A multi-target cooperative tracking design method and related device

Abstract

This application provides a multi-target cooperative tracking design method and related equipment, belonging to the field of UAV swarm perception and multi-target tracking technology. The method includes: collaboratively collecting visual observation data and TDOA observation data of the target through multiple UAV nodes; processing the two types of data using a visual PHD filter and a TDOA PHD filter respectively to obtain their respective target probability density functions; fusing the two density functions using a confidence-based dynamic weighting strategy to generate a unified target probability density distribution; dividing each node into Voronoi sub-regions and calculating its target density centroid based on the fused density distribution; and guiding each node to move towards its centroid using the Lloyd algorithm to achieve cooperative tracking and area coverage. This invention integrates the high precision of vision with the wide-area coverage of TDOA, adopts a distributed filtering architecture to reduce communication burden, and significantly improves the system's tracking accuracy, robustness, and efficiency through adaptive fusion and cooperative control.

Description

Technical Field

[0001] This application relates to the field of UAV swarm perception and multi-target tracking technology, and in particular to a multi-target cooperative tracking design method and related equipment. Background Technology

[0002] In recent years, the safety and management of low-altitude airspace have faced severe challenges. The widespread use of consumer drones has led to frequent "unauthorized flights," posing a serious threat to sensitive areas such as airports and critical facilities, making counter-unmanned aerial system (C-UAS) technology a research hotspot.

[0003] Current mainstream C-UAS technologies rely on sensing methods such as radar, vision, and radio, and mostly employ a centralized processing architecture. However, single sensors have inherent limitations: 1) Visual systems depend on lighting conditions and have a limited field of view, making them susceptible to obstruction; 2) Radar systems have weak detection capabilities against small UAV targets and are prone to false alarms; 3) While radio detection (such as TDOA) has a wide coverage area, its positioning accuracy is limited and severely constrained by synchronization errors and geometric layout. Furthermore, traditional centralized systems face enormous computational and communication pressures in scenarios with dynamically changing target numbers and unpredictable trajectories, making real-time response difficult.

[0004] Multi-UAV cooperative systems offer a potential solution to the aforementioned problems, improving system coverage and robustness through distributed perception and cooperative control. However, this also introduces new challenges: the perception capabilities of each UAV are heterogeneous, making the fusion of data from different modalities a key issue; communication bandwidth is limited in a distributed structure, making it impossible to frequently exchange raw data or complete trajectories; targets appear / disappear frequently, and the uncertainty between target states is high, requiring consideration of uncertain modeling for data correlation and state estimation. Summary of the Invention

[0005] The main objective of this application is to propose a distributed target tracking method and related equipment for multiple UAVs based on probability hypothesis density (PHD) filtering. By fusing two heterogeneous sensors, vision and time difference of arrival (TDOA), it achieves collaborative detection and tracking of dynamic or static UAV targets.

[0006] To achieve the above objectives, one aspect of this application proposes a multi-target cooperative tracking design method, the method comprising: Visual observation data and TDOA observation data of the target are collected collaboratively by multiple UAV nodes; Each UAV node constructs a visual PHD filter and a TDOA PHD filter, and processes the visual observation data and TDOA observation data to obtain the visual target probability density function. and TDOA target probability density function ; Regarding the and Perform weighted fusion to generate the fused target probability density function. ; Based on the above For each UAV node, a Voronoi sub-region is divided, and the target density centroid of each sub-region is calculated. ; Based on the target density centroid The Lloyd algorithm is used to guide the movement of each UAV node in order to achieve collaborative tracking of multiple targets.

[0007] In some embodiments, the calculation formula for the weighted fusion is:

[0008] in, For fusion weighting coefficients.

[0009] In some embodiments, the fusion weighting coefficient It is based on the confidence level of visual observation. and TDOA observation confidence The calculation formula is dynamically adjusted as follows:

[0010] in, For discrete time steps. In some embodiments, the visual observation confidence level It is calculated based on one or more of the residuals, covariance matrices, or target matching degrees from recent visual observations.

[0011] In some embodiments, the target density centroid is calculated. At this time, a gradient suppression term is introduced to achieve a smooth transition, and the calculation formula is as follows:

[0012] in, For smoothing weighting coefficients, For the first The Voronoi unit corresponding to each UAV within its working area. In some embodiments, a proportional-derivative controller is used to control the UAV nodes toward the target density centroid. The movement and control law are:

[0013] in, To control the quantity, and These are the current position and speed of the drone node, respectively. and These are control parameters.

