Method and System for Remote Monitoring of Electric Scooters Based on the Internet of Things

Abstract

The present invention relates to a method, device, equipment and storage medium for remote monitoring of an electric scooter based on the Internet of Things, including the following steps. If the scooter state feature vector indicates that the electric scooter has an abnormality, then perform multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain vehicle abnormal behavior pattern features; perform risk prediction on the electric scooter based on the vehicle abnormal behavior pattern features to obtain a risk level classification result; perform distributed processing and priority sorting on the risk level classification result through an edge computing gateway to obtain a control strategy instruction set; input the control strategy instruction set into the remote motor controller to control the electric scooter, solving the technical problems that the traditional management method mainly relies on manual inspection and regular maintenance, which is not only inefficient, but also difficult to monitor the vehicle state in real time and timely discover potential safety hazards.

Description

Technical Field

[0001] The present invention relates to the technical field of electric scooters, and particularly to a remote monitoring method and system for electric scooters based on the Internet of Things. Background Art

[0002] With the acceleration of urbanization and the diversification of people's travel needs, electric scooters, as a convenient and environmentally friendly short-distance means of transportation, have been rapidly popularized worldwide. However, this has led to an increasing demand for the management and safety monitoring of electric scooters. Traditional management methods mainly rely on manual inspections and regular maintenance. This method is not only inefficient but also difficult to monitor the vehicle status in real time and detect potential safety hazards in a timely manner. Especially in the sharing economy model, a large number of electric scooters have been put on the market, and how to efficiently and accurately monitor the status of these vehicles has become an urgent problem to be solved.

[0003] At the technical level, although existing electric scooters are already equipped with basic sensor devices such as GPS positioning systems and battery management systems, the data of these sensors often work independently and lack an effective integration and analysis mechanism. This results in the difficulty of extracting valuable information from the large amount of raw data collected to guide actual operations. In addition, due to the complex and changeable environment in which electric scooters are used, the data of a single type of sensor is difficult to comprehensively reflect the actual operating conditions of the vehicle, thus limiting the accuracy of the assessment of the vehicle's health status. Therefore, a more intelligent method is needed to achieve the comprehensive monitoring and management of the status of electric scooters.

[0004] Based on the above background, researchers have begun to explore the use of Internet of Things technology combined with advanced methods such as edge computing to improve the remote monitoring ability of electric scooters. By integrating the data of multiple sensors and applying advanced data analysis algorithms for processing, it is possible to more accurately identify abnormal vehicle behaviors and predict potential risks. This method can not only improve the safety of electric scooters but also provide decision-making support for operators, optimize resource allocation, and reduce operating costs. Nevertheless, achieving this goal still faces many challenges, including but not limited to the maturity of sensor data fusion technology, the stability of edge computing performance, and data privacy protection. The effective solution of these problems is crucial for promoting the development of electric scooter intelligent monitoring technology. Summary of the Invention

[0005] The main object of the present invention is to provide a remote monitoring method and system for electric scooters based on the Internet of Things, which solves the technical problem that traditional management methods mainly rely on manual inspections and regular maintenance, which is not only inefficient but also difficult to monitor the vehicle status in real time and detect potential safety hazards in a timely manner.

[0006] To achieve the above object, the present invention provides a method for remote monitoring of an electric scooter based on the Internet of Things. An electric scooter is provided with a motor controller, and the method includes the following steps:

[0007] Collect data from multiple sensors mounted on the electric scooter to obtain an original sensing data stream; wherein, the original sensing data stream includes a gyroscope sensor, an acceleration sensor, a GPS positioning sensor, and a battery management sensor;

[0008] Perform multi-dimensional feature extraction and data fusion on the original sensing data stream to obtain a scooter state feature vector;

[0009] If the scooter state feature vector indicates that the electric scooter has an abnormality, perform multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain a vehicle abnormal behavior pattern feature;

[0010] Perform risk prediction on the electric scooter based on the vehicle abnormal behavior pattern feature to obtain a risk level classification result;

[0011] Perform distributed processing and priority sorting on the risk level classification result through an edge computing gateway to obtain a control strategy instruction set;

[0012] Input the control strategy instruction set into the remote motor controller to control the electric scooter.

[0013] Further, the performing multi-dimensional feature extraction and data fusion on the original sensing data stream to obtain a scooter state feature vector includes:

[0014] Perform time-frequency domain decomposition on the original sensing data stream to obtain a vibration feature sequence of the electric scooter, and perform wavelet packet decomposition on the vibration feature sequence to obtain a multi-scale vibration spectrogram;

[0015] Perform data analysis of the scooter attitude on the electric scooter based on the multi-scale vibration spectrogram to obtain an attitude feature set, perform time series alignment on the attitude feature set to obtain an aligned state sequence, and perform sparse coding on the aligned state sequence to obtain a compressed feature representation;

[0016] Perform multi-modal feature fusion on the compressed feature representation to obtain a fusion feature matrix, and perform tensor decomposition on the fusion feature matrix to obtain a scooter state feature vector.

[0017] Further, the performing multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain a vehicle abnormal behavior pattern feature includes:

[0018] Perform multi-dimensional space projection on the state feature vector of the scooter to obtain a feature distribution mapping atlas, and perform non-linear manifold embedding on the feature distribution mapping atlas to obtain a topological feature matrix;

[0019] Perform similarity measurement on the topological feature matrix through a preset local sensitive hashing technique to obtain a feature similarity spectrum, and perform spectral clustering analysis on the feature similarity spectrum to obtain a set of behavior pattern prototypes; wherein, the set of behavior pattern prototypes includes an acceleration pattern, a steering pattern, and a braking pattern;

[0020] Extract spatio-temporal features from the set of behavior pattern prototypes to obtain a spatio-temporal feature tensor, and perform tensor decomposition on the spatio-temporal feature tensor to obtain multi-modal behavior features;

[0021] Perform temporal dependence analysis on the multi-modal behavior features through a preset conditional random field technique to obtain a state transition sequence, and perform Markov jump analysis on the state transition sequence to obtain a behavior pattern transition graph; wherein, the behavior pattern transition graph includes state duration, state transition probability, and state stability;

[0022] Perform distributed association mining on the behavior pattern transition graph based on preset historical behavior data to obtain behavior pattern association rules, and perform probabilistic graph reasoning on the behavior pattern association rules to obtain a behavior prediction vector; wherein, the behavior prediction vector includes behavior occurrence probability, behavior duration, and behavior transition trend;

[0023] Perform multi-scale feature fusion on the behavior prediction vector to obtain a multi-dimensional behavior feature space, and perform abnormal behavior detection based on the multi-dimensional behavior feature space to obtain vehicle abnormal behavior pattern features; wherein, the vehicle abnormal behavior pattern features include abnormal occurrence probability, abnormal degree score, and abnormal behavior type.

[0024] Furthermore, perform risk prediction on the electric scooter based on the vehicle abnormal behavior pattern features to obtain a risk level classification result, including:

[0025] Perform structured decomposition on the vehicle abnormal behavior pattern features to obtain a sequence of behavior feature components, and perform dynamic evolution analysis on the sequence of behavior feature components through a recursive quantization analysis method to obtain a behavior evolution trajectory graph, and perform fractal dimension calculation on the behavior evolution trajectory graph to obtain a set of fault evolution features; wherein, the set of fault evolution features includes component wear trend, fault diffusion rate, and system degradation degree;

[0026] Perform causal chain analysis on the set of fault evolution features based on preset historical fault data to obtain a fault propagation network, and perform topological feature extraction on the fault propagation network to obtain a set of key fault nodes;

[0027] Perform probabilistic inference on the set of key fault nodes through a preset Bayesian network algorithm to obtain component reliability evaluation values, and perform fuzzy comprehensive evaluation on the electric scooter based on the component reliability evaluation values to obtain a health status score table; wherein, the health status score table includes motor health, battery life value, and bearing wear degree;

[0028] Perform operating risk analysis on the electric scooter based on the health status score table to obtain a risk level classification result.

[0029] Furthermore, perform distributed processing and priority sorting on the risk level classification result through an edge computing gateway to obtain a control strategy instruction set, including:

[0030] Perform data structure parsing on the risk level classification result to obtain a risk feature sequence, and perform multi-level threshold quantization on the risk feature sequence to obtain a hierarchical risk numerical matrix; wherein, the hierarchical risk numerical matrix includes motor control risk value, battery safety threshold, and bearing loss index;

[0031] Perform distributed parallel decoupling on the hierarchical risk numerical matrix to obtain a risk priority mapping table, and perform network topology adaptive allocation based on the risk priority mapping table to obtain a task scheduling sequence; wherein, the task scheduling sequence includes an emergency task queue, a regular task queue, and a low-priority task queue;

[0032] Optimize resource allocation for the task scheduling sequence through an edge computing gateway to obtain an edge node allocation plan, and perform load balancing adjustment on the edge node allocation plan to obtain an edge computing optimization strategy;

[0033] Perform real-time adjustment mapping on preset scooter control parameters based on the edge computing optimization strategy to obtain a control strategy instruction set; wherein, the control strategy instruction set includes speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies.

[0034] Furthermore, input the control strategy instruction set into the remote motor controller to control the electric scooter, including:

[0035] Perform instruction decoding and restoration on the control strategy instruction set to obtain a control parameter sequence, and perform parameter boundary verification on the control parameter sequence to obtain a safety control threshold matrix; wherein, the safety control threshold matrix includes maximum speed limit, maximum steering angle, maximum power output, and maximum braking force;

[0036] Parallelly segment and map the safety control threshold matrix through a multi-threaded channel to obtain a multi-channel control signal stream, and perform real-time timing synchronization on the multi-channel control signal stream to obtain a synchronized control instruction sequence;

[0037] Perform waveform modulation transformation on the synchronized control instruction sequence to obtain a cluster of drive waveforms, and perform pulse width encoding on the cluster of drive waveforms to obtain a motor drive signal chain;

[0038] Perform hardware timing analysis on the motor controller based on the motor drive signal chain to obtain a controller execution timing table, and perform parallel task allocation on the controller execution timing table to obtain a multi-core execution instruction group;

[0039] Perform hardware drive mapping on the multi-core execution instruction group to obtain a motor control drive sequence, and perform real-time motion control on the electric scooter based on the motor control drive sequence.

