A distributed NLOS detection and robust positioning method and system based on dynamic security root

Abstract

This invention discloses a distributed NLOS detection and robust localization method and system based on dynamic security foundation. Multiple intelligent agent nodes share a group DSF, independently generate logical timelines, and obtain TOA distance measurements by exchanging security pilot signals. Each node independently executes MDS to reconstruct relative coordinates and calculates residuals. Utilizing the global synchronization characteristic of the logical timeline, the residuals and NLOS judgment results of each node for the same link are naturally consistent, automatically achieving distributed consensus without additional signaling. Relative coordinates are re-obtained after removing or downweighting links judged as NLOS. Furthermore, IMU data is aligned with TOA observations based on the logical timeline, and the usage of observations is controlled according to the NLOS detection results (discarding, downweighting, or replacing with IMU predictions). Simulations show that in a 40% NLOS link environment, this invention reduces the standard deviation of ranging residuals from 5.006 meters to 0.122 meters, a reduction of 97.6%, while maintaining continuous and stable trajectories in NLOS regions. This invention features zero signaling, decentralization, and significantly improves the localization robustness of the cluster in NLOS environments.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication and cooperative positioning technology, specifically relating to a method, system, and apparatus for distributed detection and consensus-based elimination of ranging anomalies caused by non-line-of-sight (NLOS) propagation in a distributed intelligent agent swarm based on Dynamic Security Foundation (DSF) and logical time axis in environments without infrastructure and where GNSS is denied. This method enhances the robustness of inertial measurement unit (IMU) and time of arrival (TOA) fusion positioning. This invention is particularly suitable for applications requiring high-precision and robust cooperative positioning in NLOS scenarios such as UAV swarms, robot formations, and missile clusters, including urban canyons and mountainous terrain. Background Technology

[0002] Distributed intelligent agent swarms need to obtain their relative positions in real time when performing tasks such as cooperative reconnaissance, formation flying, and target tracking. Time-of-arrival (TOA) ranging is a common method: each agent sends a known signal to the others, measures the propagation delay, and thus obtains the distance. However, in non-line-of-sight (NLOS) environments, signals are blocked by obstacles such as buildings and mountains, and can only reach the other party through reflection and diffraction paths. This causes a positive deviation in the TOA measurement (the measured distance is greater than the actual straight-line distance), with deviations reaching tens or even hundreds of meters. This deviation can severely damage the accuracy of algorithms such as multidimensional scaling (MDS) and multilateral localization, leading to the failure of relative coordinate reconstruction.

[0003] Existing NLOS detection and mitigation methods mainly include: (1) single-node classifiers based on signal amplitude, delay spread, kurtosis and other features, which require a large number of prior statistical models and have poor adaptability; (2) residual analysis based on multi-round ranging, which usually requires the central node to collect global measurement values ​​for optimization, resulting in large signaling overhead and the risk of single-point failure; (3) using inertial navigation (IMU) or vision-assisted positioning, but traditional schemes rely on GPS second pulses or Network Time Protocol (NTP) to align multi-node data, which cannot work in GNSS denial environments.

[0004] The Dynamic Security Foundation (DSF) framework previously proposed by the applicant (see patents 2026100015014, 2026100463580, 2026101229059, etc.) and the "A Synesthesia-Computing Integrated System and Method Based on a Logical Time Axis" (hereinafter referred to as the "Synesthesia-Computing Patent") have achieved a logical time axis that does not require an external time reference, enabling communication, sensing (including IMU), and computing tasks to be aligned on a unified logical counter n. Specifically, the "real-time mode" of the Synesthesia-Computing Patent stipulates that at each logical decision moment n, the node triggers IMU sampling and binds it to n; the Kalman filter directly uses n as the state index for recursion, thereby completely eliminating the time synchronization error between sensors.

