High-precision positioning and navigation remote emergency takeover system

Abstract

The invention discloses a high-precision positioning and navigation remote emergency takeover system, and relates to the technical field of navigation and position service, and the system comprises a safety communication control module which is in communication connection with a data perception fusion module, a data optimization conversion module and an intelligent decision response module; the data sensing fusion module collects centimeter-level positioning, acceleration and angular velocity, an environment model is created in combination with a laser radar and a visual sensor, the data optimization conversion module optimizes data, the intelligent decision response module recognizes abnormity and generates a control instruction, and the safety communication control module constructs a two-way communication link. According to the high-precision positioning and navigation remote emergency takeover system, by integrating multi-sensor data sensing and fusion, data optimization and intelligent decision making, centimeter-level accurate positioning and environment model construction are achieved, abnormity is effectively recognized, a control instruction is intelligently generated, meanwhile, safe communication is ensured, and the navigation precision and the emergency response capability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of navigation and location services, and in particular to a high-precision positioning and navigation remote emergency takeover system. Background Art

[0002] With the acceleration of urbanization and the frequent occurrence of various emergencies, the demand for emergency management capabilities is increasing. The remote emergency takeover system can realize remote monitoring and control of emergency equipment through wireless communication technology. In the fields of public safety, traffic management, environmental monitoring, etc., the system can play an important role. In addition, with the continuous development of technologies such as cloud computing, big data, and artificial intelligence, the remote emergency takeover system is also gradually becoming intelligent and automated. For example, by integrating intelligent algorithms, the system can realize real-time monitoring and early warning of potential risks, further improving the speed and accuracy of emergency response. In the future, with the continuous advancement of technology and the continuous expansion of application scenarios, the remote emergency takeover system is expected to be widely used in more fields, providing more powerful technical support for improving social emergency management capabilities and protecting the lives and property of the people.

[0003] In the existing technology, there is a problem of low system performance and response speed in emergency situations. Therefore, how to improve the response speed and response efficiency of the remote emergency takeover system in emergency situations, and how to ensure the performance stability of the remote emergency takeover system when facing unexpected situations are the problems we need to solve. To this end, a high-precision positioning and navigation remote emergency takeover system is proposed. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a high-precision positioning and navigation remote emergency takeover system, which solves the problems in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a high-precision positioning and navigation remote emergency takeover system, including a safety communication control module, the safety communication control module is communicatively connected with a data perception fusion module, a data optimization conversion module and an intelligent decision response module;

[0006] The data perception fusion module collects centimeter-level positioning data through the global satellite positioning system, collects acceleration and angular velocity through the inertial measurement unit, and creates an environmental model through laser radar and visual sensors in combination with synchronous positioning and map building algorithms;

[0007] The data optimization conversion module uses a multi-sensor fusion algorithm to optimize data quality;

[0008] The intelligent decision response module uses a long short-term memory network combined with a convolutional neural network algorithm to identify abnormal conditions and assess their severity, determine whether it is necessary to start a remote takeover procedure and generate control instructions;

[0009] The safety communication control module establishes a two-way communication link between the local remote emergency takeover system and the remote operation station, and monitors the communication status.

[0010] Furthermore, in the data sensing and fusion module, the process of collecting centimeter-level positioning data through the global satellite positioning system includes:

[0011] Installing an enhanced global positioning system on the mobile device, the global positioning system receiver determines the three-dimensional position of the mobile device by receiving satellite signals;

[0012] Unify the raw GPS data into RINEX format, identify and exclude abnormal observations caused by multipath and ionospheric delay, and perform time synchronization and error correction;

[0013] Using carrier phase observation, differential measurement is performed through the differential technology RTK to solve the integer cycle number in the carrier phase measurement, centimeter-level positioning data is collected, and differential operations are performed using continuous time series positioning data. The rate of change of the distance between two points is calculated to obtain the velocity, and the direction of the velocity is determined by the azimuth angles of adjacent position points. The global satellite positioning system receiver calculates different types of geometric precision factors DOP based on the received satellite signals. The geometric precision factors include position DOP, horizontal DOP and vertical DOP to evaluate the quality of positioning data. The global satellite positioning system reflection signal analysis technology is used to detect the signal reflection surface information. By comparing the signal characteristics of the direct path and the reflection path, the ground type and the height of the surrounding buildings are inferred.