[0014] In some embodiments, the observation noise covariance of the TDOA PHD filter It is a dynamic estimate, and the estimation formula is:

[0015] in, Historical weighting coefficients For the current observation value, These are the observed and predicted values.

[0016] In some embodiments, the method further includes a consensus coordination step: minimizing the difference function of the target density centroids among neighboring drone nodes. To optimize motion planning, the difference function is:

[0017] in, This is a set of indices for all drones. Indicates the first The set of neighboring nodes of a drone node. These are the target density centroids of the two drone nodes.

[0018] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0019] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.

[0020] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the method described above. The embodiments of this application include at least the following beneficial effects: This application provides a multi-target cooperative tracking design method, electronic device, storage medium, and program product. It proposes a dual-channel PHD filtering architecture (visual PHD filter and TDOA PHD filter), and combines Voronoi partitioning and the Lloyd algorithm to achieve dynamic region coverage and cooperative control, effectively improving the overall system detection capability and resource utilization efficiency. This invention integrates heterogeneous observations and achieves multi-target cooperative tracking without relying on a central server, providing a feasible solution to address the complex and ever-changing threats posed by low-altitude unmanned aerial vehicles (UAVs) in the future. Attached Figure Description

[0021] Figure 1 This is a flowchart of a multi-target cooperative tracking design method provided in an embodiment of this application.

[0022] Figure 2 This is a system framework diagram of an embodiment of this application; it illustrates the dual sensor input, PHD filtering, and motion control process.

[0023] Figure 3 This is an OSPA error curve of TDOA under different target densities in the embodiments of this application; Figure 4 This is an OSPA error curve diagram of three tracking methods (visual / TDOA / fusion) under different target densities in the embodiments of this application; Figure 5 These are OSPA error curves under different target densities and different mixing ratios in the embodiments of this application; Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0026] Before providing a detailed description of the embodiments of this application, some of the nouns and terms involved in the embodiments of this application will be explained first. The nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0027] 1) TDOA: This is a wireless positioning technology. TDOA positioning is a method of positioning using time difference. By measuring the time it takes for a signal to arrive at a monitoring station, the distance to the signal source can be determined. Using the distances from the signal source to each monitoring station (drawing a circle with the monitoring station as the center and the distance as the radius), the signal's location can be determined. However, absolute time is generally difficult to measure. By comparing the absolute time differences of the signal arriving at each monitoring station, a hyperbola can be plotted with the monitoring station as the focus and the distance difference as the major axis. The intersection of the hyperbola is the signal's location.

[0028] The Probability Hypothesis Density (PHD) filter is based on the Random Finite Set (RFS) theory. It can directly estimate the number and spatial distribution of targets, avoid complex data correlation, and has high computational efficiency. It is very suitable for multi-target tracking on distributed UAV platforms with limited computing resources.

[0029] To overcome the problems faced by existing multi-UAV systems, such as uncertain target numbers, strong heterogeneity of observations, low system coordination efficiency, and severe dependence on central nodes, this application proposes a multi-target cooperative tracking design method and related equipment that integrates observation information from visual and TDOA sensors. It utilizes dual-channel PHD filtering to achieve dynamic multi-target state estimation and density distribution modeling, and improves the system's task partitioning efficiency and target coverage capability through Voronoi partitioning and a Lloyd controller. This design exhibits high robustness, scalability, and real-time performance, making it suitable for target recognition, localization, and tracking tasks in dynamic and complex environments. It has broad application prospects, particularly in applications with high security and reliability requirements, such as key area defense, urban surveillance, and border patrol. In this scheme, each UAV is equipped with two heterogeneous sensors: a visual sensor and a TDOA sensor. The former is suitable for short-range, high-resolution detection, while the latter is suitable for long-range, global perception supplementation. The observation data from each sensor are locally processed to construct target density estimates, followed by heterogeneous multimodal fusion to overcome the limitations of single sensors in terms of detection range, resolution, and error robustness.

[0030] The multi-target cooperative tracking design method provided in this application relates to the field of UAV swarm perception and multi-target tracking technology. This multi-target cooperative tracking design method can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the multi-target cooperative tracking design method, but is not limited to the above forms.

