A method and system for maintaining a reference based on unified space-time

Abstract

The present invention relates to the field of satellite navigation and surveying technology, and specifically to a reference maintenance method and system based on unified space-time, including: obtaining real-time positioning signals of a first satellite navigation system and a second satellite navigation system at a specified position, generating unified space-time reference data through a CORS network and alignment, and sending it to distributed reference nodes in the tunnel to generate a consistent distributed space-time reference covering the tunnel range; after receiving by the vehicle terminal, using the time reference and coordinate reference to synchronously correct the sampling time and pose solution of the local sensor, and generating a vehicle pose reference; if the positioning accuracy continuously deviates from a preset threshold, sending an error correction request to an adjacent distributed reference node; the adjacent node generates an updated reference by combining historical cache data and a collaborative compensation mechanism, and performs fusion calculation by combining data collected by the local sensor to obtain accurate vehicle positioning and three-dimensional surveying. It helps to improve the positioning accuracy and navigation performance in the tunnel.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite navigation and surveying and mapping, and specifically to a reference maintenance method and system based on unified space-time. Background Art

[0002] In the modern urban environment, high-precision positioning, navigation, and timing services have become indispensable support technologies in fields such as intelligent transportation, emergency rescue, and precision surveying and mapping. As an important part of the global satellite navigation system, the Beidou satellite navigation system has wide applications in urban scenarios due to its technical advantages of high precision, global coverage, and multi-frequency signals. However, the urban environment is complex and changeable, and the factors affecting the performance of the Beidou system are also relatively prominent.

[0003] In the urban environment, when vehicles are driving in complex scenarios (such as tunnels, viaducts, and underground parking lots), the stability and reliability of Beidou satellite signals are often difficult to guarantee. Among them, in the tunnel environment, due to terrain occlusion and spatial enclosure, Beidou satellite signals may be completely blocked, and the receiving device cannot obtain direct path signals, resulting in the interruption of positioning information. In addition, the signal propagation in the tunnel is also easily affected by the multipath effect of the vehicle's metal structure and the tunnel wall surface. The path received after signal reflection may be longer, further exacerbating the positioning error.

[0004] To address these problems, the Beidou system usually combines ground-based augmentation technology. By establishing a network of CORS stations (Continuously Operating Reference Stations), high-precision positioning and timing services are provided for users. CORS stations receive satellite signals and generate differential correction data in combination with ground reference equipment, and provide correction information to users in real time, significantly improving the positioning accuracy. At the same time, CORS stations can also monitor the quality of satellite signals and provide a reliable time reference for users. However, in the tunnel scenario, the satellite signal sources relied on by the CORS station network are difficult to obtain stably due to occlusion problems, resulting in the inability to continuously provide a unified space-time reference for users. In addition, the special environment in the tunnel makes it difficult for existing ground-based augmentation technologies to effectively solve the problems of signal attenuation and multipath errors, affecting the continuous positioning and navigation performance of vehicles in the tunnel.

[0005] Therefore, a reference maintenance method and system based on unified space-time are proposed. Summary of the Invention

[0006] The object of the present invention is to provide a reference maintenance method and system based on unified space-time, which are applicable to high-precision positioning and navigation in complex tunnel environments, including: obtaining real-time positioning signals of a first satellite navigation system and a second satellite navigation system at a specified position, generating differential correction data through a CORS network, and aligning according to the national geodetic coordinate system to generate unified space-time reference data; sending the reference data to distributed reference nodes in the tunnel to form a consistent distributed space-time reference covering the tunnel range; after receiving the reference, the vehicle terminal synchronously corrects the local sensor data using the time reference and the coordinate reference to generate a vehicle pose reference; if the positioning accuracy deviates from a preset threshold, an error correction request is sent to adjacent nodes through a feedback mechanism, and the adjacent nodes generate an updated reference by combining historical cache data and a collaborative compensation mechanism, and fuse and calculate with the sensor data to obtain high-precision vehicle positioning and three-dimensional mapping results. The present invention also optimizes the scoring of reference nodes through a dynamic weight learning mechanism, improving the reference synchronization stability and navigation performance in complex scenarios.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A reference maintenance method based on unified space-time, including:

[0009] Obtaining real-time positioning signals from a first satellite navigation system and a second satellite navigation system at a specified position, and obtaining differential correction data through a CORS network deployed on the ground;

[0010] Aligning the differential correction data according to the national geodetic coordinate system to obtain unified space-time reference data;

[0011] Sending the unified space-time reference data to distributed reference nodes in a specified tunnel, and generating a consistent distributed space-time reference covering the entire range of the specified tunnel with the reference point at the specified position as the starting point to obtain a consistent distributed space-time reference in the tunnel; the consistent distributed space-time reference in the tunnel includes a time reference and a coordinate reference;

[0012] After the vehicle terminal receives the consistent distributed space-time reference in the tunnel, synchronously correcting the sampling time and pose solution of the local sensor using the time reference and the coordinate reference to obtain a vehicle pose reference; if it is detected that the positioning accuracy of the vehicle pose reference deviates from a preset threshold in N consecutive sampling periods, an error correction request is sent to the adjacent distributed reference nodes through a feedback mechanism; the local sensor includes at least one of an inertial measurement unit, a vision module, a lidar, and an odometer;

[0013] Receive the updated distributed spatio-temporal reference generated by the adjacent distributed reference nodes, and perform fusion calculation by combining the real-time data collected by the local sensors to obtain vehicle precise positioning and vehicle three-dimensional mapping; the updated distributed spatio-temporal reference is generated by the adjacent distributed reference nodes after receiving the error correction request, in combination with historical cache data and a collaborative compensation mechanism.