[0014] Furthermore, in the data perception fusion module, the process of collecting acceleration and angular velocity through the inertial measurement unit includes:

[0015] Install an inertial measurement unit at the center of gravity of the mobile device to reduce the influence of centrifugal force and Coriolis force caused by rotation, and perform temperature compensation and calibration to eliminate installation errors;

[0016] The accelerometer in the inertial measurement unit detects the displacement of the mobile device relative to the accelerometer housing, collects the acceleration, and uses the gyroscope in the inertial measurement unit to measure the angular velocity using the principle of optical interference;

[0017] Time-synchronize acceleration data with angular rate data, apply low-pass filters to remove high-frequency noise, integrate inertial measurement unit readings using Kalman filtering techniques to reduce drift errors, adjust accelerometer and gyroscope outputs to offset the effects of temperature changes, and perform static and dynamic calibrations to correct for zero bias, scale factor errors, and inter-axis non-orthogonality;

[0018] The mobile device velocity data is obtained by integrating the acceleration data, and the mobile device position data is obtained by integrating it again. The angular velocity data of the gyroscope and the gravity direction indication provided by the accelerometer are integrated, and the pitch, roll and heading angles are calculated using a complementary filtering algorithm. The time series data of the inertial measurement unit is analyzed to identify different motion modes, including acceleration, deceleration and turning.

[0019] Furthermore, in the data perception fusion module, the process of creating an environment model by combining laser radar and visual sensors with a synchronous positioning and map building algorithm includes:

[0020] Install a lidar on top of the mobile device to provide 360-degree coverage, and install an RGB-D camera to work with the lidar to provide complementary visual information;

[0021] The LiDAR calculates the distance of the mobile device by emitting laser pulses and measuring the reflection time, generating point cloud data to analyze the position and shape of the mobile device in the environment. The LiDAR captures image sequences through an RGB-D camera, and uses computer vision algorithms to extract feature points, line segments, and distances of pixels in the image to the RDB-D camera.

[0022] Aligning the lidar point cloud data and the visual image data in space and time, expressing them in the same coordinate system, applying filters to remove outliers in the lidar point cloud data and noise in the visual image data, identifying natural features from the visual image data, the natural features including corners and edges, and correcting the deviation of the point cloud data and the image data caused by the movement of the mobile device based on the posture information provided by the inertial measurement unit;

[0023] Extract plane and cylindrical elements from LiDAR point cloud data, extract straight line and circle features from visual image data to model static environment structures, analyze color and brightness changes in visual image data, and identify surface textures;

[0024] Combine LiDAR data and visual image data to distinguish between static background and moving objects, use SLAM algorithm to integrate continuous LiDAR point cloud data and visual image data features, build a map representation, which includes global map and local map updates, match current observation features with existing maps, and update the position and direction of mobile devices in real time.

[0025] Furthermore, in the data optimization conversion module, the process of optimizing data quality using a multi-sensor fusion algorithm includes:

[0026] By combining the absolute position information provided by the global satellite positioning system with the relative kinematic parameters of the inertial measurement unit, the global satellite positioning system and the inertial measurement unit are combined to realize real-time estimation of the attitude, velocity and position state vector of the mobile device, and the prediction error covariance matrix is ​​adjusted online. The environmental features obtained by the lidar and RGB-D camera are introduced, and the SLAM algorithm is applied to build and update the map representation to achieve self-positioning.

[0027] Furthermore, in the intelligent decision response module, the process of using a convolutional neural network algorithm to identify abnormal conditions and assess their severity includes:

[0028] Preprocess and fuse the collected centimeter-level positioning data, acceleration, angular velocity, point cloud data, and image data to synchronize the collected data in time and align them in space, forming a unified spatiotemporal feature sequence as the input of the convolutional neural network model and the long short-term memory network model;

[0029] A convolutional neural network model is constructed, which includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer. The input layer receives image data captured by an RGB-D camera. The convolution layer extracts edge and corner features in the image through a convolution operation, uses a filter to capture local features at different scales and directions, and adopts a ReLU activation function to enhance nonlinear expression capabilities. The pooling layer reduces the spatial dimension of the feature map through a maximum pooling operation, reduces the amount of calculation and keeps the extracted features unchanged. The fully connected layer flattens the extracted features into a one-dimensional vector, performs high-level feature combination, and outputs a fixed-length feature representation. The output layer outputs a visual feature vector for describing the environment structure and identifying objects.