[0005] However, the aforementioned sensing-based ranging scheme assumes that TOA ranging is obtained under line-of-sight (LOS) conditions, failing to consider the positive ranging bias caused by NLOS propagation. When agent clusters operate in complex environments such as urban areas and mountainous terrain, signal obstruction can cause additional delays of tens or even hundreds of meters in some links. Directly substituting this bias into a Kalman filter indexed by n will result in the estimation of severely contaminated states, leading to positioning divergence. Furthermore, while the existing DSF framework provides secure pilot-to-pilot ranging capabilities, it has not yet solved the problem of automatic detection and elimination of NLOS links. Centralized optimization methods based on residual analysis require a central node, which contradicts the decentralized philosophy of the DSF framework.

[0006] Therefore, there is an urgent need for a distributed, zero-additional-signaling NLOS detection and consensus elimination method that utilizes the logical synchronization characteristics within the DSF logical time axis framework. Furthermore, it is necessary to enhance the robustness of IMU / TOA fusion on this basis, enabling the integrated sensor-computing system to operate robustly in an NLOS environment. Summary of the Invention

[0007] Core concept definition This invention, while fully inheriting the definitions of prior patents (especially synesthetic calculation patents), expands the following concepts: - Dynamic Security Foundation (DSF): As defined in Patent 2026100015014, it is a shared cryptographic triple (K_sec, Init_Anchor, Rule_ID).

[0008] - Protocol Secure State Machine (PSSM): A deterministic state machine based on DSF, outputting a logical state S(t), with all agents' logical counters n strictly synchronized.

[0009] - Logical Timeline: A discrete sequence consisting of logical decision moments n output by the PSSM, decoupled from the physical clock. As described in the Tongsensing patent, all communication, sensing (IMU), and computing tasks use n as a unified time index.

[0010] - Safety Pilot: A pseudo-random sequence generated based on the individual DSF and the current logical state, used for mutual ranging and communication. The pilots of different agents are approximately orthogonal.

[0011] - TOA measurement model: at line-of-sight, tau_ij = d_ij / c + delta_i - delta_j + epsilon_ij; at NLOS, add a positive deviation b_ij > 0.

[0012] - IMU Data Logical Binding: As described in the Tonggan Computing patent, IMU sampling is triggered at each logical decision moment n and bound to n. IMU data from all agents at the same n are logically aligned.

[0013] Purpose of the invention This invention aims to address the problems of drastic decrease in positioning accuracy of distributed intelligent agent clusters in NLOS environments, NLOS detection relying on centralized signaling, and the failure of existing integrated sensing and computing solutions to handle NLOS ranging bias. It provides a distributed NLOS detection and robust positioning method based on DSF and a logical time axis. This method utilizes the logical synchronization characteristics of DSF to achieve NLOS consensus marking with zero additional signaling, and performs robust IMU / TOA fusion based on logical time axis alignment, significantly improving positioning robustness in NLOS environments.

[0014] Technical solution I. Distributed NLOS Detection and Consensus Removal Methods Step N1: Each agent generates a safety pilot based on DSF and transmits TOA measurements to each other. Each agent i sends a safety pilot P_i(n) at logical decision time n and receives pilots from other agents. The TOA measurement value tau_ij^(n) is obtained through matched filtering (correlation peak detection), which is converted into the distance d_ij^(n) = c * tau_ij^(n). Due to DSF synchronization, all agents obtain their distance measurements from each other at the same logical time n, forming a symmetric distance matrix D(n).

[0015] Step N2: Each agent independently performs MDS reconstruction of relative coordinates. Each agent collects the available {d_ij^(n)} at the current time (the maximum tolerance for missing values ​​can be set), and performs classic MDS: 1. Calculate the squared distance matrix D2; 2. Bicentric B = -0.5 * J * D² * J, where J = I - (1 / N)*1*1^T; 3. Perform eigenvalue decomposition on B, and take the eigenvectors corresponding to the first three largest eigenvalues ​​to obtain the relative coordinate matrix X(n).

[0016] Since all agents use the same distance matrix (with consistent rules for handling missing values) and the MDS algorithm is deterministic, the X(n) calculated independently by each agent is consistent in the sense of rotation / reflection / translation, requiring no additional alignment signaling.