[0031] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0032] Figure 1 This is an optional flowchart in the field of unmanned aerial vehicle (UAV) swarm perception and multi-target tracking technology provided in this application embodiment. Figure 1 The method may include, but is not limited to, steps S1 to S5.

[0033] Step S1: Collect visual observation data and TDOA observation data of the target through the collaborative collection of multiple UAV nodes.

[0034] In some embodiments, multiple drone nodes form a cooperative network. Each drone node is equipped with a visual sensor (such as an RGB camera) and a TDOA sensor (such as a microphone array or a UWB receiver). During system initialization, two independent sets of PHD filter parameters are configured for each drone node, including particle number, survival probability, detection probability, target birth model, and measurement noise model.

[0035] Step S2: Each UAV node constructs a visual PHD filter and a TDOA PHD filter, respectively, and processes the visual observation data and TDOA observation data to obtain the visual target probability density function. and TDOA target probability density function .

[0036] In some embodiments, a visual sensor acquires high-precision image observations within a preset range (e.g., 50m), extracts target location information using target detection algorithms such as YOLO, and models the observation model as a linear Gaussian model. A TDOA sensor collaborates with neighboring nodes to receive target signals and calculate the time difference of arrival; the observation model is then modeled as a nonlinear hyperbolic equation.

[0037] Subsequently, the visual PHD filter and TDOAPHD filter were updated using the visual observation set and the TDOA observation set, respectively, to obtain the visual target probability density function. and TDOA target probability density function To improve the robustness of the TDOA filter, a dynamic noise estimation mechanism is adopted to adaptively adjust the observation noise covariance.

[0038] Step S3, for the and Perform weighted fusion to generate the fused target probability density function. .

[0039] In some embodiments, the outputs of the two filters are weighted and fused to generate a fused target probability density function. :

[0040] Among them, the weighting coefficient It is not a fixed value, but is dynamically adjusted based on the real-time confidence levels of the vision and TDOA sensors:

[0041] The It can be obtained through recent observation residuals, eigenvalues ​​of the covariance matrix, or target matching similarity. This strategy ensures that when the performance of a certain sensor degrades, the system automatically tends to rely on more reliable sensor information.

[0042] Step S4, based on the For each UAV node, a Voronoi sub-region is divided, and the target density centroid of each sub-region is calculated. .

[0043] In some embodiments, each drone node calculates its Voronoi subregion based on its own location and the received location information of its neighboring nodes. In each Inside, calculate the centroid of the merged PHD distribution. This serves as the desired target point for that node. To prevent abrupt changes in the target density field from causing centroid instability, a gradient suppression term is introduced to achieve smooth calculation:

[0044] in This is the smoothing coefficient.

[0045] Step S5, based on the target density centroid The Lloyd algorithm is used to guide the movement of each UAV node in order to achieve collaborative tracking of multiple targets.

[0046] In some embodiments, the Lloyd algorithm is used to control each UAV node toward its center of gravity. Movement. Control commands are generated using a proportional-derivative (PD) controller:

[0047] To enhance the collaborative consistency among multiple drones, a consistency constraint term is introduced. And minimize it:

[0048] This constraint effectively reduces redundant coverage and improves group efficiency.

[0049] The following detailed description and explanation of the embodiments of the present invention are provided in conjunction with the accompanying drawings and specific application examples.

[0050] See Figure 2 This embodiment provides a multi-target collaborative tracking design method based on the fusion of vision and TDOA, including the following steps: Step 1: System Initialization - Initialize two sets of PHD filters for each UAV, one for visual observation and the other for TDOA observation. Initialize the target state density represented by particles, and set the survival probability, target generation model, and measurement noise model.

[0051] Visual observation is modeled as a linear Gaussian model:

[0052] in, Let the visual measurement vector be the time step 𝑘. The visual measurement matrix maps the state vector to the measurement space. Let k be the system state vector at time step k. For visual measurement noise at time step k, With a mean of zero and a covariance of It follows a multivariate normal distribution.

[0053] Step 2: Data Acquisition - The visual sensor acquires high-precision observations within a 50m range. The TDOA sensor collaborates with Voronoi neighbors to receive target synchronization signals and construct TDOA measurements. The TDOA observations are modeled as a nonlinear hyperbolic geometry:

[0054] in, The locations of the two receiving base stations. This is the target state. The TDOA measurement value at time step k. The TDOA measurement noise at time step k is zero-mean Gaussian noise, including arrival time estimation error and synchronization. With a mean of zero and a covariance of It follows a multivariate normal distribution.