[0040] Further, the performing pulse width encoding on the cluster of drive waveforms to obtain a motor drive signal chain includes:

[0041] Perform waveform feature decomposition on the cluster of drive waveforms to obtain a fundamental wave parameter matrix, and perform multiple sampling quantization on the fundamental wave parameter matrix to obtain a waveform sampling sequence;

[0042] Perform carrier signal synthesis based on the waveform sampling sequence to obtain a set of carrier modulation parameters, and perform phase compensation calibration on the set of carrier modulation parameters to obtain a phase correction matrix;

[0043] Perform duty cycle modulation calculation on the phase correction matrix to obtain a PWM modulation sequence, and perform dead time compensation on the PWM modulation sequence to obtain a compensation control vector;

[0044] Perform edge detection synchronization on the compensation control vector through a preset synchronization trigger to obtain a set of trigger timings, perform multi-level interleaved encoding on the set of trigger timings to obtain a drive waveform sequence, and perform output parameter mapping on the drive waveform sequence to obtain a motor drive signal chain; wherein, the motor drive signal chain includes a PWM waveform, a phase trigger point, a duty cycle parameter, and a frequency modulation value.

[0045] The present invention also provides an Internet of Things-based remote monitoring system for an electric scooter. A motor controller is provided inside the electric scooter, including:

[0046] An acquisition module, configured to collect data from multi-source sensors carried by the electric scooter to obtain an original sensing data stream; wherein, the original sensing data stream includes a gyroscope sensor, an acceleration sensor, a GPS positioning sensor, and a battery management sensor;

[0047] An extraction module, configured to perform multi-dimensional feature extraction and data fusion on the original sensing data stream to obtain a scooter state feature vector;

[0048] A mapping module, configured to perform multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain vehicle abnormal behavior pattern features;

[0049] A prediction module, configured to perform risk prediction on the electric scooter based on the vehicle abnormal behavior pattern features to obtain a risk level classification result;

[0050] A sorting module, configured to perform distributed processing and priority sorting on the risk level classification result through an edge computing gateway to obtain a control strategy instruction set;

[0051] A control module, configured to input the control strategy instruction set into the remote motor controller to control the electric scooter.

[0052] The present invention further provides a computer device, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0053] The present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0054] A method for remote monitoring of an electric scooter based on the Internet of Things provided by the present invention includes the following steps: collecting data from multi-source sensors carried by the electric scooter to obtain an original sensing data stream; performing multi-dimensional feature extraction and data fusion on the original sensing data stream to obtain a scooter state feature vector; if the scooter state feature vector indicates that the electric scooter has an abnormality, performing multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain a vehicle abnormal behavior pattern feature; predicting the risk of the electric scooter based on the vehicle abnormal behavior pattern feature to obtain a risk level classification result; performing distributed processing and priority sorting on the risk level classification result through an edge computing gateway to obtain a control strategy instruction set; inputting the control strategy instruction set into the remote motor controller to control the electric scooter. By the above technical means, the technical problem that the traditional management method mainly relies on manual inspection and regular maintenance, which is not only inefficient but also difficult to monitor the vehicle state in real time and timely discover potential safety hazards, is solved. The distributed processing and priority sorting are realized by using the edge computing gateway, ensuring that the control strategy can be quickly responded to and executed even in the case of poor network conditions. This greatly improves the stability and reliability of the system, reduces the latency at the same time, and enhances the user experience. Description of the Drawings

[0055] Figure 1 is a schematic diagram of the steps of a method for remote monitoring of an electric scooter based on the Internet of Things in an embodiment of the present invention;

[0056] Figure 2 is a block diagram of the structure of a remote monitoring system for an electric scooter based on the Internet of Things in an embodiment of the present invention;

[0057] Figure 3 is a schematic block diagram of the structure of a computer device in an embodiment of the present invention.

[0058] The implementation, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0059] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0060] As Figure 1 shown, Figure 1 is a schematic diagram of the steps of a method for remote monitoring of an electric scooter based on the Internet of Things in an embodiment of the present invention;

[0061] In an embodiment of the present invention, a remote monitoring method for an electric scooter based on the Internet of Things is provided. An electric scooter is provided with a motor controller, and the method includes the following steps:

[0062] Step S1, collect data from multiple sensors mounted on the electric scooter to obtain an original sensor data stream; wherein, the original sensor data stream includes a gyroscope sensor, an acceleration sensor, a GPS positioning sensor, and a battery management sensor.

[0063] Specifically, in the remote monitoring method for an electric scooter based on the Internet of Things, collecting data from multiple sensors mounted on the electric scooter is one of the key steps to achieve comprehensive monitoring. Specifically, this process involves obtaining the original sensor data stream from the gyroscope sensor, the acceleration sensor, the GPS positioning sensor, and the battery management sensor. First of all, the gyroscope sensor can detect the angular velocity change of the scooter, thereby helping to determine its attitude and direction; the acceleration sensor can measure the linear acceleration of the scooter in three dimensions, which is crucial for understanding the dynamic behavior of the vehicle. At the same time, the GPS positioning sensor provides accurate position information, enabling operators to track the specific location of each scooter in real time, which is particularly important in the sharing economy model as it helps to optimize scheduling and manage resources. In addition, the battery management sensor monitors the state of the battery, including parameters such as voltage, current, and temperature, to ensure the safe use of the battery and extend its life. To better illustrate this process, we can envision an application scenario: in a busy urban environment, a shared scooter company needs to ensure that all scooters put on the market can operate efficiently and safely. Through the collaborative work of the above sensors, whenever a user starts a scooter, the system begins to collect data from these sensors. For example, when the scooter is turning, the gyroscope and acceleration sensors will record the corresponding angular velocity and acceleration changes, while the GPS positioning sensor continuously updates the position information of the scooter so that the background management system can monitor its driving path in real time. At the same time, the battery management sensor continuously monitors the battery state. Once it detects that the battery power is too low or the temperature rises abnormally, the system will immediately notify the maintenance team to take corresponding measures. In this way, not only the user experience is improved, but also the safety is enhanced and potential risks are reduced. Throughout the process, the rich data stream provided by the multiple sensors lays a solid foundation for subsequent data processing and analysis, making the state evaluation of the scooter more accurate and the management more efficient.

[0064] Step S2, perform multi-dimensional feature extraction and data fusion on the original sensor data stream to obtain a scooter state feature vector.

[0065] Specifically, in the method for remotely monitoring an electric scooter based on the Internet of Things, multi-dimensional feature extraction and data fusion of the original sensing data stream to generate a scooter state feature vector is an important step in achieving precise monitoring. This process first requires extracting key features from the data collected by gyro sensors, acceleration sensors, GPS positioning sensors, and battery management sensors. For example, extract the change trend of angular velocity from the gyro sensor, the change of linear acceleration from the acceleration sensor, obtain the speed and direction of position change from the GPS positioning sensor, and monitor parameters such as voltage, current, and temperature from the battery management sensor. Next, these different types of original data are fused through complex data processing algorithms, aiming to convert them into a feature vector that can comprehensively describe the current state of the scooter. To understand this process more clearly, we can envision an application scenario: in a shared scooter service system, when a user rides a scooter, each sensor continuously records its operating state. Suppose a scooter suddenly experiences abnormal vibration during driving. At this time, the gyro sensor will capture abnormal angular velocity fluctuations, and the acceleration sensor will record abnormal acceleration changes. Meanwhile, the GPS positioning sensor provides the vehicle's position information to help determine the specific location where the problem occurs, while the battery management sensor ensures the stability of the battery state to avoid problems caused by insufficient power or overheating. All the data provided by these sensors is sent to the data fusion module, where advanced algorithms such as machine learning models analyze and integrate this data to generate a state feature vector containing multi-dimensional information. This feature vector not only reflects the current physical state of the scooter (such as whether it is driving smoothly or whether there are mechanical failures), but also includes important information such as environmental factors (such as location and speed) and battery health status. This comprehensive state assessment provides a solid foundation for subsequent risk prediction and control strategy formulation, enabling operators to more accurately judge the operating condition of the scooter and take corresponding measures to ensure user safety and service quality. In this way, not only the intelligent level of the system is improved, but also the user experience and safety are enhanced.

[0066] Step S3, if the scooter state feature vector indicates that the electric scooter is abnormal, then perform multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain vehicle abnormal behavior pattern features.

[0067] Specifically, in the remote monitoring method of electric scooters based on the Internet of Things, once the scooter state feature vector obtained by multi-dimensional feature extraction and data fusion shows an abnormal situation, the system will perform multimodal feature mapping on the electric scooter based on this feature vector to identify the abnormal behavior pattern characteristics of the vehicle. Specifically, this process first requires analyzing the various indicators in the state feature vector, including data provided by the gyroscope sensor, accelerometer sensor, GPS positioning sensor and battery management sensor. For example, if the state feature vector shows significant fluctuations in angular velocity and acceleration, accompanied by irregularities in position changes or abnormal battery parameters, this may mean that the scooter has a mechanical failure or is experiencing an abnormal usage scenario. In this case, the system will further map these features using a pre-trained model to convert different types of sensor data into unified abnormal behavior pattern characteristics. To better understand this process, we can imagine an application scenario: in a shared scooter operation environment, a scooter suddenly shakes violently and deviates from the scheduled route during driving. At this time, the gyroscope sensor records an abnormally high angular velocity value, the accelerometer detects frequent acceleration changes, the GPS positioning sensor shows an irregular position movement trajectory, and the battery management sensor reports an increase in battery temperature. All this information is integrated into the state feature vector of the scooter and passed to the multimodal feature mapping module. This module uses machine learning algorithms to conduct in-depth analysis of these features, trying to find the root cause of these abnormal phenomena. For example, the system may find that this abnormal behavior pattern matches a known mechanical failure (such as tire damage) or an extreme usage situation (such as high-speed sharp turns). In this way, the system can not only accurately identify specific abnormal behavior patterns, but also provide key basis for subsequent risk prediction and control strategies. For example, after detecting the above anomalies, the system can immediately notify the operation team for inspection and repair, thereby avoiding potential safety hazards and ensuring the riding safety and service quality of users. In addition, this intelligent monitoring mechanism can also help optimize maintenance plans, reduce unnecessary downtime, and improve resource utilization. By combining data from multi-source sensors and advanced data analysis technology, accurate assessment and timely response to the health status of electric scooters are achieved.

[0068] Step S4: predicting the risk of the electric scooter based on the abnormal behavior pattern characteristics of the vehicle to obtain a risk level classification result.