[0030] Furthermore, in the intelligent decision response module, the process of identifying abnormal conditions and assessing their severity in combination with the long short-term memory network algorithm includes:

[0031] Constructing a long short-term memory network model, wherein the long short-term memory network model includes an input layer, an LSTM layer and an output layer;

[0032] The input layer receives acceleration and angular velocity time series data from the inertial measurement unit, as well as the speed and position change sequence of the global satellite positioning system. The LSTM layer learns long-term dependencies through its internal LSTM unit and forgets irrelevant information. The LSTM unit includes an input gate, a forget gate, and an output gate.

[0033] By controlling the information flow, the long time series data is processed by the network. The high-level time series features are abstracted layer by layer by stacking LSTM layers, and the output layer is used to output the time series feature vector containing the mobile device behavior pattern and path prediction.

[0034] Furthermore, in the intelligent decision response module, a hybrid architecture model combining a long short-term memory network and a convolutional neural network is constructed to identify abnormal conditions and assess their severity, including:

[0035] The visual feature vector from the convolutional neural network and the time series feature vector generated by the LSTM are concatenated into a fused feature representation, and the attention mechanism is introduced to enable the convolutional neural network model and the LSTM model to dynamically adjust the importance weight of the input data;

[0036] Combined with the map constructed by SLAM and the precise positioning information of mobile devices, the fused features can reflect the latest situational changes. Based on the fused features, the risk score is calculated to share the probability of abnormal events occurring in the current state and their potential threat level. Different weights are assigned to different types of abnormalities, and the context information is analyzed for adjustment. The softmax function is used to output the probability distribution and risk score of normal operation, minor abnormalities, and severe abnormalities, and the emergency response procedure is initiated.

[0037] Furthermore, in the intelligent decision response module, the process of determining whether to start the remote takeover program and generate a control instruction includes:

[0038] According to the output abnormality type and risk score, the remote emergency takeover system automatically generates an emergency response strategy, which includes emergency braking instructions, activation of steering correction mechanism and remote operator intervention. When an obstacle is detected suddenly in front of the mobile device and the distance is too close, an emergency braking instruction is issued. When the mobile device deviates from the planned route and is about to enter the dangerous area, the steering correction mechanism is activated. When the abnormal condition exceeds the threshold and conventional corrective measures cannot be effectively dealt with, it is determined whether the remote emergency takeover program needs to be activated, the remote operator intervenes, and sends control instructions to the mobile device. The emergency response results are recorded as a data source for improving model performance.

[0039] Furthermore, in the safety communication control module, a two-way communication link is established between the local remote emergency takeover system and the remote operation station, and the process of monitoring the communication status includes:

[0040] Select a satellite communication link, configure the standard transport layer TCP protocol and application layer MQTT protocol to transmit data packets, perform end-to-end encrypted communication through SSL, use digital certificates to verify the identities of the mobile device and remote operator, and use the local remote emergency takeover system on the mobile device to initiate a connection request, and establish a communication session after completing the handshake process;

[0041] A priority scheduling algorithm is used to allocate bandwidth resources for control commands, heartbeat signals are exchanged regularly to confirm the connection status, and CRC verification is implemented to ensure data integrity;

[0042] Build a multi-path communication solution to automatically switch to the backup channel when the main channel fails. The control instructions sent by the remote operation station reach the mobile device via the communication link, where they are parsed and passed to the execution unit. The execution results are monitored and the actual execution status is fed back to the remote operation station to form a closed-loop control, record the instruction interaction and its results, deploy an intrusion detection system to monitor abnormal behavior, and provide a one-click disconnect function to deal with extreme situations.

[0043] The present invention provides a high-precision positioning and navigation remote emergency takeover system. It has the following beneficial effects:

[0044] First, through the comprehensive use of global satellite positioning system, inertial measurement unit, lidar and visual sensor, the data perception fusion module can collect and fuse data from multiple sensors in real time and accurately to create a high-precision environmental model. This provides clear and detailed on-site information for remote emergency takeover, greatly improving the accuracy and efficiency of emergency response. Secondly, the data optimization conversion module uses a multi-sensor fusion algorithm to optimize the original data, effectively reducing data errors and noise, and improving the reliability and stability of the data. This provides a solid foundation for subsequent intelligent decision-making and emergency response. Furthermore, the intelligent decision response module uses advanced deep learning algorithms to automatically identify abnormal conditions and assess their severity, quickly determine whether it is necessary to start the remote takeover program, and generate control instructions in real time. The existence of this module enables the system to respond quickly in the face of emergencies, effectively avoiding potential risks caused by delays in human judgment. Finally, the secure communication control module builds a two-way communication link between the local remote emergency takeover system and the remote operation station, and monitors the communication status in real time. This ensures the accurate transmission of instructions and the stable operation of the system, providing reliable communication guarantee for remote emergency takeover. To sum up, this high-precision positioning and navigation remote emergency takeover system solution has significant advantages in improving emergency response speed, accuracy and stability, and is of great significance for ensuring public safety and reducing disaster losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a block diagram of a high-precision positioning and navigation remote emergency takeover system according to an embodiment of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] like Figure 1 As shown, the present invention provides a technical solution: a high-precision positioning and navigation remote emergency takeover system, including a safety communication control module, wherein the safety communication control module is communicatively connected with a data perception fusion module, a data optimization conversion module and an intelligent decision response module;