[0055] To address the challenges of an unknown number of targets and their dynamic changes, the system employs a Propagation High-Density Filtering (PHD) framework. This method avoids the combinatorial complexity associated with explicit data association by propagating a density function rather than independent target states. This framework effectively models phenomena such as target birth, disappearance, and distribution aggregation, and exhibits good numerical stability. Each UAV runs an independent PHD filter on its visual observation set and TDOA observation set, and information fusion is achieved through adjustable weighting factors. The fusion result is used to construct a local target distribution for use by the task partitioning and path planning modules. Because traditional TDOA data suffers from excessive measurement errors and poor stability, a dynamic noise estimation mechanism is introduced to improve filtering robustness.

[0056] in Controlling the weight of historical memory, For current observations, The predicted value is used to dynamically adjust the covariance, improving the filter's adaptability to weak signals at long distances. This method can effectively adapt to noise characteristics in different environments.

[0057] Step 3: Local Filtering Update - The visual filter uses high-confidence observations for particle-weighted updates; the TDOA filter constructs a pseudo-likelihood function, which is then updated in conjunction with measurement geometry and synchronization errors. The core expression for PHD update is as follows:

[0058] Step 4: Multi-sensor fusion - Employing a weighted fusion strategy (e.g., weight ratio) (0.5), the particle weights output by the two filters are fused to retain their respective advantages.

[0059]

[0060] in The weighting is adaptively adjusted based on parameters such as real-time signal-to-noise ratio and target distribution. During multimodal fusion, to prevent misleading fusion due to visual or TDOA observation degradation at specific time points, a confidence-guided dynamic weighting strategy is employed.

[0061] in It can be obtained by calculating the observation residuals, covariance information, and matching degree of the most recent frames, thereby improving the adaptability of fusion decision-making.

[0062] Step 5: Cooperative Motion Planning - Each UAV calculates its Voronoi region based on its current location information; within its respective region, it calculates the fused PHD centroid and sets it as the target for the next motion step. In scenarios with sparse target distribution or blurred boundaries, Voronoi sub-regions are prone to jitter. Therefore, a gradient suppression term is introduced to make the region centroid smoother with changes in target density.

[0063] in These are the smoothing weighting coefficients.

[0064] Step 6: Use the Lloyd algorithm to guide the UAV towards the target area. To handle high-speed targets or rapid response requirements, a first-order predictive control mechanism is introduced:

[0065] In the formula As the current sub-region centroid, This is the current location of the drone; yes The derivative of yes The derivative of .

[0066] Step 7: Dynamic Target Processing - The system supports target appearance and disappearance, employing target birth-death modeling; it combines motion models to predict target states, maintaining tracking continuity. Furthermore, to enhance coordination and consistency among multiple UAVs, the system introduces consistency constraints. :

[0067] By minimizing This ensures consistency in target guidance within the neighborhood, thereby reducing redundant coverage among drones and improving swarm efficiency.

[0068] As a further preferred embodiment, step 5 further includes the following steps: The UAV delineates spatial sub-regions based on its own position and broadcast information from surrounding UAVs, and calculates the target aggregation center within each sub-region based on the density field distribution. By iteratively updating the UAV's position and sub-region boundaries using the Lloyd algorithm, it achieves rapid response and continuous focusing on high-density target areas. The UAV's final navigation strategy is based on a density field centroid-driven mechanism.

[0069] Subsequently To target the location, a gradient descent control strategy is used to drive the platform to move, enabling focused detection of high-density areas.

[0070] (1) Experimental verification This embodiment performs simulation verification in a two-dimensional space, such as a 1000×1000 meter two-dimensional space, in an unobstructed area. The target is a stationary or moving drone simulated target, which is placed evenly distributed on the map at the beginning of the experiment.

[0071] The implementation parameters are as follows: Number of UAVs: Nr = {10, 20, …, 100}; Number of targets: Nt = {10, 30, 50}; Maximum speed of UAVs: 8 m / s; Sampling frequency: 2 Hz; Visual sensor observation range: 50 m; TDOA maximum sensing radius: 1000 m; Initial number of particles: 1000 per UAV; Initial weight of each particle: 1e-4.