[0069] Specifically, in the method for remote monitoring of electric scooters based on the Internet of Things, once the abnormal behavior pattern features of the vehicle are obtained through multi-modal feature mapping, the system will perform risk prediction on the electric scooter based on these features to obtain the risk level classification result. Specifically, this process first requires using a machine learning model or statistical analysis method to evaluate the impact degree of the abnormal behavior pattern features. For example, if the state feature vector shows significant fluctuations in the data of the acceleration sensor and the gyroscope sensor, at the same time, the GPS positioning sensor records irregular position changes, and the battery management sensor reports an abnormal increase in battery temperature, then this information will be input into a pre-trained risk prediction model. The model will calculate the probability of a possible failure or accident under the current abnormal behavior pattern according to historical data and preset risk thresholds. To understand this process more clearly, we can imagine an application scenario: in a shared scooter service system, a certain scooter shows abnormal vibrations, frequent sharp turns, and overheating of the battery during driving. At this time, the system has identified these abnormal behavior pattern features through multi-modal feature mapping and transmitted them to the risk prediction module. In this module, the system uses historical data and known risk factors (such as mechanical failures, overuse, or environmental impacts) to evaluate the potential risks in the current situation. Suppose the system detects that this abnormal behavior pattern is highly similar to high-risk events in previous records, such as loss of control caused by tire damage or safety hazards caused by overheating of the battery, it will immediately activate the risk prediction algorithm to calculate the corresponding risk level. For example, the system may divide the risk into three levels: low, medium, and high according to the severity of the abnormal behavior pattern. If the prediction result shows that the current scooter is in a high-risk state, the system will generate a high-risk level result and trigger the corresponding alarm mechanism. Further, based on the risk level classification result, the system can take different countermeasures. For example, in the case of a high-risk level, the system will immediately notify the operation team for emergency inspection and repair; while in the case of a medium-risk level, the system may recommend that the user slow down and end the journey as soon as possible. This intelligent risk prediction mechanism can not only detect potential safety hazards in a timely manner but also provide a scientific basis for operators to formulate reasonable maintenance plans, reduce the occurrence of accidents, and improve user experience and safety. By combining advanced data analysis techniques and real-time monitoring means, the accurate evaluation and effective management of the operating state of electric scooters are realized. In this way, not only can the riding safety of users be guaranteed, but also the resource utilization efficiency can be improved, and the operation cost can be reduced.

[0070] Step S5: Perform distributed processing and priority sorting on the risk level classification result through the edge computing gateway to obtain a control strategy instruction set.

[0071] Specifically, in the remote monitoring method of electric scooters based on the Internet of Things, the distributed processing and prioritization of the risk level classification results by the edge computing gateway to generate a control strategy instruction set is a key step to ensure efficient response and optimize resource management. Specifically, when the system completes risk prediction based on the abnormal behavior pattern characteristics of the vehicle and obtains the risk level classification results, these results will be sent to the edge computing gateway. As a processing node close to the data source, the edge computing gateway can quickly analyze and process a large amount of data without uploading all the data to the cloud, thus reducing latency and improving the real-time performance of the system. In this process, the edge computing gateway will first perform distributed processing on the risk level classification results of each scooter, which means it will process the data of multiple scooters simultaneously and classify them according to their risk levels. For example, in a shared scooter service system, assume that multiple scooters report different risk levels at the same time: some vehicles are in a high-risk state (such as battery overheating or mechanical failure), while other vehicles are in a medium or low-risk state (such as minor sensor anomalies). The edge computing gateway will first identify those scooters in a high-risk state and prioritize them. This sorting mechanism can determine which vehicles need to take immediate action based on the severity and urgency of the risk. For high-risk scooters, the system may generate an emergency stop instruction or recommend that the user end the trip immediately; for medium-risk scooters, it may generate an instruction to slow down or check the device status. All these control strategy instruction sets will be integrated together to form specific operation guidelines for each scooter. To better understand this process, we can imagine such an application scenario: in a busy urban environment, a shared scooter company operates thousands of scooters. One day, due to bad weather and frequent use, many scooters showed varying degrees of anomalies. After receiving the risk level classification results of these vehicles, the edge computing gateway quickly activates the distributed processing mechanism to evaluate the status of each vehicle. For those scooters with severe battery overheating or mechanical failure, the system immediately generates an emergency shutdown instruction and sends it to the corresponding scooter through the network to ensure user safety. At the same time, for those scooters with only minor anomalies (such as unstable GPS signals), the system generates an instruction to remind the user to pay attention and recommend parking nearby for inspection. In this way, the edge computing gateway can not only process a large amount of data in a short time but also dynamically adjust the control strategy according to the actual situation to ensure that each scooter can be managed in a timely and effective manner. This efficient distributed processing and prioritization mechanism significantly improves the response speed and reliability of the system, while also optimizing resource allocation, reducing operating costs, and enhancing user experience and safety.

[0072] Step S6, input the control strategy instruction set into the remote motor controller to control the electric scooter.

[0073] Specifically, in the remote monitoring method of electric scooters based on the Internet of Things, inputting the control strategy instruction set into the remote motor controller to control the electric scooter is a key step in realizing intelligent management and ensuring user safety. Specifically, when the edge computing gateway completes the distributed processing and priority sorting of the risk level division results, the generated control strategy instruction set will be sent to the motor controller of each scooter. This process relies on a stable and reliable communication network to ensure that the instructions can be transmitted to the target vehicle in a timely and accurate manner. Once the motor controller receives these instructions, it will adjust the operation mode of the scooter according to the instruction content, thereby achieving real-time control of the vehicle. For example, in a shared scooter service system, assume that a certain scooter is identified as a high-risk state by the system due to overheating of the battery. The edge computing gateway generates an emergency shutdown instruction as part of the control strategy. This instruction is then transmitted to the motor controller of this scooter through the wireless communication network. After receiving the instruction, the motor controller immediately executes the corresponding operations, such as gradually decelerating and finally stopping the operation of the scooter, and at the same time sending a warning message to the user, indicating that there is a safety hazard in the current device and suggesting that the user end the trip immediately and contact the customer service for inspection. For scooters in a medium-risk state, such as detecting minor mechanical failures or sensor abnormalities, the system may generate instructions to slow down or suggest that the user park nearby for inspection. After receiving such instructions, the motor controller will adjust the speed of the scooter accordingly and notify the user to take appropriate measures through the on-board display or mobile application. To further illustrate this process, we can imagine such an application scenario: in a busy urban environment, a shared scooter company operates thousands of scooters. During the use of one scooter, a serious mechanical failure suddenly occurs, resulting in abnormal fluctuations in the data of the acceleration sensor and gyroscope sensor. The edge computing gateway quickly analyzes this data, determines that the scooter is in a high-risk state, and generates an emergency shutdown instruction. This instruction is quickly transmitted to the motor controller of the scooter through the wireless communication network, and the motor controller immediately responds by first gradually reducing the speed of the scooter and then smoothly stopping it. At the same time, the system sends an alarm message through the user's mobile application, informing the user that there is a safety hazard in the current device and suggesting that the user contact the customer service as soon as possible for follow-up processing. This instant response mechanism not only effectively avoids potential safety accidents, but also improves the user experience and sense of security. In addition, for scooters with only minor abnormalities (such as unstable GPS signals), the system generates instructions to remind the user to pay attention and suggest parking nearby for inspection. The motor controller then adjusts the vehicle's state according to the instructions to ensure that the user can continue to use the scooter safely. In this way, the entire system realizes a closed-loop management from data collection, risk assessment to control strategy execution, significantly improving the safety and reliability of electric scooters. This efficient control mechanism not only optimizes resource utilization, but also reduces operating costs and enhances the overall service quality.

[0074] In a specific embodiment, the multi-dimensional feature extraction and data fusion of the original sensing data stream to obtain a scooter state feature vector includes:

[0075] Performing time-frequency domain decomposition on the original sensing data stream to obtain a vibration feature sequence of the electric scooter, and performing wavelet packet decomposition on the vibration feature sequence to obtain a multi-scale vibration spectrogram;

[0076] Based on the multi-scale vibration spectrogram, performing data analysis on the attitude of the electric scooter to obtain an attitude feature set, performing time series alignment on the attitude feature set to obtain an aligned state sequence, and performing sparse coding on the aligned state sequence to obtain a compressed feature representation;

[0077] Performing multi-modal feature fusion on the compressed feature representation to obtain a fusion feature matrix, and performing tensor decomposition on the fusion feature matrix to obtain a scooter state feature vector.

[0078] Specifically, in the remote monitoring method of electric scooters based on the Internet of Things, extracting multi-dimensional features and fusing data from the original sensor data stream to generate a scooter state feature vector is a complex and crucial process. This process involves multiple steps, each of which involves complex signal processing and data analysis techniques to ensure a comprehensive and accurate description of the scooter's state. First, the original sensor data stream obtained from gyroscope sensors, acceleration sensors, GPS positioning sensors, and battery management sensors needs to be decomposed in the time-frequency domain to obtain the vibration feature sequence of the electric scooter, and further perform wavelet packet decomposition on these vibration feature sequences to generate a multi-scale vibration spectrogram. Specifically, time-frequency domain decomposition converts the original sensor data into a data representation in two dimensions, time and frequency, enabling the simultaneous analysis of vibration characteristics in different time periods. For example, in a shared scooter service system, assume that a certain scooter experiences abnormal vibrations during driving. Through time-frequency domain decomposition, the specific time and frequency distribution of this vibration can be captured. Then, using wavelet packet decomposition technology, these vibration features can be further refined into a multi-scale vibration spectrogram, providing more detailed frequency component information to help identify potential problems. Next, based on the multi-scale vibration spectrogram, data analysis of the electric scooter's attitude is performed to obtain an attitude feature set. In this process, the attitude feature set not only contains the attitude information of the scooter (such as tilt angle, rotation speed, etc.), but also combines position and motion data provided by other sensors. To ensure the consistency and comparability of these data, it is necessary to perform temporal alignment on the attitude feature set, that is, synchronize the timestamps of different sensors to generate an aligned state sequence. This step is crucial for subsequent data processing because it ensures that all sensor data can be compared and analyzed under the same time reference. For example, when the scooter is turning or accelerating and decelerating, the aligned state sequence can accurately capture its attitude changes, thus better evaluating the vehicle's operating state. Then, sparse coding is performed on the aligned state sequence to obtain a compressed feature representation. Sparse coding is a data compression technique that achieves efficient data representation by finding a small number of basis vectors that best represent the original data. This method can not only reduce the amount of data, but also retain key information and improve the efficiency of subsequent processing. For example, in the above application scenario, by performing sparse coding on the aligned state sequence, a large amount of original sensor data can be compressed into a compact feature representation, facilitating fast transmission and processing. This compressed feature representation not only reduces the burden of data storage and transmission, but also improves the real-time response ability of the system. Subsequently, multi-modal feature fusion is performed on the compressed feature representation to generate a fused feature matrix. Multi-modal feature fusion aims to integrate data from different sensors to form a unified feature representation. In this process, different types of sensor data (such as acceleration, angular velocity, position, etc.) are combined to form a multi-dimensional feature matrix.For example, in the application scenario of shared scooters, by fusing the data of the accelerometer, gyroscope sensor, and GPS positioning sensor, a feature matrix that comprehensively reflects the dynamic behavior of the scooter can be generated. This fused feature matrix can provide more comprehensive information and help to more accurately identify the status and potential problems of the scooter. Finally, the fused feature matrix is ​​tensor decomposed to obtain the scooter state feature vector. Tensor decomposition is an advanced data analysis method that can decompose high-dimensional data into multiple low-dimensional subspaces to reveal hidden structures and patterns in the data. In this way, the most representative feature vector can be extracted from the fused feature matrix for subsequent risk prediction and control strategy formulation. For example, in the above application scenario, by performing tensor decomposition on the fused feature matrix, a feature vector that can accurately describe the current state of the scooter can be obtained. This feature vector not only contains the posture, position, and motion information of the scooter, but also reflects its health status and potential risks. Based on this feature vector, the system can perform accurate risk assessment and early warning, and take corresponding control measures in time to ensure the safety of users and service quality. To better understand the whole process, we can imagine a specific application scenario: In a busy urban environment, a shared scooter company operates thousands of scooters. One day, due to bad weather and frequent use, many scooters experienced different degrees of abnormal conditions. Assume that one of the scooters suddenly shook violently and deviated from the scheduled route during driving. At this time, the system first collects the original sensor data stream from each sensor and generates a multi-scale vibration spectrum through time-frequency domain decomposition and wavelet packet decomposition. Then, posture data analysis is performed based on these spectrum graphs to obtain a posture feature set, and an aligned state sequence is generated through time alignment. Sparse coding is performed on the aligned state sequence to generate a compressed feature representation, thereby reducing the amount of data and retaining key information. Subsequently, the compressed feature representation is fused with other sensor data for multimodal features to generate a fused feature matrix. Finally, the fused feature matrix is ​​tensor-decomposed to generate a scooter state feature vector. Based on this feature vector, the system can accurately identify the abnormal conditions of the scooter and take corresponding control measures according to the risk level classification results, such as emergency shutdown or suggesting that users slow down to ensure user safety and normal operation of the equipment. Through this multi-level data processing and analysis mechanism, comprehensive monitoring and intelligent management of the status of electric scooters are achieved, significantly improving the reliability of the system and user experience.