[0048] The data perception fusion module collects centimeter-level positioning data through the global satellite positioning system, collects acceleration and angular velocity through the inertial measurement unit, and creates an environmental model through laser radar and visual sensors in combination with synchronous positioning and map building algorithms;

[0049] The data optimization conversion module uses a multi-sensor fusion algorithm to optimize data quality;

[0050] The intelligent decision response module uses a long short-term memory network combined with a convolutional neural network algorithm to identify abnormal conditions and assess their severity, determine whether it is necessary to start a remote takeover procedure and generate control instructions;

[0051] The safety communication control module establishes a two-way communication link between the local remote emergency takeover system and the remote operation station, and monitors the communication status.

[0052] In the data perception fusion module, the process of collecting centimeter-level positioning data through the global satellite positioning system includes:

[0053] Installing an enhanced global positioning system on the mobile device, the global positioning system receiver determines the three-dimensional position of the mobile device by receiving satellite signals;

[0054] Unify the raw GPS data into RINEX format, identify and exclude abnormal observations caused by multipath and ionospheric delay, and perform time synchronization and error correction;

[0055] Using carrier phase observation, differential measurement is performed through the differential technology RTK to solve the integer cycle number in the carrier phase measurement, centimeter-level positioning data is collected, and differential operations are performed using continuous time series positioning data. The rate of change of the distance between two points is calculated to obtain the velocity, and the direction of the velocity is determined by the azimuth angles of adjacent position points. The global satellite positioning system receiver calculates different types of geometric precision factors DOP based on the received satellite signals. The geometric precision factors include position DOP, horizontal DOP and vertical DOP to evaluate the quality of positioning data. The global satellite positioning system reflection signal analysis technology is used to detect the signal reflection surface information. By comparing the signal characteristics of the direct path and the reflection path, the ground type and the height of the surrounding buildings are inferred.

[0056] In the data perception fusion module, the process of collecting acceleration and angular velocity through the inertial measurement unit includes:

[0057] Install an inertial measurement unit at the center of gravity of the mobile device to reduce the influence of centrifugal force and Coriolis force caused by rotation, and perform temperature compensation and calibration to eliminate installation errors;

[0058] The accelerometer in the inertial measurement unit detects the displacement of the mobile device relative to the accelerometer housing, collects the acceleration, and uses the gyroscope in the inertial measurement unit to measure the angular velocity using the principle of optical interference;

[0059] Time-synchronize acceleration data with angular rate data, apply low-pass filters to remove high-frequency noise, integrate inertial measurement unit readings using Kalman filtering techniques to reduce drift errors, adjust accelerometer and gyroscope outputs to offset the effects of temperature changes, and perform static and dynamic calibrations to correct for zero bias, scale factor errors, and inter-axis non-orthogonality;

[0060] The mobile device velocity data is obtained by integrating the acceleration data, and the mobile device position data is obtained by integrating it again. The angular velocity data of the gyroscope and the gravity direction indication provided by the accelerometer are integrated, and the pitch, roll and heading angles are calculated using a complementary filtering algorithm. The time series data of the inertial measurement unit is analyzed to identify different motion modes, including acceleration, deceleration and turning.

[0061] In the data perception fusion module, the process of creating an environment model by combining laser radar and visual sensors with a synchronous positioning and map building algorithm includes:

[0062] Install a lidar on top of the mobile device to provide 360-degree coverage, and install an RGB-D camera to work with the lidar to provide complementary visual information;

[0063] The LiDAR calculates the distance of the mobile device by emitting laser pulses and measuring the reflection time, generating point cloud data to analyze the position and shape of the mobile device in the environment. The LiDAR captures image sequences through an RGB-D camera, and uses computer vision algorithms to extract feature points, line segments, and distances of pixels in the image to the RDB-D camera.