[0014] Further, the specified position is the entrance of the specified tunnel.

[0015] Further, the first satellite navigation system is specifically the Beidou satellite navigation system.

[0016] Further, the second satellite navigation system is at least one of the GPS satellite navigation system, the GLONASS satellite navigation system, and the Galileo satellite navigation system.

[0017] Further, the distributed reference nodes include multiple movable reference nodes arranged in the specified tunnel, and the specific composition of each movable reference node includes:

[0018] A high-precision local clock module, which is used to maintain the time reference when there is no external signal input, and online monitor the clock deviation by combining historical drift data to obtain a stable time reference under the condition of external signal loss; when receiving the high-precision time calibration information from the GNSS receiving module, update the stable time reference to obtain an updated time reference; and provide the updated time reference as the node time reference to the communication and signal forwarding module and the adaptive clock drift compensation module;

[0019] The GNSS receiving module is used to periodically receive satellite signals at the entrance of the specified tunnel to obtain periodic satellite signals; calculate calibration parameter information according to the periodic satellite signals, and the calibration parameter information includes satellite orbit correction parameters and clock error correction parameters; calibrate the high-precision local clock module according to the periodic satellite signals and the calibration parameter information to obtain the high-precision time calibration information aligned with the high-precision local clock;

[0020] The communication and signal forwarding module is used to receive the unified spatio-temporal reference data of the adjacent movable reference nodes, and the unified spatio-temporal reference data includes time information, coordinate information, and signal quality data, and at the same time receive the motion state information of the vehicle terminal; according to the preset synchronization strategy, dynamically weighted fusion of the clock deviation and signal delay between the movable reference nodes, and optimize the coordinate reference by combining the extended Kalman filter algorithm to obtain a distributed spatio-temporal reference; calculate the node quality score according to the signal quality data:

[0021] ;

[0022] Among them, represents the node quality score, represents the signal strength, represents the communication delay, represents the link stability, is the first weight parameter, is the second weight parameter, is the third weight parameter;

[0023] Select the movable reference node with the highest node quality score as the priority target node, send a synchronization instruction to the priority target node, and receive the calibration data returned by the priority target node; update the distributed spatio-temporal reference according to the calibration data to generate a preliminary calibration result; when the communication link is not interrupted and the priority target node does not fail, use the preliminary calibration result as the consistent distributed spatio-temporal reference in the tunnel; and send the distributed spatio-temporal reference to the vehicle terminal;

[0024] The adaptive clock drift compensation module is used to dynamically correct the local clock drift model according to the mutual comparison results between the movable reference nodes and environmental changes. The update process includes calculating the real-time drift rate according to the time deviation of multiple nodes, and optimizing the assignment of the time reference of each node by combining the credibility weight mechanism;

[0025] The environment perception module is used to detect the multipath effect distribution, shielding area range, interference source position, dynamic environment state and physical parameter change results in the specified tunnel to obtain detection results; and feedback the detection results to the adaptive clock drift compensation module;

[0026] The redundant data cache and fault takeover module is used to re-evaluate the node quality scores of the remaining nodes when the communication link is interrupted or the priority target node fails, select the movable reference node with the second highest node quality score as the updated priority target node, and dynamically correct the preliminary calibration result by combining the historical cache data and the cooperative compensation mechanism of adjacent nodes to generate an updated calibration result; use the updated calibration result to recalculate and generate a distributed spatio-temporal reference, and the distributed spatio-temporal reference includes the corrected time reference, the corrected coordinate reference and the synchronization status information; and send the distributed spatio-temporal reference to the vehicle terminal and the adjacent movable reference nodes.

[0027] Furthermore, the first weight parameter and the second weight parameter and the third weight parameter are adjusted according to the dynamic weight learning mechanism, specifically including:

[0028] Construct a state vector based on the signal strength, communication delay, link stability, historical performance, and environmental interference intensity of each of the movable reference nodes;

[0029] Input the state vector of each of the movable reference nodes, combined with historical data and real-time environmental parameters, into a deep reinforcement learning model, which adopts an Actor-Critic architecture and includes an Actor network and a Critic network;

[0030] Among them, the Actor network is used to input the state vector and output a weight increment, and the weight increment includes a first weight increment , a second weight increment and a third weight increment ;

[0031] The Critic network is used to evaluate the performance of the current weight allocation scheme and optimize the updates of the first weight parameter, the second weight parameter, and the third weight parameter through a reward function;

[0032] The current weight allocation scheme is:

[0033] ;

[0034] The reward function is defined by the following formula:

[0035] ;

[0036] Among them, is the reward function; is the reduced value of the positioning error after update; is the improvement amplitude of the reference synchronization accuracy after update; is the improvement of the environmental adaptability after update; is the first reward weight coefficient; is the second reward weight coefficient; is the third reward weight coefficient.