[0079] In a specific embodiment, the multimodal feature mapping of the electric scooter based on the scooter state feature vector to obtain the abnormal behavior pattern feature of the vehicle includes:

[0080] Perform multi-dimensional space projection on the state feature vector of the scooter to obtain a feature distribution mapping atlas, and perform non-linear manifold embedding on the feature distribution mapping atlas to obtain a topological feature matrix;

[0081] Perform similarity measurement on the topological feature matrix through a preset local sensitive hashing technique to obtain a feature similarity spectrum, and perform spectral clustering analysis on the feature similarity spectrum to obtain a set of behavior pattern prototypes; among them, the set of behavior pattern prototypes includes an acceleration pattern, a steering pattern, and a braking pattern;

[0082] Extract spatio-temporal features from the set of behavior pattern prototypes to obtain a spatio-temporal feature tensor, and perform tensor decomposition on the spatio-temporal feature tensor to obtain multi-modal behavior features;

[0083] Perform time series dependence analysis on the multi-modal behavior features through a preset conditional random field technique to obtain a state transition sequence, and perform Markov jump analysis on the state transition sequence to obtain a behavior pattern transition graph; among them, the behavior pattern transition graph includes state duration, state transition probability, and state stability;

[0084] Perform distributed association mining on the behavior pattern transition graph based on preset historical behavior data to obtain behavior pattern association rules, and perform probabilistic graph reasoning on the behavior pattern association rules to obtain a behavior prediction vector; among them, the behavior prediction vector includes behavior occurrence probability, behavior duration, and behavior transition trend;

[0085] Perform multi-scale feature fusion on the behavior prediction vector to obtain a multi-dimensional behavior feature space, and perform abnormal behavior detection based on the multi-dimensional behavior feature space to obtain vehicle abnormal behavior pattern features; among them, the vehicle abnormal behavior pattern features include abnormal occurrence probability, abnormal degree score, and abnormal behavior type.

[0086] Specifically, in the method for remotely monitoring electric scooters based on the Internet of Things, performing multi-modal feature mapping on electric scooters based on the scooter state feature vector to obtain vehicle abnormal behavior pattern features is a complex and multi-level process. First, the system performs multi-dimensional space projection on the scooter state feature vector to generate a feature distribution mapping atlas, and further performs non-linear manifold embedding on these atlases to obtain a topological feature matrix. Specifically, multi-dimensional space projection converts high-dimensional feature vectors into representations in a low-dimensional space, enabling more intuitive observation and analysis of the distribution characteristics of data. For example, in a shared scooter service system, assume that a certain scooter experiences abnormal vibration during driving. Through multi-dimensional space projection, the state feature vector of this scooter can be mapped into a two-dimensional or three-dimensional space to form a feature distribution mapping atlas. Then, using non-linear manifold embedding techniques (such as t-SNE or LLE), these features can be further embedded into a topological structure to reveal hidden data relationships and generate a topological feature matrix. Next, the system performs similarity measurement on the topological feature matrix through a preset locality-sensitive hashing (LSH) technique to obtain a feature similarity spectrum, and performs spectral clustering analysis on the feature similarity spectrum to generate a set of behavior pattern prototypes. In this process, the locality-sensitive hashing technique can efficiently calculate the similarity between different features, while spectral clustering analysis can group samples with similar features to form different behavior pattern prototypes. For example, in the above application scenario, the similarity between the topological feature matrices of each vehicle is calculated through the locality-sensitive hashing technique, and they are divided into several clusters using the spectral clustering algorithm. Each cluster represents a typical behavior pattern, such as an acceleration pattern, a steering pattern, and a braking pattern. This classification not only helps to identify normal driving behaviors but also helps to discover potential abnormal behavior patterns. Subsequently, the system extracts spatio-temporal features from the set of behavior pattern prototypes to generate a spatio-temporal feature tensor, and performs tensor decomposition on it to obtain multi-modal behavior features. Spatio-temporal feature extraction aims to capture the dynamic changes of behavior patterns in time and space, while tensor decomposition can decompose high-dimensional spatio-temporal features into multiple low-dimensional subspaces to reveal the key information in the data. For example, in the above application scenario, by extracting spatio-temporal features from the acceleration pattern, steering pattern, and braking pattern, a spatio-temporal feature tensor containing time series and spatial position information can be generated. Then, through tensor decomposition techniques, this high-dimensional tensor can be decomposed into multiple low-dimensional features to form multi-modal behavior features, which better describe the behavior patterns of the scooter. Then, the system performs temporal dependence analysis on the multi-modal behavior features through a preset conditional random field (CRF) technique to generate a state transition sequence, and performs Markov jump analysis on the state transition sequence to obtain a behavior pattern transition graph. The conditional random field technique can effectively model the dependence relationships in sequence data, while Markov jump analysis can reveal the transition probabilities and durations between states.For example, in the above application scenario, the temporal dependence of multimodal behavior characteristics is analyzed through conditional random field technology to generate a state transition sequence, such as the transition process from the acceleration mode to the steering mode and then to the braking mode. Next, through Markov jump analysis, the duration of each state, the state transition probability, and the state stability can be calculated to form a behavior pattern transition graph. Subsequently, the system performs distributed association mining on the behavior pattern transition graph based on preset historical behavior data to generate behavior pattern association rules, and performs probabilistic graphical inference on the behavior pattern association rules to obtain a behavior prediction vector. Distributed association mining can discover potential behavior pattern association rules from a large amount of historical data, while probabilistic graphical inference can quantify the occurrence probability and trend of these rules. For example, in the above application scenario, through distributed association mining of historical behavior data, the association between certain specific behavior pattern combinations (such as frequent hard acceleration and hard braking) and the failure rate can be discovered. Then, through probabilistic graphical inference technology, the probability of occurrence, the duration, and the transition trend of these behavior patterns in the future can be calculated to form a behavior prediction vector. Finally, the system performs multi-scale feature fusion on the behavior prediction vector to generate a multi-dimensional behavior feature space, and performs abnormal behavior detection based on the multi-dimensional behavior feature space to generate vehicle abnormal behavior pattern features. Multi-scale feature fusion can integrate feature information at different levels, while abnormal behavior detection can identify behaviors that do not conform to the normal pattern. For example, in the above application scenario, through multi-scale feature fusion of the behavior prediction vector, a multi-dimensional behavior feature space can be generated, which contains feature representations of various behavior patterns. Then, through abnormal behavior detection algorithms, behaviors that deviate from the normal pattern, such as abnormal acceleration fluctuations or irregular steering behaviors, can be identified, and vehicle abnormal behavior pattern features, including the probability of occurrence of the abnormality, the abnormality degree score, and the type of abnormal behavior, can be generated. Through this multi-level data processing and analysis mechanism, the comprehensive monitoring and intelligent management of the electric scooter state are realized, significantly improving the reliability of the system and the user experience. This detailed analysis can not only timely discover potential safety hazards but also provide a scientific basis for subsequent risk prediction and control strategies to ensure user safety and service quality.

[0087] In a specific embodiment, the risk prediction of the electric scooter based on the vehicle abnormal behavior pattern features to obtain a risk level classification result includes:

[0088] Structurally decompose the vehicle abnormal behavior pattern features to obtain a sequence of behavior feature components, and perform dynamic evolution analysis on the sequence of behavior feature components through the recursive quantification analysis method to obtain a behavior evolution trajectory graph, and calculate the fractal dimension of the behavior evolution trajectory graph to obtain a fault evolution feature set; wherein, the fault evolution feature set includes component wear trend, fault diffusion rate, and system degradation degree;

[0089] Perform a causal chain analysis on the fault evolution feature set based on preset historical fault data to obtain a fault propagation network, and extract topological features of the fault propagation network to obtain a set of key fault nodes;

[0090] Perform probability inference on the set of key fault nodes through a preset Bayesian network algorithm to obtain component reliability evaluation values, and perform a fuzzy comprehensive evaluation on the electric scooter based on the component reliability evaluation values to obtain a health status score table; wherein, the health status score table includes motor health, battery life value, and bearing wear degree;

[0091] Perform an operating risk analysis on the electric scooter based on the health status score table to obtain a risk level classification result.