[0017] Step N3: Calculate the theoretical distance and residual Based on the reconstructed coordinates, each agent calculates the theoretical distance d_hat_ij^(n) = || p_hat_i^(n) -p_hat_j^(n) || and obtains the residual r_ij^(n) = | d_ij^(n) - d_hat_ij^(n) |.

[0018] Step N4: Distributed consensus voting based on logical synchronization (zero additional signaling) Utilizing the logical synchronization characteristic of DSF, all agents achieve identical residual calculation results for the same link (numerical errors are negligible). Specifically: - Each agent has a preset NLOS decision threshold gamma (e.g., 2 to 3 times the standard deviation of TOA noise, or configured according to the false positive rate required by the system).

[0019] - If r_ij^(n) > gamma, then the agent determines that link (i,j) is "suspected NLOS".

[0020] Since all agents have the same calculation results for the residuals, their judgment results are also naturally consistent, and a global consensus can be automatically formed without any communication negotiation.

[0021] Step N5: Remove NLOS links and re-MDS Each agent sets the d_ij^(n) corresponding to the link determined to be NLOS as missing (or assigns it a very low weight), and then performs MDS again to obtain a more accurate relative coordinate after removing outliers.

[0022] Step N6: Iterative optimization (optional) Steps N3-N5 can be repeated, removing several links with the largest residuals each time, until the residual statistics converge (e.g., the maximum residual is less than a threshold). Stability is usually achieved in 1-2 iterations.

[0023] II. Robust Fusion of IMU / TOA Based on Logical Time Axis (Enhancement Based on Synesthesia Calculation Patent) This section enhances NLOS robustness on top of the IMU and TOA fusion framework already implemented in the Tongyin Computing patent. The Tongyin Computing patent already provides a Kalman filter architecture with a logical step size T_step as the discrete time interval and n as the state index. This invention adds the following enhancements to this architecture: Step F1: IMU data logic binding (Tongtonggan calculation patent) At each logical decision moment n, agent i triggers IMU sampling to obtain acceleration a_i(n) and angular velocity omega_i(n), and stores them in conjunction with the logical counter n.

[0024] Step F2: IMU state recursion based on logical step size (Tongtongsen calculation patent) Using the logical step size T_step as the discrete time interval, the state is recursively predicted using IMU data. The state vector X_i(n) = [p_i^T, v_i^T, q_i^T]^T, and the prediction equation is: p_i(n+1) = p_i(n) + v_i(n) T_step + 0.5 a_i(n) T_step^2 v_i(n+1) = v_i(n) + a_i(n) T_step q_i(n+1) = q_i(n) × exp(0.5 omega_i(n) T_step) Where × represents quaternion multiplication. This recursion is based entirely on the logic counter n and does not depend on absolute physical time.

[0025] Step F3: Observation quality control based on NLOS detection results When obtaining TOA ranging with other agents j, the method in Part 1 is first used to determine whether the link is NLOS: - If the link is determined to be LOS, then the TOA distance is treated as a normal observation and used in the Kalman filter update.

[0026] - If the link is determined to be NLOS, then execute one of the following strategies: Strategy 1: Discard the observation completely and rely solely on IMU predictions; Strategy 2: Reduce the weight of the observation (e.g., increase the measurement noise covariance by a factor of 10); Strategy 3: Use the relative distance ||p_i_pred(n) - p_j_pred(n)|| predicted by the IMU to replace the measured distance as a virtual observation.

[0027] Step F4: Enhanced Consistency Check To further verify the NLOS determination, the consistency between the relative distance predicted by the IMU and the measured TOA distance can be checked. If |d_ij^(n) - ||p_i_pred(n) - p_j_pred(n)|| | > gamma2 (the second threshold, usually greater than gamma), then the NLOS confidence is strengthened, and strategy three (predicted value substitution) can be triggered.

[0028] Step F5: Distributed Kalman Filtering (Patent for Tongtong Inductance Calculation) A distributed Kalman consensus filter is employed, where each agent maintains its own state estimate and exchanges predictions with its neighbors. Because the logical time axis is synchronized, the states of different agents at the same time step n can be directly used for collaborative updates without the need for time interpolation.