[0037] A reference maintenance system based on unified space-time, comprising:

[0038] A signal acquisition module, configured to acquire real-time positioning signals from a first satellite navigation system and a second satellite navigation system at a specified position, and obtain differential correction data through a CORS network deployed on the ground; a reference alignment module, configured to align the differential correction data according to the national geodetic coordinate system to obtain unified space-time reference data;

[0039] A reference distribution module for sending the unified spatio-temporal reference data to distributed reference nodes in a specified tunnel, and generating a consistent distributed spatio-temporal reference covering the entire range of the specified tunnel with the reference point at the specified position as the starting point, to obtain a consistent distributed spatio-temporal reference within the tunnel;

[0040] A terminal calibration module for, after the vehicle terminal receives the consistent distributed spatio-temporal reference within the tunnel, using the consistent distributed spatio-temporal reference within the tunnel to synchronously calibrate the sampling time and pose solution of the local sensor, to obtain a vehicle pose reference;

[0041] A feedback mechanism module for, if it is detected that the positioning accuracy of the vehicle pose reference deviates from a preset threshold for N consecutive sampling periods, sending an error correction request to the adjacent distributed reference nodes through a feedback mechanism;

[0042] An updated reference fusion module for receiving the updated distributed spatio-temporal reference generated by the adjacent distributed reference nodes, and performing fusion calculation in combination with the real-time data collected by the local sensor, to obtain vehicle precise positioning and vehicle three-dimensional mapping; the updated distributed spatio-temporal reference is generated by the adjacent distributed reference nodes after receiving the error correction request, in combination with historical cache data and a collaborative compensation mechanism.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] 1. The present invention constructs a consistent distributed spatio-temporal reference covering the entire range of the tunnel by arranging distributed reference nodes in the tunnel and combining the differential correction data generated by the CORS network and the unified alignment of the national geodetic coordinate system, effectively solving the problem of reduced positioning accuracy caused by satellite signal occlusion in the tunnel scenario. After the vehicle terminal receives the spatio-temporal reference, it calibrates the local sensor data by combining the time reference and the coordinate reference, improving the positioning accuracy and data synchronization; when the positioning accuracy is abnormal, it interacts with adjacent nodes through a feedback mechanism to dynamically generate an updated reference, and performs fusion calculation in combination with multi-sensor data, ensuring the continuous high-precision positioning and three-dimensional mapping capabilities of the vehicle in a complex environment. The present invention significantly improves the navigation performance and the reliability of the spatio-temporal reference in special environments such as tunnels, providing strong support for high-precision positioning and mapping technologies.

[0045] 2. The present invention realizes the high-precision maintenance of a consistent distributed spatio-temporal reference in a tunnel by deploying multiple movable distributed reference nodes in the tunnel, in combination with a high-precision local clock module, a GNSS receiving module, and an adaptive clock drift compensation module. The priority target nodes are dynamically selected through a node quality scoring mechanism, and the generation and update of the distributed spatio-temporal reference are optimized by combining historical cache data and a collaborative compensation mechanism. When the communication link is interrupted or a node fails, the system function can be quickly restored through a redundant data cache and a fault takeover module. The environment perception module further enhances the system's adaptability to complex environments, significantly improving the positioning accuracy, spatio-temporal reference synchronization stability, and navigation performance in complex shielding environments such as tunnels, providing a reliable technical guarantee for high-precision navigation and three-dimensional mapping.

[0046] 3. The present invention realizes the adaptive optimization of the quality scoring of movable reference nodes by introducing a dynamic weight learning mechanism. By constructing the state vector of the nodes, combining historical data and real-time environmental parameters, and using a deep reinforcement learning model to dynamically adjust the first weight parameter, the second weight parameter, and the third weight parameter, the node quality scoring becomes more accurate and adapts to the dynamic changes of complex environments. The Actor-Critic architecture ensures the rationality of the weight increment and the efficiency of reward optimization, and dynamically balances the improvement of positioning accuracy, reference synchronization stability, and environmental adaptability through a reward function. This method significantly enhances the reliability of distributed reference nodes and the stability of spatio-temporal references in complex scenarios such as tunnels, helps to improve the positioning continuity and navigation performance of vehicle terminals, reduces the accuracy degradation caused by signal interference or environmental changes, and provides a solid technical guarantee for high-precision navigation and surveying. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flowchart of a reference maintenance method based on unified space-time provided by an embodiment of the present invention;

[0048] Figure 2 It is a system structure diagram of a movable reference node provided by an embodiment of the present invention;

[0049] Figure 3 It is a system structure diagram of a reference maintenance system based on unified space-time provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] Please refer toFigures 1 to 3 , the present invention provides a reference maintenance method and system based on unified space-time, and the technical solution is as follows:

[0052] Embodiment 1

[0053] A special surveying vehicle is equipped with high-precision sensors such as lidar, cameras, inertial measurement units, and odometers. Its task is to collect structural data within a specified tunnel and generate a complete three-dimensional model of the tunnel. The special surveying vehicle starts from the entrance position (specified position) of the specified tunnel and travels uniformly along the tunnel trajectory until the tunnel exit, performing three-dimensional surveying tasks throughout the tunnel.