[0092] Specifically, in the remote monitoring method of electric scooters based on the Internet of Things, predicting the risk of electric scooters based on the characteristics of abnormal vehicle behavior patterns to obtain the risk level classification result is a multi-level and complex process. First, the system will perform a structured decomposition of the characteristics of abnormal vehicle behavior patterns, generate a sequence of behavioral characteristic components, and conduct a dynamic evolution analysis of these sequences through the recursive quantification analysis method to obtain a behavioral evolution trajectory graph. Specifically, the structured decomposition breaks down the complex abnormal behavior pattern characteristics into multiple independent behavioral characteristic components, enabling a more detailed analysis of the changing trends of each component. For example, in a shared scooter service system, assume that a certain scooter exhibits frequent rapid acceleration and sudden braking during driving. Through structured decomposition, these abnormal behavior pattern characteristics can be broken down into independent behavioral characteristic component sequences such as acceleration change and steering angle change. Then, the recursive quantification analysis method (such as recurrence plots or recurrence quantification analysis) is used to conduct a dynamic evolution analysis of these sequences of behavioral characteristic components to generate a behavioral evolution trajectory graph. This trajectory graph can not only show the changes in behavioral characteristics over time but also reveal potential trends and patterns. Next, the system calculates the fractal dimension of the behavioral evolution trajectory graph to obtain a set of fault evolution characteristics. Fractal dimension calculation is a mathematical tool used to describe the dynamic characteristics of complex systems and can quantify the self-similarity and irregularity of the system. In this process, by calculating the fractal dimension of the behavioral evolution trajectory graph, key characteristics reflecting the degree of system degradation can be extracted, such as component wear trends, fault propagation rates, and system degradation levels. For example, in the above application scenario, it can be found through fractal dimension calculation that certain key components of the scooter (such as the motor or bearings) show a gradually deteriorating trend, generating a set of fault evolution characteristics. These characteristic sets not only provide detailed fault information but also provide a scientific basis for subsequent risk assessment. Subsequently, the system conducts a causal chain analysis of the set of fault evolution characteristics based on preset historical fault data, generates a fault propagation network, and extracts topological characteristics of the fault propagation network to obtain a set of key fault nodes. Causal chain analysis aims to mine the causal relationships between faults from historical fault data, while topological characteristic extraction can identify the key nodes in the fault propagation network. For example, in the above application scenario, through causal chain analysis, it can be found that there are obvious causal relationships between phenomena such as rapid acceleration and sudden braking and problems such as motor overheating and bearing wear, forming a fault propagation network. Then, by extracting the topological characteristics of this network, key fault nodes that have the greatest impact on system reliability can be identified, such as motor overheating problems or battery life attenuation. Then, the system conducts probability inference on the set of key fault nodes through a preset Bayesian network algorithm to obtain component reliability evaluation values, and based on these evaluation values, conducts a fuzzy comprehensive evaluation of the electric scooter to generate a health status score table. A Bayesian network is a graphical model based on probability inference that can effectively handle uncertainty and complex dependencies.In this process, by performing probabilistic reasoning on critical fault nodes, the reliability evaluation values of each component can be calculated, such as motor health, battery life, and bearing wear. For example, in the above application scenario, through the Bayesian network algorithm, it can be obtained that the current health of the motor is 80%, the battery life is 75%, and the bearing wear is 60%. Then, using the fuzzy comprehensive evaluation method, these specific values are converted into an overall health status score table, comprehensively reflecting the operating conditions of the scooter. Finally, the system conducts an operating risk analysis on the electric scooter based on the health status score table and generates a risk level classification result. The operating risk analysis aims to evaluate the likelihood of future failures of the scooter based on its overall health status and classify it into different risk levels. For example, in the above application scenario, based on the health status score table (motor health 80%, battery life 75%, bearing wear 60%), the system can assess that the probability of a major failure of this scooter in the future is relatively high, so its risk level is classified as high risk. On the contrary, if the health status score table shows that the status of all key components is at a good level, its risk level may be classified as low risk. In this way, the system can not only identify potential safety hazards in a timely manner but also provide scientific maintenance suggestions for operators to ensure the safety and reliability of the scooter. To better understand the entire process, we can envision a specific application scenario: in a busy urban environment, a shared scooter company operates thousands of scooters. One day, due to bad weather and frequent use, many scooters showed varying degrees of abnormalities. Suppose one of the scooters suddenly exhibited frequent rapid acceleration and rapid braking during operation. The system first structurally decomposes the abnormal behavior pattern characteristics of the scooter to generate a sequence of behavior feature components. Then, through the recursive quantification analysis method, dynamic evolution analysis is performed on these sequences to generate a behavior evolution trajectory graph, and its fractal dimension is calculated to obtain a set of fault evolution characteristics. These characteristic sets reveal the wear trends of key components such as the motor, battery, and bearings of the scooter and the degree of system degradation. Subsequently, the system conducts a causal chain analysis on the set of fault evolution characteristics based on historical fault data to generate a fault propagation network, and identifies critical fault nodes through topological feature extraction, such as motor overheating problems and battery life attenuation. Then, probabilistic reasoning is performed on these critical fault nodes through the Bayesian network algorithm to obtain the reliability evaluation values of each component, and an overall health status score table is generated based on these evaluation values to comprehensively reflect the operating conditions of the scooter. Finally, the system conducts an operating risk analysis on the scooter based on the health status score table and generates a risk level classification result to help operators take corresponding maintenance measures in a timely manner to ensure user safety and service quality. Through this multi-level data processing and analysis mechanism, comprehensive monitoring and intelligent management of the state of electric scooters are achieved, significantly improving the reliability of the system and the user experience.This meticulous analysis can not only promptly detect potential security hazards but also provide a scientific basis for subsequent risk prediction and control strategies, ensuring user safety and service quality.

[0093] In a specific embodiment, the distributed processing and priority ranking of the risk level division results are performed through an edge computing gateway to obtain a control strategy instruction set, including:

[0094] Perform data structure parsing on the risk level division results to obtain a risk feature sequence, and perform multi-level threshold quantization on the risk feature sequence to obtain a hierarchical risk value matrix; wherein, the hierarchical risk value matrix includes a motor control risk value, a battery safety threshold, and a bearing loss index;

[0095] Perform distributed parallel decoupling on the hierarchical risk value matrix to obtain a risk priority mapping table, and perform network topology adaptive allocation based on the risk priority mapping table to obtain a task scheduling sequence; wherein, the task scheduling sequence includes an emergency task queue, a regular task queue, and a low-priority task queue;

[0096] Optimize resource allocation for the task scheduling sequence through an edge computing gateway to obtain an edge node allocation plan, and adjust the load balance of the edge node allocation plan to obtain an edge computing optimization strategy;

[0097] Perform real-time adjustment mapping on the preset scooter control parameters based on the edge computing optimization strategy to obtain a control strategy instruction set; wherein, the control strategy instruction set includes a speed limit instruction, a steering angle constraint, battery power control, and an emergency braking strategy.

[0098] Specifically, in the method for remotely monitoring electric scooters based on the Internet of Things, the distributed processing and prioritization of the risk level classification results by the edge computing gateway to generate a control strategy instruction set is a complex and multi-level process. First, the system parses the data structure of the risk level classification results, generates risk feature sequences, and performs multi-level threshold quantization on these sequences to obtain a hierarchical risk value matrix. Specifically, the data structure parsing decomposes the complex and multi-level risk level classification results into multiple independent risk feature components, enabling a more detailed analysis of the specific values of each component. For example, in a shared scooter service system, assume that a certain scooter is evaluated as being in a high-risk state. The system first parses the risk level classification results of this scooter to generate a risk feature sequence containing motor control risk values, battery safety thresholds, and bearing wear indicators. Then, the multi-level threshold quantization method (such as setting different risk threshold intervals) is used to quantify these risk feature sequences to generate a hierarchical risk value matrix. This matrix can not only clearly show the risk levels of each key component but also provide a scientific basis for subsequent task scheduling. Next, the system performs distributed parallel decoupling on the hierarchical risk value matrix to generate a risk priority mapping table and, based on this mapping table, performs network topology adaptive allocation to generate a task scheduling sequence. The purpose of distributed parallel decoupling is to separate each risk feature component in the hierarchical risk value matrix to form an independent risk priority mapping table for subsequent parallel processing and optimization. For example, in the above application scenario, through distributed parallel decoupling of the hierarchical risk value matrix, it is possible to identify which components have the highest risk and map them to the corresponding priority list. Then, based on these priority mapping tables, network topology adaptive allocation is performed to generate a task scheduling sequence including an emergency task queue, a regular task queue, and a low-priority task queue. This scheduling mechanism ensures that high-risk tasks can be processed first, thus minimizing potential safety hazards. Then, the system optimizes the resource allocation of the task scheduling sequence through the edge computing gateway to generate an edge node allocation plan and adjusts the load balancing of this plan to generate an edge computing optimization strategy. The purpose of resource allocation optimization is to reasonably allocate the computing resources of the edge computing gateway according to the requirements of the task scheduling sequence to ensure efficient task processing capabilities. For example, in the above application scenario, the edge computing gateway dynamically adjusts the allocation of its computing resources according to the emergency task queue, regular task queue, and low-priority task queue in the task scheduling sequence to ensure that emergency tasks can be quickly responded to. At the same time, through load balancing adjustment, the workload of the edge nodes is further optimized to avoid the situation where some nodes are overloaded while other nodes are idle. This optimization strategy not only improves the overall efficiency of the system but also enhances the reliability and stability of the system. Finally, based on the edge computing optimization strategy, real-time adjustment mapping is performed on the preset scooter control parameters to generate a control strategy instruction set.Specifically, the control strategy instruction set is a series of control instructions generated in real time by the edge computing gateway according to the current state and risk level of the scooter, which are used to guide the specific operations of the scooter. For example, in the above application scenario, assume that a certain scooter is evaluated as a high-risk state due to overheating of the battery. The edge computing gateway generates a control strategy instruction set including speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies according to the edge computing optimization strategy. These instructions are transmitted to the motor controller of the scooter through the wireless communication network, and the corresponding operations are immediately executed, such as gradually decelerating and finally stopping the operation of the scooter, while sending a warning message to the user, indicating that there are safety hazards in the current device and the user needs to end the journey immediately and contact the customer service for inspection. For scooters in a medium-risk state, such as detecting minor mechanical failures or sensor abnormalities, the system may generate instructions to slow down or suggest that the user park nearby for inspection. After receiving such instructions, the motor controller will adjust the speed of the scooter accordingly and notify the user to take appropriate measures through the on-vehicle display or mobile application. To better understand the whole process, we can imagine a specific application scenario: in a busy urban environment, a shared scooter company operates thousands of scooters. One day, due to bad weather and frequent use, many scooters showed varying degrees of abnormalities. Assume that one of the scooters suddenly had a serious mechanical failure during driving, resulting in abnormal fluctuations in the data of the acceleration sensor and gyroscope sensor. First, the system parses the data structure of the risk level classification result of the scooter, generating a risk feature sequence including motor control risk values, battery safety thresholds, and bearing loss indicators. Then, a hierarchical risk numerical matrix is generated through a multi-level threshold quantization method, revealing a relatively high risk level in the key components of the scooter. Subsequently, the system performs distributed parallel decoupling on the hierarchical risk numerical matrix, generating a risk priority mapping table, and based on this mapping table, performs network topology adaptive allocation to generate a task scheduling sequence. During this process, the system identifies that this scooter belongs to a high-risk state and adds its task to the emergency task queue to ensure priority processing. Then, the edge computing gateway optimizes the resource allocation for the task scheduling sequence, generating an edge node allocation scheme, and through load balancing adjustment, generates an edge computing optimization strategy. In this case, the edge computing gateway allocates more computing resources to the tasks in the emergency task queue to ensure that high-risk scooters can be processed in a timely manner. Finally, based on the edge computing optimization strategy, the system performs real-time adjustment mapping on the preset scooter control parameters, generating a control strategy instruction set. For example, for this high-risk scooter, the system generates a control strategy instruction set including speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies, and sends it to the motor controller of the scooter through the wireless communication network.After receiving the instruction, the motor controller immediately performs corresponding operations, gradually decelerates and finally stops the operation of the scooter. At the same time, it sends an alarm message through the user's mobile application, informing the user that there is a safety hazard in the current device and suggesting that they contact the customer service as soon as possible for follow-up processing. Through this multi-level data processing and optimization mechanism, the comprehensive monitoring and intelligent management of the electric scooter status are realized, significantly improving the reliability of the system and the user experience. This efficient control strategy can not only detect and respond to potential safety hazards in a timely manner but also provide scientific maintenance suggestions for operators to ensure user safety and service quality. The entire process, from the analysis of the risk level classification results, to the generation of the task scheduling sequence, to the formulation of the edge computing optimization strategy, and finally the generation of the control strategy instruction set, forms a complete closed-loop management system to ensure that each scooter can be managed and maintained in a timely and effective manner.