[0064] Aligning the lidar point cloud data and the visual image data in space and time, expressing them in the same coordinate system, applying filters to remove outliers in the lidar point cloud data and noise in the visual image data, identifying natural features from the visual image data, the natural features including corners and edges, and correcting the deviation of the point cloud data and the image data caused by the movement of the mobile device based on the posture information provided by the inertial measurement unit;

[0065] Extract plane and cylindrical elements from LiDAR point cloud data, extract straight line and circle features from visual image data to model static environment structures, analyze color and brightness changes in visual image data, and identify surface textures;

[0066] Combine LiDAR data and visual image data to distinguish between static background and moving objects, use SLAM algorithm to integrate continuous LiDAR point cloud data and visual image data features, build a map representation, which includes global map and local map updates, match current observation features with existing maps, and update the position and direction of mobile devices in real time.

[0067] In the data optimization conversion module, the process of optimizing data quality using a multi-sensor fusion algorithm includes:

[0068] By combining the absolute position information provided by the global satellite positioning system with the relative kinematic parameters of the inertial measurement unit, the global satellite positioning system and the inertial measurement unit are combined to realize real-time estimation of the attitude, velocity and position state vector of the mobile device, and the prediction error covariance matrix is ​​adjusted online. The environmental features obtained by the lidar and RGB-D camera are introduced, and the SLAM algorithm is applied to build and update the map representation to achieve self-positioning.

[0069] In the intelligent decision response module, the process of using a convolutional neural network algorithm to identify abnormal conditions and assess their severity includes:

[0070] Preprocess and fuse the collected centimeter-level positioning data, acceleration, angular velocity, point cloud data, and image data to synchronize the collected data in time and align them in space, forming a unified spatiotemporal feature sequence as the input of the convolutional neural network model and the long short-term memory network model;

[0071] A convolutional neural network model is constructed, which includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer. The input layer receives image data captured by an RGB-D camera. The convolution layer extracts edge and corner features in the image through a convolution operation, uses a filter to capture local features at different scales and directions, and adopts a ReLU activation function to enhance nonlinear expression capabilities. The pooling layer reduces the spatial dimension of the feature map through a maximum pooling operation, reduces the amount of calculation and keeps the extracted features unchanged. The fully connected layer flattens the extracted features into a one-dimensional vector, performs high-level feature combination, and outputs a fixed-length feature representation. The output layer outputs a visual feature vector for describing the environment structure and identifying objects.

[0072] In the intelligent decision response module, the process of identifying abnormal conditions and assessing their severity in combination with the long short-term memory network algorithm includes:

[0073] Constructing a long short-term memory network model, wherein the long short-term memory network model includes an input layer, an LSTM layer and an output layer;

[0074] The input layer receives acceleration and angular velocity time series data from the inertial measurement unit, as well as the speed and position change sequence of the global satellite positioning system. The LSTM layer learns long-term dependencies through its internal LSTM unit and forgets irrelevant information. The LSTM unit includes an input gate, a forget gate, and an output gate.

[0075] By controlling the information flow, the long time series data is processed by the network. The high-level time series features are abstracted layer by layer by stacking LSTM layers, and the output layer is used to output the time series feature vector containing the mobile device behavior pattern and path prediction.

[0076] In the intelligent decision response module, a hybrid architecture model combining a long short-term memory network and a convolutional neural network is constructed to identify abnormal conditions and assess their severity, including:

[0077] The visual feature vector from the convolutional neural network and the time series feature vector generated by the LSTM are concatenated into a fused feature representation, and the attention mechanism is introduced to enable the convolutional neural network model and the LSTM model to dynamically adjust the importance weight of the input data;

[0078] Combined with the map constructed by SLAM and the precise positioning information of mobile devices, the fused features can reflect the latest situational changes. Based on the fused features, the risk score is calculated to share the probability of abnormal events occurring in the current state and their potential threat level. Different weights are assigned to different types of abnormalities, and the context information is analyzed for adjustment. The softmax function is used to output the probability distribution and risk score of normal operation, minor abnormalities, and severe abnormalities, and the emergency response procedure is initiated.