[0054] A reference maintenance method based on unified space-time, as Figure 1 shown, the specific process is as follows:

[0055] Referring to Figure 1 S10 in, after the vehicle starts, when the vehicle travels to the specified position, it receives real-time positioning signals from the first satellite navigation system and the second satellite navigation system through the vehicle terminal device, and obtains differential correction data through the CORS network deployed on the ground; further, the specified position is the entrance of the specified tunnel. Setting the specified position as the entrance of the tunnel can ensure the efficient generation and accurate distribution of the unified space-time reference. By obtaining real-time positioning signals from the first satellite navigation system and the second satellite navigation system at the entrance and combining with the CORS network to generate differential correction data, a high-precision time reference and coordinate reference can be established before entering the tunnel, serving as the reference point for the entire tunnel range. This not only simplifies the system deployment and management but also effectively avoids the impact of signal occlusion in the tunnel on reference generation, thereby ensuring the continuity, consistency, and high precision of the space-time reference for the entire tunnel. In addition, this design can also provide a reliable starting point for generating consistent distributed space-time references for subsequent distributed reference nodes, improving the efficiency and accuracy of reference transmission, and laying a solid foundation for vehicle positioning and navigation in the tunnel.

[0056] Further, the first satellite navigation system is specifically the Beidou satellite navigation system, and the second satellite navigation system is at least one of the GPS satellite navigation system, the GLONASS satellite navigation system, and the Galileo satellite navigation system. Further, by using the Beidou satellite navigation system as the first satellite navigation system and combining at least one of the GPS, GLONASS, or Galileo satellite navigation systems as the second satellite navigation system, the global coverage characteristics and high-precision positioning capabilities of multiple satellite systems can be fully utilized to improve the positioning stability and anti-interference capabilities of the system. The Beidou system has the advantages of high precision and high reliability in the Asia-Pacific region, while the GPS, GLONASS, and Galileo systems provide global supplementary and redundant support. Through the combined design of multiple satellite systems, richer satellite signals can be obtained at the tunnel entrance, enhancing the anti-occlusion ability and effectively reducing the risk of a decrease in positioning accuracy caused by signal occlusion or errors in a single system. In addition, multi-system fusion can improve the accuracy and reliability of reference generation through differential correction algorithms, providing a more stable spatio-temporal reference for subsequent distributed reference nodes and meeting the requirements of high-precision positioning and navigation in complex environments inside tunnels. This design achieves the unity of high precision, continuity, and global adaptability, and is particularly suitable for positioning applications in complex occlusion environments such as tunnels.

[0057] The reference point at the first position is specifically a fixed reference station integrating a high-precision GNSS receiving module, a high-precision local clock module, a data processing module, a communication module, and an environmental perception module. By receiving satellite signals in real time, generating unified spatio-temporal reference data, and reliably transmitting it to the distributed reference nodes inside the tunnel, it provides unified spatio-temporal reference support for the entire tunnel range.

[0058] At the reference point at the specified location, the satellite positioning signal is received by the GNSS receiving module, and high-precision differential correction data is generated in combination with the CORS network to form unified spatio-temporal reference data. The unified spatio-temporal reference data is distributed to the distributed reference nodes in the tunnel through the communication and signal forwarding module using a low-latency and highly reliable communication protocol. After receiving the reference data, each reference node optimizes the coordinate reference through the extended Kalman filter algorithm in combination with the signal quality, time reference, and spatial position information received by itself, and at the same time adjusts the time deviation and signal delay with adjacent nodes using a dynamic synchronization strategy to generate a distributed spatio-temporal reference. In order to achieve full-tunnel coverage, each reference node establishes a cooperative network based on the node quality score, optimizes the reference transfer process through a dynamic weighted fusion mechanism, and gradually propagates the reference data to the deep part of the tunnel. During the process, the redundant data caching and fault takeover module monitors the node status in real time. When a node fails or the communication is interrupted, it quickly switches to the node with the second-highest quality score to ensure that the reference transfer is not interrupted. Finally, the reference nodes work together to generate a distributed spatio-temporal reference with good consistency for the entire tunnel range and send it to the vehicle terminals in the tunnel to achieve high-precision spatio-temporal synchronization and navigation positioning.

[0059] Refer to Figure 1 In S20 of , align the differential correction data according to the national geodetic coordinate system to obtain unified spatio-temporal reference data;

[0060] Refer to Figure 1 In S30 of , send the unified spatio-temporal reference data to the distributed reference nodes in the specified tunnel, and generate a consistent distributed spatio-temporal reference covering the entire range of the specified tunnel starting from the reference point at the specified location to obtain a consistent distributed spatio-temporal reference in the tunnel;

[0061] Furthermore, the distributed reference nodes include multiple movable reference nodes arranged in the specified tunnel, as Figure 2 shown. The specific composition of each movable reference node includes:

[0062] A high-precision local clock module, which is used to maintain the time reference when there is no external signal input, and online monitor the clock deviation in combination with historical drift data to obtain a stable time reference under the condition of external signal loss; when receiving high-precision time calibration information from the GNSS receiving module, update the stable time reference to obtain an updated time reference; and provide the updated time reference as the node time reference to the communication and signal forwarding module and the adaptive clock drift compensation module;