[0099] In a specific embodiment, inputting the control strategy instruction set into the remote motor controller to control the electric scooter includes:

[0100] Decoding and restoring the control strategy instruction set to obtain a control parameter sequence, and performing parameter boundary verification on the control parameter sequence to obtain a safety control threshold matrix; wherein, the safety control threshold matrix includes the maximum rotational speed limit, the maximum steering angle, the maximum power output, and the maximum braking force;

[0101] Performing parallel segmented mapping on the safety control threshold matrix through a multi-threaded channel to obtain a multi-channel control signal flow, and performing real-time timing synchronization on the multi-channel control signal flow to obtain a synchronous control instruction sequence;

[0102] Performing waveform modulation transformation on the synchronous control instruction sequence to obtain a driving waveform cluster, and performing pulse width encoding on the driving waveform cluster to obtain a motor drive signal chain;

[0103] Performing hardware timing analysis on the motor controller based on the motor drive signal chain to obtain a controller execution timing table, and performing parallel task allocation on the controller execution timing table to obtain a multi-core execution instruction group;

[0104] Performing hardware drive mapping on the multi-core execution instruction group to obtain a motor control drive sequence, and performing real-time motion control on the electric scooter based on the motor control drive sequence.

[0105] Specifically, in the remote monitoring method of electric scooters based on the Internet of Things, inputting the control strategy instruction set into a remote motor controller to achieve real-time motion control of the electric scooter is a complex and multi-level process. First, the system decodes and restores the control strategy instruction set to generate a control parameter sequence, and performs parameter boundary verification on these sequences to obtain a safety control threshold matrix. Specifically, instruction decoding and restoration is to convert the high-level control strategy instruction set into a specific control parameter sequence, enabling it to be directly applied to the motor controller. For example, in a shared scooter service system, assume that a certain scooter is evaluated as a high-risk state due to overheating of the battery. The system generates a control strategy instruction set including speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies. By decoding and restoring these instructions, a control parameter sequence including specific values such as maximum rotational speed limit, maximum steering angle, maximum power output, and maximum braking force can be generated. Then, through parameter boundary verification, it is ensured that these control parameters are within a safe range, generating a safety control threshold matrix. This verification mechanism not only ensures the safety of the system but also prevents potential hazards caused by misoperations. Next, the system performs parallel segmented mapping on the safety control threshold matrix through a multi-threaded channel to generate a multi-channel control signal stream, and performs real-time timing synchronization on these signal streams to generate a synchronous control instruction sequence. The multi-threaded channel allows multiple control tasks to be processed simultaneously, thereby improving the response speed and efficiency of the system. For example, in the above application scenario, assume that it is necessary to adjust the speed, steering angle, and braking force of the scooter simultaneously. The system can process these tasks in parallel through the multi-threaded channel to generate a multi-channel control signal stream. Then, through real-time timing synchronization of these signal streams, it is ensured that each control signal can reach the motor controller at the correct time point, generating a synchronous control instruction sequence. This synchronization mechanism not only improves the real-time performance of the system but also ensures the coordination and consistency among control signals. Subsequently, the system performs waveform modulation transformation on the synchronous control instruction sequence to generate a drive waveform cluster, and performs pulse width encoding on these waveform clusters to generate a motor drive signal chain. Waveform modulation transformation is a key step in converting digital control signals into analog drive waveforms, and pulse width encoding further converts these waveforms into pulse signals suitable for processing by the motor controller. For example, in the above application scenario, by performing waveform modulation transformation on the synchronous control instruction sequence, a series of waveforms suitable for motor drive can be generated. Then, through pulse width encoding technology, these waveforms are converted into specific pulse signals to form a motor drive signal chain. These signal chains not only contain specific control information on speed, steering angle, and braking force but also can precisely guide the motor controller to perform corresponding operations. Then, the system performs hardware timing analysis on the motor controller based on the motor drive signal chain to generate a controller execution timing table, and performs parallel task allocation on this table to generate a multi-core execution instruction group.Hardware timing analysis aims to generate a detailed controller execution timing table based on the specific content of the motor drive signal chain, ensuring that each control signal can be executed at the correct time point. For example, in the above application scenario, by performing hardware timing analysis on the motor drive signal chain, a detailed execution timing table can be generated, specifying the trigger time and duration of each control signal. Then, through parallel task allocation technology, these timing tasks are assigned to multiple processor cores to generate a multi-core execution instruction set. This allocation mechanism not only improves the system's processing power but also ensures the coordination and synchronization among various control tasks. Finally, the system performs hardware drive mapping on the multi-core execution instruction set to generate a motor control drive sequence, and based on this sequence, real-time motion control of the electric scooter is carried out. Hardware drive mapping is a crucial step in converting abstract control instructions into specific hardware operations, ensuring that the motor controller can accurately execute control commands. For example, in the above application scenario, by performing hardware drive mapping on the multi-core execution instruction set, a detailed motor control drive sequence can be generated, specifying the specific operation details of each control signal. Then, based on this drive sequence, the motor controller can adjust the speed, steering angle, and braking force of the scooter in real time to ensure the safe operation of the vehicle. For example, when the system detects that the scooter battery is overheating, the motor controller will gradually reduce the speed of the scooter and finally stop its operation, while sending an alarm message to the user, indicating that there is a safety hazard in the current device and the user needs to end the journey immediately and contact the customer service for inspection. To better understand the entire process, we can envision a specific application scenario: In a busy urban environment, a shared scooter company operates thousands of scooters. One day, due to bad weather and frequent use, many scooters showed varying degrees of abnormalities. Suppose one of the scooters suddenly had a serious mechanical failure during operation, resulting in abnormal fluctuations in the data of the acceleration sensor and gyroscope sensor. First, the system analyzes the risk level classification result of the scooter to generate a control strategy instruction set, including speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies. By decoding and restoring these instructions, a control parameter sequence containing specific values such as maximum rotational speed limit, maximum steering angle, maximum power output, and maximum braking force is generated, and a safety control threshold matrix is generated through parameter boundary verification. Then, the system performs parallel segmented mapping on the safety control threshold matrix through multi-threaded channels to generate multi-channel control signal flows, and generates a synchronous control instruction sequence through real-time timing synchronization. For example, suppose it is necessary to adjust the speed, steering angle, and braking force of the scooter simultaneously. The system can process these tasks in parallel through multi-threaded channels to generate multi-channel control signal flows. Then, through real-time timing synchronization of these signal flows, it is ensured that each control signal can reach the motor controller at the correct time point to generate a synchronous control instruction sequence.Subsequently, the system performs waveform modulation transformation on the synchronous control instruction sequence to generate a cluster of drive waveforms, and generates a motor drive signal chain through pulse width encoding. For example, by performing waveform modulation transformation on the synchronous control instruction sequence, a series of waveform clusters suitable for motor drive are generated. Then, through pulse width encoding technology, these waveforms are converted into specific pulse signals to form a motor drive signal chain. These signal chains not only contain specific control information such as speed, steering angle, and braking force, but also can accurately guide the motor controller to perform corresponding operations. Then, the system performs hardware timing analysis on the motor drive signal chain to generate a controller execution timing table, and generates a multi-core execution instruction group through parallel task allocation. For example, by performing hardware timing analysis on the motor drive signal chain, a detailed execution timing table is generated to clarify the trigger time and duration of each control signal. Then, through parallel task allocation technology, these timing tasks are assigned to multiple processor cores to generate a multi-core execution instruction group. Finally, the system performs hardware drive mapping on the multi-core execution instruction group to generate a motor control drive sequence, and performs real-time motion control on the electric scooter based on this sequence. For example, by performing hardware drive mapping on the multi-core execution instruction group, a detailed motor control drive sequence is generated to clarify the specific operation details of each control signal. Then, based on this drive sequence, the motor controller can adjust the speed, steering angle, and braking force of the scooter in real time to ensure the safe operation of the vehicle. For example, when the system detects that the battery of the scooter is overheated, the motor controller will gradually reduce the speed of the scooter and finally stop its operation, while sending an alarm message to the user to prompt that there is a safety hazard in the current device and it is necessary to end the journey immediately and contact the customer service for inspection. Through this multi-level data processing and optimization mechanism, the comprehensive monitoring and intelligent management of the state of the electric scooter are realized, significantly improving the reliability of the system and the user experience. This efficient control strategy can not only detect and respond to potential safety hazards in a timely manner, but also provide scientific maintenance suggestions for operators to ensure the safety of users and service quality. The entire process from the decoding and restoration of the control strategy instruction set, to the parallel processing of the multi-threaded channel, to the hardware drive mapping and real-time motion control, forms a complete closed-loop management system to ensure that each scooter can be managed and maintained in a timely and effective manner.

[0106] In a specific embodiment, the pulse width encoding of the drive waveform cluster to obtain a motor drive signal chain includes:

[0107] Performing waveform feature decomposition on the drive waveform cluster to obtain a fundamental wave parameter matrix, and performing multiple sampling quantization on the fundamental wave parameter matrix to obtain a waveform sampling sequence;

[0108] Based on the waveform sampling sequence, a carrier signal is synthesized to obtain a set of carrier modulation parameters, and the set of carrier modulation parameters is subjected to phase compensation calibration to obtain a phase correction matrix;

[0109] The duty cycle modulation calculation is performed on the phase correction matrix to obtain a PWM modulation sequence, and the dead time compensation is performed on the PWM modulation sequence to obtain a compensation control vector;

[0110] The edge detection synchronization is performed on the compensation control vector through a preset synchronous trigger to obtain a set of trigger timings, and the multi-level interleaved coding is performed on the set of trigger timings to obtain a driving waveform sequence, and the output parameter mapping is performed on the driving waveform sequence to obtain a motor drive signal chain; wherein, the motor drive signal chain includes a PWM waveform, a phase trigger point, a duty cycle parameter, and a frequency modulation value.