[0079] In the intelligent decision response module, the process of determining whether to start the remote takeover program and generate a control instruction includes:

[0080] According to the output abnormality type and risk score, the remote emergency takeover system automatically generates an emergency response strategy, which includes emergency braking instructions, activation of steering correction mechanism and remote operator intervention. When an obstacle is detected suddenly in front of the mobile device and the distance is too close, an emergency braking instruction is issued. When the mobile device deviates from the planned route and is about to enter the dangerous area, the steering correction mechanism is activated. When the abnormal condition exceeds the threshold and conventional corrective measures cannot be effectively dealt with, it is determined whether the remote emergency takeover program needs to be activated, the remote operator intervenes, and sends control instructions to the mobile device. The emergency response results are recorded as a data source for improving model performance.

[0081] In the safety communication control module, the process of establishing a two-way communication link between the local remote emergency takeover system and the remote operation station and monitoring the communication status includes:

[0082] Select a satellite communication link, configure the standard transport layer TCP protocol and application layer MQTT protocol to transmit data packets, perform end-to-end encrypted communication through SSL, use digital certificates to verify the identities of the mobile device and remote operator, and use the local remote emergency takeover system on the mobile device to initiate a connection request, and establish a communication session after completing the handshake process;

[0083] A priority scheduling algorithm is used to allocate bandwidth resources for control commands, heartbeat signals are exchanged regularly to confirm the connection status, and CRC verification is implemented to ensure data integrity;

[0084] Build a multi-path communication solution to automatically switch to the backup channel when the main channel fails. The control instructions sent by the remote operation station reach the mobile device via the communication link, where they are parsed and passed to the execution unit. The execution results are monitored and the actual execution status is fed back to the remote operation station to form a closed-loop control, record the instruction interaction and its results, deploy an intrusion detection system to monitor abnormal behavior, and provide a one-click disconnect function to deal with extreme situations.

[0085] First, after starting the system, the data perception fusion module will start working immediately, using multi-sensors such as the global satellite positioning system, inertial measurement unit, lidar and visual sensor to collect real-time positioning data, acceleration, angular velocity and detailed images of the surrounding environment. These data quickly generate high-precision environmental models through synchronous positioning and map construction algorithms, providing intuitive and accurate on-site views for subsequent emergency responses. Next, the data optimization conversion module will perform in-depth processing on these raw data, using multi-sensor fusion algorithms to eliminate redundant and erroneous information and improve the accuracy and stability of the data. This step ensures that subsequent decisions are based on high-quality data. Subsequently, the intelligent decision response module uses advanced deep learning algorithms, such as long short-term memory networks and convolutional neural networks, to perform intelligent analysis on the processed data. It can quickly identify abnormal conditions, assess their potential risks, and autonomously determine whether to start the remote takeover procedure. Once it is determined that takeover is required, the module will quickly generate corresponding control instructions. Finally, the secure communication control module sends the control instructions to the local remote emergency takeover system through a two-way communication link, while monitoring the communication status to ensure accurate transmission of instructions and stable operation of the system. The staff at the remote operation station can also view the on-site environmental model and the execution of control instructions in real time, and make necessary remote interventions and adjustments. The entire process realizes closed-loop management from data collection, optimization processing, intelligent decision-making to remote control, ensuring the system's rapid response and efficient response in emergency situations.

[0086] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions. The sentence "includes an element defined by ... does not exclude the existence of other identical elements in the process, method, article or device including the element".

[0087] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-precision positioning and navigation remote emergency takeover system, characterized by: It includes a safety communication control module, which is communicatively connected to a data perception fusion module, a data optimization conversion module and an intelligent decision response module; The data perception fusion module collects centimeter-level positioning data through the global satellite positioning system, collects acceleration and angular velocity through the inertial measurement unit, and creates an environmental model through laser radar and visual sensors in combination with synchronous positioning and map building algorithms; The data optimization conversion module uses a multi-sensor fusion algorithm to optimize data quality; The intelligent decision response module uses a long short-term memory network combined with a convolutional neural network algorithm to identify abnormal conditions and assess their severity, determine whether it is necessary to start a remote takeover procedure and generate control instructions; The safety communication control module establishes a two-way communication link between the local remote emergency takeover system and the remote operation station, and monitors the communication status.