[0063] The GNSS receiving module is used to periodically receive satellite signals at the entrance of the specified tunnel to obtain periodic satellite signals; calculate correction parameter information according to the periodic satellite signals, where the correction parameter information includes satellite orbit correction parameters and clock error correction parameters; calibrate the high-precision local clock module according to the periodic satellite signals and the correction parameter information to obtain the high-precision time calibration information aligned with the high-precision local clock;

[0064] The communication and signal forwarding module is used to receive the unified spatio-temporal reference data of the adjacent movable reference nodes, where the unified spatio-temporal reference data includes time information, coordinate information, and signal quality data, and at the same time receive the motion state information of the vehicle terminal; according to the preset synchronization strategy, dynamically weighted fusion of the clock deviation and signal delay between the movable reference nodes is performed, and the coordinate reference is optimized by combining the extended Kalman filter algorithm to obtain a distributed spatio-temporal reference; according to the signal quality data, calculate the node quality score:

[0065] ;

[0066] where, represents the node quality score, represents the signal strength, represents the communication delay, represents the link stability, is the first weight parameter, is the second weight parameter, is the third weight parameter;

[0067] Select the movable reference node with the highest node quality score as the priority target node, send a synchronization instruction to the priority target node, and receive the correction data returned by the priority target node; update the distributed spatio-temporal reference according to the correction data to generate a preliminary correction result; when the communication link is not interrupted and the priority target node does not fail, use the preliminary correction result as the consistent distributed spatio-temporal reference in the tunnel; and send the distributed spatio-temporal reference to the vehicle terminal;

[0068] The adaptive clock drift compensation module is used to receive the updated time reference, dynamically correct the local clock drift model according to the mutual comparison results between the movable reference nodes and environmental changes, and the update process includes calculating the real-time drift rate according to the time deviation of multiple nodes and optimizing the assignment of the time reference of each node by combining the credibility weight mechanism;

[0069] The environmental perception module is used to detect the multipath effect distribution, shielding area range, interference source location, dynamic environmental state, and physical parameter change results in the specified tunnel, obtain the detection results, and feedback the detection results to the adaptive clock drift compensation module;

[0070] The redundant data cache and fault takeover module is used to re-evaluate the node quality scores of the remaining nodes when the communication link is interrupted or the priority target node fails, select the movable reference node with the second-highest node quality score as the updated priority target node, and dynamically correct the preliminary correction result by combining historical cache data and the cooperative compensation mechanism of adjacent nodes to generate an updated correction result. Use the updated correction result to recalculate and generate a distributed spatio-temporal reference, where the distributed spatio-temporal reference includes a corrected time reference, a corrected coordinate reference, and synchronization status information. And send the distributed spatio-temporal reference to the vehicle terminal and the adjacent movable reference node.

[0071] By deploying multiple movable reference nodes in a specified tunnel to construct a distributed reference system, it can effectively address problems such as satellite signal occlusion, multipath effects, and dynamic environmental changes in the tunnel environment. The movable reference node maintains the stability of the time reference through a high-precision local clock module when the signal is missing and is updated in real time in combination with GNSS time calibration information to ensure the high precision and reliability of the reference. At the same time, the communication and signal forwarding module optimizes the coordinate reference according to the dynamic weighted fusion and extended Kalman filtering algorithms through the cooperative work of adjacent nodes, effectively improving the accuracy and consistency of the distributed reference. In addition, the adaptive clock drift compensation module dynamically corrects the clock drift according to the time deviation between nodes and environmental changes, and the environmental perception module detects complex interference factors in the tunnel to provide real-time reference for the adaptive adjustment of the reference node. Through the node quality scoring mechanism, the optimal node is preferentially selected for reference update and distribution, further improving the robustness and efficiency of the system. Even in the case of communication link interruption or node failure, the redundant data cache and fault takeover module can still ensure the continuity and accuracy of the reference, thus providing reliable technical support for the high-precision positioning and navigation of vehicles in the tunnel, and realizing the high efficiency and stability of the reference maintenance system in a complex environment.

[0072] Furthermore, the moving mode of the movable reference node is specifically as follows: adopting the joint path planning technology that combines the fusion reinforcement learning algorithm and the multi-objective genetic algorithm, calculating the overall node efficiency value periodically according to the multipath effect distribution, the shielding area range, the interference source position, the dynamic environment state, the physical parameter change result, and the node quality score in the specified tunnel; when the overall node efficiency value is lower than the preset threshold, performing multiple rounds of simulation evaluation on the candidate target positions through online reinforcement learning; during the node movement, globally and locally optimizing the movement path through the multi-objective genetic algorithm, exchanging the environment state and the node quality score with the adjacent movable reference nodes through the real-time communication and signal forwarding module, and performing adaptive fine-tuning on the movement path; when the movable reference node reaches the target position, re-broadcasting its own position, and providing a new spatio-temporal reference and environment change information to the adaptive clock drift compensation module. By adopting the joint path planning technology that combines the fusion reinforcement learning algorithm and the multi-objective genetic algorithm, the movable reference node can achieve intelligent and dynamic deployment in a complex tunnel environment, significantly improving the flexibility and robustness of the system. By calculating the overall node efficiency value periodically, the operating state of the reference node in complex environments such as multipath effect distribution, shielding area range, and interference source position can be evaluated in real time. When the efficiency value is lower than the preset threshold, the system performs multiple rounds of simulation evaluation on the candidate target positions through online reinforcement learning to ensure the optimality of the target position. During the movement, the multi-objective genetic algorithm combines the global path planning and local path optimization capabilities to provide an efficient and accurate movement path for the node. At the same time, through the real-time communication and quality score exchange with adjacent nodes, the path can be adaptively fine-tuned to avoid interference areas and optimize the node distribution. When the node reaches the target position, it can re-broadcast its own position information and update the spatio-temporal reference, effectively covering the new working area. This design not only improves the dynamic adaptation ability of the reference node but also optimizes the overall performance of the reference system, especially ensuring the efficiency, stability, and continuous coverage ability of the distributed reference in a complex tunnel environment, thus providing a reliable guarantee for the precise positioning and navigation of vehicles.