[0111] Specifically, in the method for remote monitoring of electric scooters based on the Internet of Things, pulse-width encoding of the drive waveform cluster to generate a motor drive signal chain is a complex and multi-level process. First, the system decomposes the waveform features of the drive waveform cluster to generate a fundamental wave parameter matrix, and performs multiple sampling quantizations on this matrix to obtain a waveform sampling sequence. Specifically, waveform feature decomposition is the process of disassembling a complex drive waveform cluster into multiple basic waveform components, enabling more detailed analysis of the specific characteristics of each component. For example, in a shared scooter service system, assume that a certain scooter needs to adjust its speed and steering angle. The system first generates a drive waveform cluster containing this control information. By performing waveform feature decomposition on these waveform clusters, the specific parameters of each basic waveform component (such as sine waves, square waves, etc.) can be extracted to form a fundamental wave parameter matrix. Then, through multiple sampling quantization techniques, these fundamental wave parameters are further converted into discrete waveform sampling sequences to ensure that the characteristics of the original waveform can be accurately represented. Next, the system synthesizes carrier signals based on the waveform sampling sequence to generate a set of carrier modulation parameters, and performs phase compensation calibration on this set to obtain a phase correction matrix. Carrier signal synthesis aims to combine multiple basic waveform components into a carrier signal suitable for processing by the motor controller. For example, in the above application scenario, the system synthesizes a carrier signal suitable for motor drive according to the parameters in the waveform sampling sequence to generate a set of carrier modulation parameters. Then, by performing phase compensation calibration on these parameter sets, the phase consistency between each carrier signal is ensured, and a phase correction matrix is generated. This calibration mechanism not only improves the accuracy of the system but also prevents potential problems caused by out-of-phase synchronization. Subsequently, the system performs duty cycle modulation calculation on the phase correction matrix to generate a PWM modulation sequence, and performs dead-time compensation on this sequence to generate a compensation control vector. Duty cycle modulation calculation is the key step in converting the phase-corrected carrier signal into a specific PWM (pulse-width modulation) signal, and dead-time compensation further optimizes these PWM signals to ensure their stability and reliability in practical applications. For example, in the above application scenario, by performing duty cycle modulation calculation on the phase correction matrix, a series of PWM modulation sequences can be generated, specifying the duty cycle parameters of each PWM signal. Then, by performing dead-time compensation on these PWM modulation sequences, a compensation control vector is generated to ensure that there is no conflict or overlap between each PWM signal, thereby improving the stability of the system. Then, the system performs edge detection synchronization on the compensation control vector through a preset synchronization trigger to generate a set of trigger timings, and performs multi-level interleaved encoding on this set to generate a drive waveform sequence. The role of the synchronization trigger is to ensure that each PWM signal can be triggered at the correct time point to generate an accurate set of trigger timings. For example, in the above application scenario, the system uses a synchronization trigger to perform edge detection synchronization on the compensation control vector to generate a set of trigger timings, specifying the trigger time and duration of each PWM signal.Next, by performing multilevel interleaved coding on this set, a driving waveform sequence is generated to ensure that each PWM signal can be output at appropriate time points, forming a continuous and stable driving waveform. Finally, the system performs output parameter mapping on the driving waveform sequence to generate a motor drive signal chain. Output parameter mapping is a key step in converting the specific control information in the driving waveform sequence into a signal format suitable for processing by the motor controller. For example, in the above application scenario, by performing output parameter mapping on the driving waveform sequence, a motor drive signal chain including PWM waveforms, phase trigger points, duty cycle parameters, and frequency modulation values can be generated. These signal chains not only contain specific control instructions but also can precisely guide the motor controller to perform corresponding operations to ensure the safe operation of the scooter. For example, when the system detects that the scooter battery is overheating, the motor controller will gradually reduce the speed of the scooter and finally stop its operation, while sending an alarm message to the user, indicating that there is a safety hazard in the current device and the user needs to end the journey immediately and contact the customer service for inspection. To better understand the entire process, we can imagine such a specific application scenario: In a busy urban environment, a shared scooter company operates thousands of scooters. One day, due to bad weather and frequent use, many scooters showed varying degrees of abnormalities. Suppose one of the scooters suddenly had a serious mechanical failure during driving, resulting in abnormal fluctuations in the data of the acceleration sensor and gyroscope sensor. First, the system analyzes the risk level classification result of this scooter to generate a control strategy instruction set, including speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies. By decoding and restoring these instructions, a control parameter sequence including specific values such as maximum rotational speed limit, maximum steering angle, maximum power output, and maximum braking force is generated, and a safety control threshold matrix is generated through parameter boundary verification. Next, the system performs parallel segmented mapping on the safety control threshold matrix through a multi-threaded channel to generate a multi-channel control signal flow, and generates a synchronous control instruction sequence through real-time timing synchronization. For example, suppose it is necessary to adjust the speed, steering angle, and braking force of the scooter simultaneously. The system can process these tasks in parallel through a multi-threaded channel to generate a multi-channel control signal flow. Then, through real-time timing synchronization of these signal flows, ensure that each control signal can reach the motor controller at the correct time point to generate a synchronous control instruction sequence. Subsequently, the system performs waveform modulation transformation on the synchronous control instruction sequence to generate a driving waveform cluster, and generates a motor drive signal chain through pulse width coding. Specifically, the system first decomposes the waveform characteristics of the driving waveform cluster to generate a fundamental wave parameter matrix, and generates a waveform sampling sequence through multiple sampling quantization. For example, in the above application scenario, the system generates a driving waveform cluster according to the synchronous control instruction sequence, decomposes its waveform characteristics, extracts the specific parameters of each basic waveform component, and forms a fundamental wave parameter matrix.Next, through the multi-sampling quantization technique, these fundamental wave parameters are converted into a discrete waveform sampling sequence to ensure that the characteristics of the original waveform can be accurately represented. Then, the system synthesizes a carrier signal based on the waveform sampling sequence, generates a set of carrier modulation parameters, and performs phase compensation calibration on this set to generate a phase correction matrix. For example, in the above application scenario, the system synthesizes a carrier signal suitable for motor drive according to the parameters in the waveform sampling sequence and generates a set of carrier modulation parameters. Then, by performing phase compensation calibration on these parameter sets, the phase consistency between each carrier signal is ensured, and a phase correction matrix is generated. Next, the system performs duty cycle modulation calculation on the phase correction matrix to generate a PWM modulation sequence, and compensates the dead time for this sequence to generate a compensation control vector. For example, in the above application scenario, the system performs duty cycle modulation calculation on the phase correction matrix to generate a series of PWM modulation sequences, and determines the duty cycle parameters of each PWM signal. Then, by compensating the dead time for these PWM modulation sequences, a compensation control vector is generated to ensure that there is no conflict or overlap between each PWM signal, thereby improving the stability of the system. Then, the system performs edge detection synchronization on the compensation control vector through a preset synchronization trigger to generate a set of trigger timings, and performs multi-level interleaved coding on this set to generate a drive waveform sequence. For example, in the above application scenario, the system uses a synchronization trigger to perform edge detection synchronization on the compensation control vector to generate a set of trigger timings, and determines the trigger time and duration of each PWM signal. Then, by performing multi-level interleaved coding on this set, a drive waveform sequence is generated to ensure that each PWM signal can be output at the appropriate time point to form a continuous and stable drive waveform. Finally, the system performs output parameter mapping on the drive waveform sequence to generate a motor drive signal chain. For example, in the above application scenario, by performing output parameter mapping on the drive waveform sequence, a motor drive signal chain including PWM waveforms, phase trigger points, duty cycle parameters, and frequency modulation values is generated. These signal chains not only contain specific control instructions but also can accurately guide the motor controller to execute corresponding operations to ensure the safe operation of the scooter. For example, when the system detects that the scooter battery is overheated, the motor controller will gradually reduce the speed of the scooter and finally stop its operation, while sending an alarm message to the user, indicating that there is a safety hazard in the current device and the user needs to end the journey immediately and contact the customer service for inspection. Through this multi-level data processing and optimization mechanism, the comprehensive monitoring and intelligent management of the electric scooter state are realized, significantly improving the reliability of the system and the user experience. This efficient control strategy can not only detect and respond to potential safety hazards in a timely manner but also provide scientific maintenance suggestions for operators to ensure user safety and service quality.The entire process, from waveform feature decomposition, to carrier signal synthesis, then to PWM modulation and dead-time compensation, and finally generating the motor drive signal chain, forms a complete closed-loop management system to ensure that each scooter can be managed and maintained in a timely and effective manner.

[0112] The above described the method for remote monitoring of electric scooters based on the Internet of Things in the embodiments of the present invention. Next, the remote monitoring system for electric scooters based on the Internet of Things in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the remote monitoring system for electric scooters based on the Internet of Things in the embodiments of the present invention includes:

[0113] An acquisition module 21, configured to collect data from multi-source sensors carried by the electric scooter to obtain an original sensing data stream; wherein, the original sensing data stream includes a gyroscope sensor, an acceleration sensor, a GPS positioning sensor, and a battery management sensor;

[0114] An extraction module 22, configured to perform multi-dimensional feature extraction and data fusion on the original sensing data stream to obtain a scooter state feature vector;

[0115] A mapping module 23, configured to perform multi-modal feature mapping on the electric scooter based on the scooter state feature vector to obtain vehicle abnormal behavior pattern features;

[0116] A prediction module 24, configured to perform risk prediction on the electric scooter based on the vehicle abnormal behavior pattern features to obtain a risk level classification result;

[0117] A sorting module 25, configured to perform distributed processing and priority sorting on the risk level classification result through an edge computing gateway to obtain a control strategy instruction set;

[0118] A control module 26, configured to input the control strategy instruction set into the remote motor controller to control the electric scooter.

[0119] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to that described in the above method embodiment, and details will not be repeated here.

[0120] Refer to Figure 3 , the embodiments of the present invention also provide a computer device, the internal structure of which can be as Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected via a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0121] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0122] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0123] Those of ordinary skill in the art can understand that all or part of the processes in the above embodiment methods can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0124] It should be noted that, in this text, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, apparatus, article or method comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, apparatus, article or method. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, apparatus, article or method comprising such element.