2. According to claim 1, a high-precision positioning and navigation remote emergency takeover system is characterized by: In the data perception fusion module, the process of collecting centimeter-level positioning data through the global satellite positioning system includes: Installing an enhanced global positioning system on the mobile device, the global positioning system receiver determines the three-dimensional position of the mobile device by receiving satellite signals; Unify the raw GPS data into RINEX format, identify and exclude abnormal observations caused by multipath and ionospheric delay, and perform time synchronization and error correction; Using carrier phase observation, differential measurement is performed through the differential technology RTK to solve the integer cycle number in the carrier phase measurement, centimeter-level positioning data is collected, and differential operations are performed using continuous time series positioning data. The rate of change of the distance between two points is calculated to obtain the velocity, and the direction of the velocity is determined by the azimuth angles of adjacent position points. The global satellite positioning system receiver calculates different types of geometric precision factors DOP based on the received satellite signals. The geometric precision factors include position DOP, horizontal DOP and vertical DOP to evaluate the quality of positioning data. The global satellite positioning system reflection signal analysis technology is used to detect the signal reflection surface information. By comparing the signal characteristics of the direct path and the reflection path, the ground type and the height of the surrounding buildings are inferred.

3. A high-precision positioning and navigation remote emergency takeover system according to claim 2, characterized in that: In the data perception fusion module, the process of collecting acceleration and angular velocity through the inertial measurement unit includes: Install an inertial measurement unit at the center of gravity of the mobile device to reduce the influence of centrifugal force and Coriolis force caused by rotation, and perform temperature compensation and calibration to eliminate installation errors; The accelerometer in the inertial measurement unit detects the displacement of the mobile device relative to the accelerometer housing, collects the acceleration, and uses the gyroscope in the inertial measurement unit to measure the angular velocity using the principle of optical interference; Time-synchronize acceleration data with angular rate data, apply low-pass filters to remove high-frequency noise, integrate inertial measurement unit readings using Kalman filtering techniques to reduce drift errors, adjust accelerometer and gyroscope outputs to offset the effects of temperature changes, and perform static and dynamic calibrations to correct for zero bias, scale factor errors, and inter-axis non-orthogonality; The mobile device velocity data is obtained by integrating the acceleration data, and the mobile device position data is obtained by integrating it again. The angular velocity data of the gyroscope and the gravity direction indication provided by the accelerometer are integrated, and the pitch, roll and heading angles are calculated using a complementary filtering algorithm. The time series data of the inertial measurement unit is analyzed to identify different motion modes, including acceleration, deceleration and turning.

4. A high-precision positioning and navigation remote emergency takeover system according to claim 3, characterized in that: In the data perception fusion module, the process of creating an environment model by combining laser radar and visual sensors with a synchronous positioning and map building algorithm includes: Install a lidar on top of the mobile device to provide 360-degree coverage, and install an RGB-D camera to work with the lidar to provide complementary visual information; The LiDAR calculates the distance of the mobile device by emitting laser pulses and measuring the reflection time, generating point cloud data to analyze the position and shape of the mobile device in the environment. The LiDAR captures image sequences through an RGB-D camera, and uses computer vision algorithms to extract feature points, line segments, and distances of pixels in the image to the RDB-D camera. Aligning the lidar point cloud data and the visual image data in space and time, expressing them in the same coordinate system, applying filters to remove outliers in the lidar point cloud data and noise in the visual image data, identifying natural features from the visual image data, the natural features including corners and edges, and correcting the deviation of the point cloud data and the image data caused by the movement of the mobile device based on the posture information provided by the inertial measurement unit; Extract plane and cylindrical elements from LiDAR point cloud data, extract straight line and circle features from visual image data to model static environment structures, analyze color and brightness changes in visual image data, and identify surface textures; Combine LiDAR data and visual image data to distinguish between static background and moving objects, use SLAM algorithm to integrate continuous LiDAR point cloud data and visual image data features, build a map representation, which includes global map and local map updates, match current observation features with existing maps, and update the position and direction of mobile devices in real time.

5. A high-precision positioning and navigation remote emergency takeover system according to claim 4, characterized in that: In the data optimization conversion module, the process of optimizing data quality using a multi-sensor fusion algorithm includes: By combining the absolute position information provided by the global satellite positioning system with the relative kinematic parameters of the inertial measurement unit, the global satellite positioning system and the inertial measurement unit are combined to realize real-time estimation of the attitude, velocity and position state vector of the mobile device, and the prediction error covariance matrix is ​​adjusted online. The environmental features obtained by the lidar and RGB-D camera are introduced, and the SLAM algorithm is applied to build and update the map representation to achieve self-positioning.