[0029] III. System Architecture and Devices The system implementing the method of the present invention includes multiple intelligent agent nodes, each node comprising: - DSF storage and PSSM processing module: maintains group DSF, generates logical timelines, and manages logical counters; - Safety pilot transceiver module: transmits and receives safety pilots generated based on individual DSFs at the time of logical decision-making; - TOA measurement module: Performs matched filtering on the received signal, estimates the time of arrival, and calculates the distance; - MDS calculation module: Reconstructs relative coordinates based on the distance matrix; - NLOS detection and consensus module: calculates residuals, determines NLOS links, and automatically reaches consensus using logical synchronization characteristics; - Elimination and Recalculation Module: Set the distance corresponding to the NLOS link as missing and re-execute MDS; - IMU sensor module and logic binding module (Tongtonggansuan patent): Collect IMU data at each logic decision moment and bind it to n; - Predictive update module (based on synesthesia patent enhancement): It recursively calculates the state based on the logical step size and controls the usage of observations (discard, downweight, or replace) according to the NLOS detection results.

[0030] The group DSF root key is pre-distributed to all agents. The system has no central server; all processing is distributed.

[0031] Beneficial effects 1. NLOS consensus with zero additional signaling: By leveraging the consistency of DSF logical synchronization and MDS reconstruction results, all agents do not need any communication negotiation to determine the NLOS link, completely avoiding the signaling overhead caused by exchanging NLOS tags in traditional schemes, which is especially suitable for highly dynamic, low-bandwidth clusters.

[0032] 2. Highly Robust Positioning: By eliminating NLOS (Not From Service) anomalies in ranging, the positioning accuracy is significantly improved compared to directly using MDS (Mean Distance Detection). Monte Carlo simulations show that, under conditions where 40% of the links are NLOS (with an additional positive deviation of 10m) and the TOA (Total Anomaly) noise standard deviation is 0.3m, the standard deviation of the ranging residuals using the traditional method is 5.006m, while the standard deviation of the residuals using the method of this invention (after eliminating NLOS links) is reduced to 0.122m, a reduction of 97.6%, verifying the effectiveness of this invention in NLOS detection and elimination.

[0033] 3. Seamless compatibility with the Tonggan Computing patent: This invention is entirely built on the logic time axis and IMU / TOA fusion framework already implemented in the Tonggan Computing patent, adding only NLOS detection and robust processing modules. Existing DSF devices can obtain this capability through firmware upgrades without damaging the original zero signaling characteristics.

[0034] 4. Robust fusion without GPS time synchronization: Inheriting the advantage of no external time synchronization from the patented Tonggan algorithm, it also increases the robustness to NLOS deviation, enabling the system to maintain high-precision positioning even in complex environments where GNSS is denied and NLOS exists.

[0035] 5. High scalability: It can be used for NLOS rejection in pure TOA as well as for IMU+TOA tight coupling; NLOS detection results can be used for MDS and Kalman filtering at the same time, forming dual protection. Attached Figure Description

[0036] Figure 1 This is a flowchart of the distributed NLOS detection and consensus elimination process of this invention.

[0037] Figure 2 This is a schematic diagram of a robust IMU / TOA fusion framework based on a logical time axis (enhanced based on the synesthesia calculation patent).

[0038] Figure 3 This chart compares the standard deviations of ranging residuals between the traditional method and the proposed DSF method in an NLOS environment. The horizontal axis represents the standard deviation of residuals (unit: meters), the left vertical axis represents the cumulative distribution function (CDF), and the right vertical axis represents the box plot. The traditional method directly calculates the standard deviation of residuals using all links; the DSF method calculates the standard deviation of residuals after removing detected NLOS links. Simulation results show that in an environment with 40% NLOS links, the mean standard deviation of residuals for the traditional method is as high as 5.006 meters, while the DSF method reduces it to 0.122 meters, a relative reduction of 97.6%, verifying the effectiveness of the present invention in NLOS detection and removal.