[0125] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A remote monitoring method for electric scooters based on the Internet of Things, characterized in that: The electric scooter is provided with a motor controller, comprising the following steps: Collect data from multiple source sensors on the electric scooter to obtain an original sensor data stream; wherein the original sensor data stream includes a gyroscope sensor, an acceleration sensor, a GPS positioning sensor, and a battery management sensor; Performing multi-dimensional feature extraction and data fusion on the original sensor data stream to obtain a scooter state feature vector; If the scooter state feature vector indicates that the electric scooter is abnormal, multimodal feature mapping is performed on the electric scooter based on the scooter state feature vector to obtain abnormal behavior pattern features of the vehicle; Based on the abnormal behavior pattern characteristics of the vehicle, risk prediction is performed on the electric scooter to obtain a risk level classification result; Distributed processing and priority sorting of the risk level classification results are performed through an edge computing gateway to obtain a control strategy instruction set; Inputting the control strategy instruction set into the remote motor controller to control the electric scooter; The multimodal feature mapping of the electric scooter based on the scooter state feature vector to obtain the abnormal behavior pattern feature of the vehicle includes: Performing multi-dimensional space projection on the scooter state feature vector to obtain a feature distribution mapping atlas, and performing nonlinear manifold embedding on the feature distribution mapping atlas to obtain a topological feature matrix; The topological feature matrix is ​​similarly measured by a preset local sensitive hashing technology to obtain a feature similarity spectrum, and the feature similarity spectrum is subjected to spectral clustering analysis to obtain a behavior pattern prototype set; wherein the behavior pattern prototype set includes an acceleration mode, a steering mode, and a braking mode; Extracting spatiotemporal features from the behavior pattern prototype set to obtain a spatiotemporal feature tensor, and performing tensor decomposition on the spatiotemporal feature tensor to obtain multimodal behavior features; The multimodal behavior characteristics are subjected to a time-series dependency analysis through a preset conditional random field technology to obtain a state transition sequence, and the state transition sequence is subjected to a Markov jump analysis to obtain a behavior pattern transition diagram; wherein the behavior pattern transition diagram includes state duration, state transition probability, and state stability; Based on the preset historical behavior data, the behavior pattern conversion graph is subjected to distributed association mining to obtain the behavior pattern association rules, and the behavior pattern association rules are subjected to probability graph reasoning to obtain the behavior prediction vector; wherein the behavior prediction vector includes the probability of behavior occurrence, the duration of the behavior, and the behavior conversion trend; Multi-scale feature fusion is performed on the behavior prediction vector to obtain a multi-dimensional behavior feature space, and abnormal behavior detection is performed based on the multi-dimensional behavior feature space to obtain vehicle abnormal behavior pattern characteristics; wherein the vehicle abnormal behavior pattern characteristics include abnormality occurrence probability, abnormality degree score and abnormal behavior type.

2. The electric scooter remote monitoring method based on the Internet of Things according to claim 1 is characterized in that: The multi-dimensional feature extraction and data fusion of the original sensor data stream to obtain the scooter state feature vector includes: Decomposing the original sensor data stream in the time-frequency domain to obtain a vibration characteristic sequence of the electric scooter, and decomposing the vibration characteristic sequence by wavelet packets to obtain a multi-scale vibration spectrum diagram; Performing scooter posture data analysis on the electric scooter based on the multi-scale vibration spectrum to obtain a posture feature set, performing time-series alignment on the posture feature set to obtain an aligned state sequence, and performing sparse coding on the aligned state sequence to obtain a compressed feature representation; Multimodal feature fusion is performed on the compressed feature representation to obtain a fused feature matrix, and tensor decomposition is performed on the fused feature matrix to obtain a scooter state feature vector.

3. The electric scooter remote monitoring method based on the Internet of Things according to claim 1 is characterized in that: The risk prediction of the electric scooter based on the abnormal behavior pattern characteristics of the vehicle to obtain a risk level classification result includes: Performing structural decomposition on the abnormal behavior pattern characteristics of the vehicle to obtain a behavior characteristic component sequence, and performing dynamic evolution analysis on the behavior characteristic component sequence through a recursive quantitative analysis method to obtain a behavior evolution trajectory diagram, and performing fractal dimension calculation on the behavior evolution trajectory diagram to obtain a fault evolution feature set; wherein the fault evolution feature set includes component wear trend, fault diffusion rate, and system degradation degree; Performing causal chain analysis on the fault evolution feature set based on preset historical fault data to obtain a fault propagation network, and extracting topological features from the fault propagation network to obtain a set of key fault nodes; Probabilistic reasoning is performed on the key fault node set through a preset Bayesian network algorithm to obtain a component reliability evaluation value, and a fuzzy comprehensive evaluation is performed on the electric scooter based on the component reliability evaluation value to obtain a health status score sheet; wherein the health status score sheet includes motor health, battery life value and bearing wear; An operation risk analysis of the electric scooter is performed based on the health status score sheet to obtain a risk level classification result.

4. The method for remote monitoring of electric scooters based on the Internet of Things according to claim 1, characterized in that: The risk level classification results are distributedly processed and prioritized by the edge computing gateway to obtain a control strategy instruction set, including: Performing data structure analysis on the risk level classification result to obtain a risk feature sequence, and performing multi-level threshold quantization on the risk feature sequence to obtain a graded risk numerical matrix; wherein the graded risk numerical matrix includes a motor control risk value, a battery safety threshold, and a bearing loss index; Performing distributed parallel decoupling on the hierarchical risk numerical matrix to obtain a risk priority mapping table, and performing adaptive allocation of network topology based on the risk priority mapping table to obtain a task scheduling sequence; wherein the task scheduling sequence includes an emergency task queue, a regular task queue, and a low-priority task queue; By using the edge computing gateway, resource allocation is optimized for the task scheduling sequence to obtain an edge node allocation scheme, and load balancing is adjusted for the edge node allocation scheme to obtain an edge computing optimization strategy; Based on the edge computing optimization strategy, the preset scooter control parameters are adjusted and mapped in real time to obtain a control strategy instruction set; wherein the control strategy instruction set includes speed limit instructions, steering angle constraints, battery power control, and emergency braking strategies.

5. The method for remote monitoring of electric scooters based on the Internet of Things according to claim 1, characterized in that: The step of inputting the control strategy instruction set into the remote motor controller to control the electric scooter comprises: Decoding and restoring the control strategy instruction set to obtain a control parameter sequence, and performing parameter boundary check on the control parameter sequence to obtain a safety control threshold matrix; wherein the safety control threshold matrix includes a maximum speed limit, a maximum steering angle, a maximum power output, and a maximum braking force; The safety control threshold matrix is ​​mapped in parallel segments through a multi-thread channel to obtain a multi-channel control signal stream, and the multi-channel control signal stream is synchronized in real time to obtain a synchronous control instruction sequence; Performing waveform modulation transformation on the synchronous control instruction sequence to obtain a drive waveform cluster, and performing pulse width encoding on the drive waveform cluster to obtain a motor drive signal chain; Performing hardware timing analysis on the motor controller based on the motor drive signal chain to obtain a controller execution timing table, and performing parallel task allocation on the controller execution timing table to obtain a multi-core execution instruction group; The multi-core execution instruction group is mapped to a hardware driver to obtain a motor control drive sequence, and the electric scooter is controlled in real time based on the motor control drive sequence.

6. The method for remote monitoring of electric scooters based on the Internet of Things according to claim 5, characterized in that: The pulse width encoding of the driving waveform cluster is performed to obtain a motor driving signal chain, including: Performing waveform feature decomposition on the driving waveform cluster to obtain a fundamental wave parameter matrix, and performing multiple sampling quantization on the fundamental wave parameter matrix to obtain a waveform sampling sequence; Carrier signal synthesis is performed based on the waveform sampling sequence to obtain a carrier modulation parameter set, and phase compensation calibration is performed on the carrier modulation parameter set to obtain a phase correction matrix; Performing duty cycle modulation calculation on the phase correction matrix to obtain a PWM modulation sequence, and performing dead time compensation on the PWM modulation sequence to obtain a compensation control vector; The compensation control vector is edge-detected and synchronized by a preset synchronization trigger to obtain a trigger timing set, and the trigger timing set is multi-level interleaved encoded to obtain a drive waveform sequence, and the drive waveform sequence is output parameter mapped to obtain a motor drive signal chain; wherein the motor drive signal chain includes a PWM waveform, a phase trigger point, a duty cycle parameter, and a frequency modulation value.

7. An electric scooter remote monitoring system based on the Internet of Things, characterized in that: The electric scooter is provided with a motor controller, including: The acquisition module is used to collect data from the multi-source sensors carried by the electric scooter to obtain the original sensor data stream; wherein the original sensor data stream includes a gyroscope sensor, an acceleration sensor, a GPS positioning sensor, and a battery management sensor; An extraction module, used for performing multi-dimensional feature extraction and data fusion on the original sensor data stream to obtain a scooter state feature vector; A mapping module, used for performing multimodal feature mapping on the electric scooter based on the scooter state feature vector to obtain vehicle abnormal behavior pattern features; A prediction module, used to predict the risk of the electric scooter based on the abnormal behavior pattern characteristics of the vehicle, and obtain a risk level classification result; A sorting module, used for performing distributed processing and priority sorting on the risk level classification results through an edge computing gateway to obtain a control strategy instruction set; A control module, used for inputting the control strategy instruction set into the remote motor controller to control the electric scooter; The multimodal feature mapping of the electric scooter based on the scooter state feature vector to obtain the abnormal behavior pattern feature of the vehicle includes: Performing multi-dimensional space projection on the scooter state feature vector to obtain a feature distribution mapping atlas, and performing nonlinear manifold embedding on the feature distribution mapping atlas to obtain a topological feature matrix; The topological feature matrix is ​​similarly measured by a preset local sensitive hashing technology to obtain a feature similarity spectrum, and the feature similarity spectrum is subjected to spectral clustering analysis to obtain a behavior pattern prototype set; wherein the behavior pattern prototype set includes an acceleration mode, a steering mode, and a braking mode; Extracting spatiotemporal features from the behavior pattern prototype set to obtain a spatiotemporal feature tensor, and performing tensor decomposition on the spatiotemporal feature tensor to obtain multimodal behavior features; The multimodal behavior characteristics are subjected to a time-series dependency analysis through a preset conditional random field technology to obtain a state transition sequence, and the state transition sequence is subjected to a Markov jump analysis to obtain a behavior pattern transition diagram; wherein the behavior pattern transition diagram includes state duration, state transition probability, and state stability; Based on the preset historical behavior data, the behavior pattern conversion graph is subjected to distributed association mining to obtain the behavior pattern association rules, and the behavior pattern association rules are subjected to probability graph reasoning to obtain the behavior prediction vector; wherein the behavior prediction vector includes the probability of behavior occurrence, the duration of the behavior, and the behavior conversion trend; Multi-scale feature fusion is performed on the behavior prediction vector to obtain a multi-dimensional behavior feature space, and abnormal behavior detection is performed based on the multi-dimensional behavior feature space to obtain vehicle abnormal behavior pattern characteristics; wherein the vehicle abnormal behavior pattern characteristics include abnormality occurrence probability, abnormality degree score and abnormal behavior type.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images