric wheelchair rollover risk detection and prevention system according to claim 8, characterized in that: The training process of the risk mitigation prediction model is: A set of functional data adjustment values ​​and environmental data are taken as a set of analysis data, d sets of analysis data are collected in advance, d is an integer greater than 1, and the analysis data and the corresponding risk degree mitigation values ​​are converted into a corresponding set of feature vectors; the risk degree mitigation value corresponding to the analysis data can be calculated by subtracting the actual risk value before the functional data adjustment from the actual risk value after the functional data adjustment, d sets of analysis data are collected, and under the conditions of each set of analysis data, the influence of the change trend of the risk degree on the rollover of the electric wheelchair under the real-time environmental data and the functional data is comprehensively analyzed; Each group of feature vectors is used as input to a risk mitigation prediction model, wherein the risk mitigation prediction model uses a group of risk mitigation values ​​corresponding to each group of analysis data as output, and uses the actual risk mitigation value corresponding to each group of analysis data as a prediction target, wherein the actual risk mitigation value is determined by the risk mitigation value corresponding to the analysis data collected in advance; The training objective is to minimize the sum of the prediction errors of all analyzed data; the formula for the prediction error is expressed as: , where ε p is the prediction error, p is the group number of the eigenvector corresponding to the analyzed data, θ p is the predicted risk mitigation value corresponding to the p-th group analysis data, μ p The actual risk level mitigation value corresponding to the p-th group of analysis data is used to train the risk mitigation prediction model until the sum of the prediction errors reaches convergence and the training is stopped.

Images