[0039] Figure 4This figure compares the localization trajectories of the traditional method and the proposed DSF method in an NLOS environment. The experimental scenario includes typical movement trajectories of straight lines and circular turns, with NLOS bias introduced in the circular turn area (grey marked area). The traditional method (blue dashed line) exhibits severe trajectory drift in the NLOS-affected area, failing to follow the true path. The proposed DSF method (red solid line), within the NLOS area, detects ranging anomalies, actively discards contaminated TOA observations, and relies solely on IMU prediction (uniform linear model) to recursively deduce the position; therefore, the trajectory extends in a straight line without spontaneous turning. After leaving the NLOS area, observations return to normal, and the trajectory quickly converges to the true path. This behavior demonstrates the robust design principle of this invention—"not relying on unreliable data and maintaining motion continuity"—under abnormal observations. This figure verifies the robustness and effectiveness of this invention in dynamic NLOS scenarios. Detailed Implementation

[0040] Example 1: Distributed NLOS Detection and Consensus Removal Scenario: Ten drones are flying in an urban environment. Due to building obstruction, approximately 20% of the links are NLOS (Not at Rest). All drones share a group DSF (Digital Subsequent Facing), with a logical step size T_step = 0.125 ms. At logical time n=1000, each drone broadcasts a safety pilot, which, upon reception, yields a TOA (Total Occurrence Allocation) distance matrix (including NLOS bias). Each drone independently performs MDS (Mean Distance Detection and Reconstruction) to reconstruct coordinates, calculates the residual, and sets a threshold γ=2σ (σ≈0.5m). All drones detect a residual of 18m at link (3,7), exceeding the threshold, and unanimously classify it as NLOS. After removing this link and re-performing MDS, the RMSE (Real-Time Sequence) decreases from 42m to 4.1m. The entire process involves no additional signaling.

[0041] Example 2: Robust Fusion of IMU / TOA Based on Logical Time Axis Scenario: In the same cluster, each drone is equipped with a low-cost IMU (MPU6050), and IMU recursion and Kalman filtering with logical time axis alignment have been implemented according to the patented Tongsensuan algorithm. At logical time n=1000, link (3,7) is identified as NLOS by the NLOS detection module. Based on this determination, the Kalman filter module increases the noise covariance of the TOA measurement from R=1 to R=100 (reducing weight), thereby almost ignoring this abnormal observation. At the same time, the consistency between the relative distance predicted by the IMU (approximately 5.2m) and the measured TOA distance (approximately 23m) is checked, and the deviation reaches 17.8m, further verifying the NLOS determination. The final fused positioning trajectory is smooth, unaffected by NLOS, and the deviation from the true trajectory is less than 2m. In contrast, if NLOS detection is not used and the original Tongsensuan patented fusion method is used directly, the positioning error will jump to more than 15m due to this NLOS observation.

[0042] Example 3: Combined robustness and predictive substitution In more extreme NLOS scenarios (such as when link (3,7) is completely blocked by a large building, and the measured TOA distance deviation exceeds 50m), the system adopts strategy three: directly replacing the measured distance with the relative distance predicted by the IMU (5.2m). Because the IMU has high short-time recursion accuracy (drift <0.5m / s), the observation error after replacement is less than 1m, and the Kalman filter can still converge stably. Simultaneously, the system continues to monitor the link at several subsequent logical moments, and once the residual returns to normal (e.g., after the drone flies over the building), it automatically resumes using the measured TOA.

[0043] This invention can be embedded in intelligent terminal devices such as drones, robots, and unmanned vehicles. As a functional extension of the DSF framework and the integrated sensor-computer interface patent, it can be widely used in cluster collaborative positioning in complex environments such as military reconnaissance, disaster relief, and urban logistics, significantly improving the system's survivability and positioning accuracy under NLOS conditions.