6. A high-precision positioning and navigation remote emergency takeover system according to claim 5, characterized in that: In the intelligent decision response module, the process of using a convolutional neural network algorithm to identify abnormal conditions and assess their severity includes: Preprocess and fuse the collected centimeter-level positioning data, acceleration, angular velocity, point cloud data, and image data to synchronize the collected data in time and align them in space, forming a unified spatiotemporal feature sequence as the input of the convolutional neural network model and the long short-term memory network model; A convolutional neural network model is constructed, which includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer. The input layer receives image data captured by an RGB-D camera. The convolution layer extracts edge and corner features in the image through a convolution operation, uses a filter to capture local features at different scales and directions, and adopts a ReLU activation function to enhance nonlinear expression capabilities. The pooling layer reduces the spatial dimension of the feature map through a maximum pooling operation, reduces the amount of calculation and keeps the extracted features unchanged. The fully connected layer flattens the extracted features into a one-dimensional vector, performs high-level feature combination, and outputs a fixed-length feature representation. The output layer outputs a visual feature vector for describing the environment structure and identifying objects.

7. A high-precision positioning and navigation remote emergency takeover system according to claim 6, characterized in that: In the intelligent decision response module, the process of identifying abnormal conditions and assessing their severity in combination with the long short-term memory network algorithm includes: Constructing a long short-term memory network model, wherein the long short-term memory network model includes an input layer, an LSTM layer and an output layer; The input layer receives acceleration and angular velocity time series data from the inertial measurement unit, as well as the speed and position change sequence of the global satellite positioning system. The LSTM layer learns long-term dependencies through its internal LSTM unit and forgets irrelevant information. The LSTM unit includes an input gate, a forget gate, and an output gate. By controlling the information flow, the long time series data is processed by the network. The high-level time series features are abstracted layer by layer by stacking LSTM layers, and the output layer is used to output the time series feature vector containing the mobile device behavior pattern and path prediction.

8. The high-precision positioning and navigation remote emergency takeover system according to claim 7, characterized in that: In the intelligent decision response module, a hybrid architecture model combining a long short-term memory network and a convolutional neural network is constructed to identify abnormal conditions and assess their severity, including: The visual feature vector from the convolutional neural network and the time series feature vector generated by the LSTM are concatenated into a fused feature representation, and the attention mechanism is introduced to enable the convolutional neural network model and the LSTM model to dynamically adjust the importance weight of the input data; Combined with the map constructed by SLAM and the precise positioning information of mobile devices, the fused features can reflect the latest situational changes. Based on the fused features, the risk score is calculated to share the probability of abnormal events occurring in the current state and their potential threat level. Different weights are assigned to different types of abnormalities, and the context information is analyzed for adjustment. The softmax function is used to output the probability distribution and risk score of normal operation, minor abnormalities, and severe abnormalities, and the emergency response procedure is initiated.

9. A high-precision positioning and navigation remote emergency takeover system according to claim 8, characterized in that: In the intelligent decision response module, the process of determining whether to start the remote takeover program and generate a control instruction includes: According to the output abnormality type and risk score, the remote emergency takeover system automatically generates an emergency response strategy, which includes emergency braking instructions, activation of steering correction mechanism and remote operator intervention. When an obstacle is detected suddenly in front of the mobile device and the distance is too close, an emergency braking instruction is issued. When the mobile device deviates from the planned route and is about to enter the dangerous area, the steering correction mechanism is activated. When the abnormal condition exceeds the threshold and conventional corrective measures cannot be effectively dealt with, it is determined whether the remote emergency takeover program needs to be activated, the remote operator intervenes, and sends control instructions to the mobile device. The emergency response results are recorded as a data source for improving model performance.

10. A high-precision positioning and navigation remote emergency takeover system according to claim 9, characterized in that: In the safety communication control module, the process of establishing a two-way communication link between the local remote emergency takeover system and the remote operation station and monitoring the communication status includes: Select a satellite communication link, configure the standard transport layer TCP protocol and application layer MQTT protocol to transmit data packets, perform end-to-end encrypted communication through SSL, use digital certificates to verify the identities of the mobile device and remote operator, and use the local remote emergency takeover system on the mobile device to initiate a connection request, and establish a communication session after completing the handshake process; A priority scheduling algorithm is used to allocate bandwidth resources for control commands, heartbeat signals are exchanged regularly to confirm the connection status, and CRC verification is implemented to ensure data integrity; Build a multi-path communication solution to automatically switch to the backup channel when the main channel fails. The control instructions sent by the remote operation station reach the mobile device via the communication link, where they are parsed and passed to the execution unit. The execution results are monitored and the actual execution status is fed back to the remote operation station to form a closed-loop control, record the instruction interaction and its results, deploy an intrusion detection system to monitor abnormal behavior, and provide a one-click disconnect function to deal with extreme situations.

Images