[0073] Furthermore, the first weight parameter and the second weight parameter and the third weight parameter are adjusted according to the dynamic weight learning mechanism, which specifically includes:

[0074] Constructing a state vector according to the signal strength, the communication delay, the link stability, the historical performance, and the environmental interference strength of each movable reference node;

[0075] The state vector of each of the movable reference nodes is input into a deep reinforcement learning model in combination with historical data and real-time environmental parameters. Specifically, in this embodiment, the historical data includes the operation records and performance of each movable reference node, such as the historical signal strength of the node, the communication delay distribution, the change trend of link stability, the task execution success rate, the calibration accuracy, and the fault records. These data can reflect the comprehensive performance of the node in different environments and task scenarios, providing long-term experience support for the reinforcement learning model. The real-time environmental data covers the dynamic state of the current environment where the node is located, such as the real-time signal strength, communication delay, the stability of the current link, the intensity of environmental interference (such as electromagnetic interference and multipath effect), the motion state of neighboring vehicles, the change of physical conditions in the tunnel (such as temperature and humidity), and the reference synchronization state. These data can dynamically reflect the current operating environment of the node, providing a reliable basis for the immediate decision-making of the model.

[0076] The deep reinforcement learning model adopts an Actor-Critic architecture, including an Actor network and a Critic network;

[0077] Among them, the Actor network is used to input the state vector and output a weight increment, and the weight increment includes a first weight increment , a second weight increment and a third weight increment ;

[0078] The Critic network is used to evaluate the performance of the current weight allocation scheme and optimize the update of the first weight parameter, the second weight parameter, and the third weight parameter through a reward function;

[0079] The current weight allocation scheme is:

[0080] ;

[0081] The reward function is defined by the following formula:

[0082] ;

[0083] Among them, is the reward function; is the reduced value of the positioning error after update; is the improvement amplitude of the reference synchronization accuracy after update; is the improvement of the environmental adaptability after update; is the first reward weight coefficient; is the second reward weight coefficient; is the third reward weight coefficient.

[0084] Specifically, , and have both been normalized.

[0085] As an implementation of this embodiment, The calculation formula of

[0086] ;

[0087] where is the signal strength consistency (represented by the standard deviation change of the signal strength of the node at different time periods in the dynamic environment, indicating the improvement degree of signal strength stability); is the error compensation efficiency (the comprehensive ratio of the reduction amount of the positioning accuracy recovery time to the reduction amplitude of the positioning error); is the node response ability (represented by the difference between the time intervals before and after the node adjusts after receiving the change of real-time data, indicating the improvement of the response speed); is 's weight coefficient; is 's weight coefficient; is 's weight coefficient.

[0088] By introducing a dynamic weight learning mechanism to adjust the weight parameters, the system can dynamically optimize the quality score according to multi-dimensional features such as the signal strength, communication delay, link stability, historical performance, and environmental interference intensity of the movable reference node. Through the Actor-Critic architecture in the deep reinforcement learning model, the system can not only evaluate the node operating state in real time but also optimize the update strategy of the weight parameters according to the reward function, thereby improving the adaptability and robustness of the system in complex environments. The reward function comprehensively considers the reduction value of the positioning error, the improvement amplitude of the reference synchronization accuracy, and the improvement of the environmental adaptability, and realizes a fair evaluation of each index through the weighted formula, effectively avoiding the excessive influence of a single index on the scoring result. At the same time, by normalizing the weight parameters, the stability and calculation accuracy of the system in different environments and conditions are further ensured. This design significantly improves the evaluation efficiency and accuracy of the distributed reference node, providing strong technical support for high-precision reference generation and vehicle positioning and navigation in complex tunnel environments.

[0089] Refer to Figure 1In S40, after the vehicle terminal receives the consistent distributed spatio-temporal reference in the tunnel, the sampling time and pose solution of the local sensor are synchronously corrected using the consistent distributed spatio-temporal reference in the tunnel to obtain the vehicle pose reference; if it is detected that the positioning accuracy of the vehicle pose reference deviates from the preset threshold for N consecutive (N is a positive integer) sampling periods, an error correction request is sent to the adjacent distributed reference node through a feedback mechanism; the local sensor includes at least one of an inertial measurement unit, a vision module, a lidar, and an odometer;

[0090] Specifically, in this embodiment, the preset position error threshold is 10 cm, the sampling frequency is 10 Hz, and N (N = 5) consecutive periods (0.5 seconds) are used as the judgment condition. If the system detects that the vehicle position error exceeds 10 cm in the last 5 samplings, an error correction request is triggered.