Claims

1. A distributed non-line-of-sight detection and robust localization method based on dynamic security foundation, characterized in that, include: Multiple intelligent agent nodes share the Group Dynamic Security Foundation (DSF), and each independently generates a logical timeline defined by the logical decision moments output by the Protocol Security State Machine (PSSM). Each logical decision moment is associated with a monotonically increasing logical counter n. Each node obtains time-of-arrival (TOA) distance measurements between nodes by exchanging safety pilot signals; Each node reconstructs its relative coordinates based on the distance measurement and calculates the residual between the theoretical distance and the measured distance; By leveraging the global synchronization characteristic of the logical time axis, the residuals and corresponding non-line-of-sight (NLOS) judgment results of each node for the same link are naturally consistent, thereby automatically achieving distributed consensus without the need for additional signaling. Each node removes or downgrades the distance measurement corresponding to the link determined to be NLOS based on the consensus, and re-obtains the relative coordinates based on the corrected distance measurement.

2. The method according to claim 1, characterized in that, The reconstructed relative coordinates are achieved using multidimensional scaling (MDS) or other relative positioning algorithms that do not require anchor nodes.

3. The method according to claim 1, characterized in that, The threshold for NLOS determination is configured based on the standard deviation of TOA measurement noise or the false positive rate required by the system.

4. The method according to claim 1, characterized in that, The achievement of the distributed consensus depends on the fact that all nodes have the same distance measurement matrix and the algorithm for reconstructing relative coordinates is consistent. Therefore, the residuals and judgment results for the same link are naturally the same, without the need for any inter-node negotiation signaling.

5. The method according to claim 1, characterized in that, Also includes: The NLOS detection and elimination process is iteratively executed at multiple logical decision points until the residuals converge.

6. The method according to claim 1, characterized in that, Also includes: For links identified as NLOS, their measurements are temporarily masked during multiple subsequent logical decision-making moments, and the links are periodically retried to detect whether the line of sight has been restored.

7. The method according to claim 1, characterized in that, Also includes: At each logical decision moment n, each node acquires inertial measurement unit (IMU) data and binds it to n; Using logical step size T_step as the discrete time interval, the IMU data is used to predict and recursively predict its own state; The method of using TOA distance measurement is controlled based on the NLOS determination result.

8. The method according to claim 7, characterized in that, The methods of using the controlled TOA distance measurement include: If the distance to line of sight (LOS) is determined, the TOA distance is used as a normal observation value in the Kalman filter update; If the observation is determined to be NLOS, at least one of the following operations will be performed: discard the observation, amplify the noise covariance of the observation by a preset factor to reduce its weight, or replace the observation with the relative distance predicted by the IMU.

9. The method according to claim 8, characterized in that, Also includes: The relative distance predicted by the IMU is used to check the consistency with the measured TOA distance. If the deviation exceeds the second preset threshold, the confidence level of NLOS judgment is strengthened, and the measured distance is automatically replaced by the IMU predicted distance.

10. The method according to claim 1, characterized in that, The group DSF includes the group root key K_sec^root, the initial anchor point Init_Anchor, and the state transition rule identifier Rule_ID; The state transition rule identified by the Rule_ID is a preset deterministic evolution rule, including at least one of the following: broadcast clock driven rule, hash chain driven rule, or logical epoch driven rule.

11. The method according to claim 1, characterized in that, The synchronization of the logical timeline includes silent calibration: the master node periodically broadcasts a verification token generated based on the current logical state, and the slave node matches the token with the candidate value of its local counter within the time deviation window. After a successful match, the local counter is silently calibrated.

12. The method according to claim 1, characterized in that, Each node derives its own individual DSF based on the group DSF root key and its own unique identifier, which is used to generate security pilots and encryption keys to achieve cryptographic isolation between nodes.

13. A distributed non-line-of-sight detection and robust positioning system based on dynamic security foundation, characterized in that, It includes at least two agent nodes, each agent node being configured to perform the method of any one of claims 1 to 12.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method of any one of claims 1 to 12.

15. An intelligent agent node, comprising a processor, a memory, a radio frequency transceiver, and an inertial measurement unit, characterized in that, The memory stores a group DSF and executable instructions, and the processor executes the instructions to implement the node-side steps of the method according to any one of claims 1 to 12.

16. A distributed intelligent agent cluster, characterized in that, It includes multiple intelligent agent nodes as described in claim 15.

Images