[0091] Refer to Figure 1 In S50, the updated distributed spatio-temporal reference generated by the adjacent distributed reference node is received, and fusion calculation is performed in combination with the real-time data collected by the local sensor to obtain vehicle precise positioning and vehicle three-dimensional mapping; the updated distributed spatio-temporal reference is generated by the adjacent distributed reference node after receiving the error correction request, in combination with historical cache data and a collaborative compensation mechanism.

[0092] To verify the performance of the reference maintenance method based on unified space-time, experiments were carried out in an actual tunnel environment. Three tunnels with lengths of 500 meters, 1 kilometer, and 2 kilometers were selected for the experiment. There are complex environmental factors such as multipath effects, signal occlusion, and interference sources in the tunnel. The vehicle is equipped with high-precision sensors such as an inertial measurement unit, a vision module, a lidar, and an odometer, and travels along the tunnel at a fixed speed, and the actual path is recorded. The experiment compares the traditional CORS reference method, the static weight allocation method, and the dynamic weight learning method, and tests the positioning accuracy, reference synchronization accuracy, and environmental adaptability respectively. During the test, the position information of the vehicle, the synchronization status of the reference node, and the fault recovery time are collected in real time, and the positioning error is analyzed by comparing with the real path. The experimental results are presented in tabular form, including key indicators such as the mean positioning error, synchronization accuracy, and recovery time of different methods under tunnel length, environmental interference level, and node fault conditions.

[0093] Table 1 Experimental result data table

[0094]

[0095] Among them, in Experimental Group 1, a reference maintenance method based on unified space-time proposed by the present invention was applied, as Figure 1As shown, the control group applied a traditional CORS reference system: a traditional CORS reference maintenance method using a single static reference node and fixed weight parameters. The control group 2 adopted an improved method based on static weight allocation: also deploying multiple reference nodes inside the tunnel, but different from the experimental group, it adopted static weight allocation, and the first weight parameter 、the second weight parameter and the third weight parameter will not be adjusted according to the dynamic weight learning mechanism.

[0096] Referring to Table 1, the fault recovery time is the total time from when the system detects a problem until the vehicle terminal regains an accurate and synchronized spatio-temporal reference. Experiments show that the experimental group has significant performance advantages in complex tunnel environments, can better adapt to multipath effects and interference conditions, the positioning accuracy and reference synchronization accuracy are significantly better than other methods, and the fault recovery time is greatly shortened, demonstrating its superiority in tunnel navigation scenarios.

[0097] In tunnels, traditional positioning methods often fail due to satellite signal blockage. This method uses distributed reference nodes to extend the unified spatio-temporal reference data to the entire tunnel range, achieving full coverage of the positioning blind area in the tunnel and ensuring the continuous positioning ability of vehicles in the tunnel. Using the differential correction data and reference alignment mechanism of the CORS network, it provides a high-precision spatio-temporal reference. Combining multi-sensor data at the vehicle end (such as inertial measurement units, lidar, and vision modules, etc.), the positioning accuracy is further improved through fusion calculation to meet the requirements of tunnel surveying and high-precision navigation. When the vehicle positioning accuracy continuously deviates from the preset threshold, the system sends an error correction request to adjacent reference nodes through a feedback mechanism to update the distributed spatio-temporal reference in real time. Combining historical cache data and a collaborative compensation mechanism, the accuracy of the reference data is dynamically adjusted to ensure the robustness and stability of the system in complex environments. By synchronously correcting sensor data through a consistent distributed spatio-temporal reference, the problems of inconsistent sampling times and pose solution errors of different sensors are solved, ensuring the unity and reliability of multi-sensor data, and providing a solid foundation for vehicle precise positioning and three-dimensional surveying. The design of the present invention is flexible and can be widely applied to scenarios such as tunnel surveying, traffic navigation, vehicle dispatching, etc., providing technical support for tunnel structure monitoring, intelligent transportation, and high-precision navigation.

[0098] Example Two

[0099] An automatic navigation vehicle is used for precise positioning and navigation in a complex tunnel environment. Its task is to complete high-precision path tracking through the tunnel, while recording the vehicle's driving path for traffic optimization and tunnel safety management.

[0100] This vehicle applied a reference maintenance system based on unified space-time, such asFigure 3 As shown, it includes:

[0101] A signal acquisition module, configured to acquire real-time positioning signals from a first satellite navigation system and a second satellite navigation system at a specified position, and obtain differential correction data through a CORS network deployed on the ground; a reference alignment module, configured to align the differential correction data according to the national geodetic coordinate system to obtain unified spatio-temporal reference data;

[0102] A reference distribution module, configured to send the unified spatio-temporal reference data to distributed reference nodes in a specified tunnel, and generate a consistent distributed spatio-temporal reference covering the entire range of the specified tunnel with the reference point at the specified position as the starting point, to obtain a consistent distributed spatio-temporal reference in the tunnel;

[0103] A terminal calibration module, configured to, after the vehicle terminal receives the consistent distributed spatio-temporal reference in the tunnel, synchronously calibrate the sampling time and pose solution of the local sensor by using the consistent distributed spatio-temporal reference in the tunnel to obtain a vehicle pose reference;

[0104] A feedback mechanism module, configured to, if it is detected that the positioning accuracy of the vehicle pose reference deviates from a preset threshold in N consecutive sampling periods, send an error correction request to the adjacent distributed reference node through the feedback mechanism;

[0105] An updated reference fusion module, configured to receive the updated distributed spatio-temporal reference generated by the adjacent distributed reference node, and perform fusion calculation by combining the real-time data collected by the local sensor to obtain vehicle precise positioning and vehicle three-dimensional mapping; the updated distributed spatio-temporal reference is generated by the adjacent distributed reference node after receiving the error correction request, in combination with historical cache data and a collaborative compensation mechanism.

[0106] When the system runs, it can achieve all the same functions as the method described in Embodiment 1.

[0107] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A benchmark maintenance method based on unified space-time, characterized in that: include: Acquire real-time positioning signals from the first satellite navigation system and the second satellite navigation system at a specified location, and obtain differential correction data through the CORS network deployed on the ground; Aligning the differential correction data according to the national geodetic coordinate system to obtain unified spatiotemporal reference data; The unified spatiotemporal reference data is sent to the distributed reference nodes in the designated tunnel, and a consistent distributed spatiotemporal reference covering the entire range of the designated tunnel is generated with the reference point at the designated location as the starting point, so as to obtain a consistent distributed spatiotemporal reference in the tunnel; After the vehicle terminal receives the consistent distributed spatiotemporal reference in the tunnel, the sampling time and posture solution of the local sensor are synchronously corrected using the consistent distributed spatiotemporal reference in the tunnel to obtain the vehicle posture reference; if it is detected that the positioning accuracy of the vehicle posture reference deviates from the preset threshold in N consecutive sampling cycles, an error correction request is sent to the adjacent distributed reference node through a feedback mechanism; the local sensor includes at least one of an inertial measurement unit, a visual module, a laser radar and an odometer; Receiving the updated distributed spatiotemporal reference generated by the adjacent distributed reference nodes, and performing fusion calculation in combination with the real-time data collected by the local sensors to obtain accurate vehicle positioning and three-dimensional vehicle mapping; The updated distributed spatiotemporal reference is generated by combining historical cache data and a collaborative compensation mechanism after the adjacent distributed reference node receives the error correction request. When the communication link is interrupted or the priority target node fails, the node quality scores of the remaining nodes are re-evaluated, and the movable reference node with the second highest node quality score is selected as the update priority target node. The preliminary correction result is dynamically corrected in combination with the historical cache data and the collaborative compensation mechanism of the adjacent nodes to generate an updated correction result. The updated correction results are used to recalculate and generate the distributed spatiotemporal benchmark.

2. The method for maintaining a unified space-time benchmark according to claim 1, characterized in that: The designated location is the entrance of the designated tunnel.

3. The method for maintaining a unified space-time benchmark according to claim 1, characterized in that: The first satellite navigation system is specifically the Beidou satellite navigation system.

4. The method for maintaining a unified space-time benchmark according to claim 1, characterized in that: The second satellite navigation system is at least one of a GPS satellite navigation system, a GLONASS satellite navigation system and a Galileo satellite navigation system.

5. A reference maintenance system based on unified space-time, characterized in that: include: A signal acquisition module, used to acquire real-time positioning signals from the first satellite navigation system and the second satellite navigation system at a specified location, and obtain differential correction data through a CORS network deployed on the ground; A reference alignment module, used to align the differential correction data according to the national geodetic coordinate system to obtain unified spatiotemporal reference data; A reference distribution module, used to send the unified spatiotemporal reference data to the distributed reference nodes in the specified tunnel, and generate a consistent distributed spatiotemporal reference covering the entire range of the specified tunnel with the reference point at the specified position as the starting point, so as to obtain a consistent distributed spatiotemporal reference in the tunnel; A terminal correction module is used to synchronously correct the sampling time and posture solution of the local sensor using the time reference and the coordinate reference after the vehicle terminal receives the consistent distributed time and space reference in the tunnel, so as to obtain the vehicle posture reference; A feedback mechanism module, configured to send an error correction request to an adjacent distributed reference node through a feedback mechanism if it is detected that the positioning accuracy of the vehicle posture reference deviates from a preset threshold value for N consecutive sampling periods; An updated reference fusion module is used to receive the updated distributed spatiotemporal reference generated by the adjacent distributed reference nodes, and perform fusion calculations in combination with the real-time data collected by the local sensor to obtain accurate vehicle positioning and three-dimensional vehicle mapping; The updated distributed spatiotemporal reference is generated by combining historical cache data and a collaborative compensation mechanism after the adjacent distributed reference node receives the error correction request. When the communication link is interrupted or the priority target node fails, the node quality scores of the remaining nodes are re-evaluated, and the movable reference node with the second highest node quality score is selected as the update priority target node. The preliminary correction result is dynamically corrected in combination with the historical cache data and the collaborative compensation mechanism of the adjacent nodes to generate an updated correction result. The updated correction results are used to recalculate and generate the distributed spatiotemporal benchmark.

Images