[0072] During the system initialization phase: In the initial deployment of the multi-UAV system, each UAV is pre-loaded with a parameter configuration file, including the visual recognition model, TDOA receiver module initialization parameters, particle filter initialization status, communication address, and clock synchronization module. ROS or DDS is used for time synchronization and communication initialization to ensure system timing consistency. Next, multimodal observation and acquisition: Each UAV acquires image frames at a fixed frequency and runs visual front-end detection algorithms such as YOLOv5 / YOLOX to extract candidate target regions; simultaneously, a microphone array / ultra-wideband (UWB) receiver is used to acquire TDOA signals, and a multipath suppression algorithm is used to eliminate abnormal propagation paths.

[0073] For each type of observation information, two independent PHD (Particle Hierarchy Processing) filters are constructed: the visual filter generates particle sets using appearance and target motion models, and performs importance sampling and resampling; the TDOA (Total Distance Occurrence) filter generates particle sets based on sound source geometry modeling combined with multi-base station range differences. The local density functions are calculated for both.

[0074]

[0075] The two sets of filtering results are then fused using confidence weighting to obtain a unified target state density estimate.

[0076] in Dynamic estimation based on current observation residuals, covariance matrix, etc.

[0077] The fused target density distribution is input into the Voronoi diagram partitioning module to divide the entire task area into sub-regions, with each UAV corresponding to a sub-region. Within each sub-region, the target density-weighted centroid is calculated.

[0078] The UAV uses a proportional-derivative (PD) controller for navigation control based on its center of mass position.

[0079] Collision-free paths are generated by combining a local trajectory optimization module with obstacle avoidance strategies. All UAVs broadcast their estimated density distribution boundaries, centroid positions, and local TDOA confidence indices at a fixed frequency. Other UAVs achieve asynchronous collaboration through message fusion and graph optimization algorithms to avoid global conflicts and redistribution delays.

[0080] In case of abnormal situations such as TDOA reception failure or visual target loss, the system will automatically call the short-term memory particle set of the most recent frame and combine it with the Bayesian prediction model to generate pseudo-observations for compensation, maintaining tracking continuity. When the target density distribution converges to a stable state, or the number of UAVs falls below a set threshold, the system enters the energy consumption optimization phase, some UAVs go into hibernation, and the master UAV uploads mission logs and updates model parameters.

[0081] Each UAV uses the centroid of the target density within the area as its navigation target and performs iterative movement using the Lloyd algorithm. The OSPA error is recorded in each iteration as a performance metric.

[0082] (1.1) Validation Example 1: Laboratory Environment Validation In an indoor UAV test field measuring 15 meters on each side, three mobile UAV platforms were deployed, each integrating an RGB camera and a UWB receiver array to form a heterogeneous observation system. The experimental targets were 1 to 3 slowly moving quadcopter UAVs. The system periodically acquired image streams and executed the YOLOv5-tiny model to identify the UAVs, and also acquired TDOA signals based on a sound source simulator.

[0083] On each UAV, visual observations are corrected by an appearance model and then enter a particle PHD filter, with the number of particles set to N=150 and an update rate of 10Hz; the TDOA filter estimates the target direction and generates particles through multi-base station differential time delay, and the confidence level is dynamically adjusted based on covariance.

[0084] The fusion module uses a weighted average method to fuse D_vis and D_tdoa, and periodically updates the Voronoi partition. After each iteration, the centroid is repositioned using the Lloyd algorithm, and the navigation module executes the PD controller to plan the trajectory. Tests show that the system can improve the tracking success rate by 27.4% under occlusion interference, and the average response time is controlled within 0.9s.

[0085] (1.2) Verification Example 2: Outdoor Simulation Test A 30m×30m open terrain simulation scene was built in the ROS-Gazebo environment, and 5 drones were deployed, each equipped with a simulated RGB-D sensor and a TDOA microphone array. The target was 5 high-speed simulated drones moving simultaneously (maximum speed 3m / s). All drones ran an autonomous navigation stack, and communication used the DDS protocol to broadcast local density and target estimates.

[0086] The filter particle number is set to 200, and the α value is dynamically adjusted using a system residual-driven confidence fusion mechanism. The fused density field is used for Voronoi partitioning and controlled by a control law. To achieve collaborative navigation.

[0087] Simulation results show that the system maintains a delay of less than 1.3 seconds under dense dynamic interference from multiple targets, the mean error of OSPA is reduced by 35% compared with the single sensor scheme, and the task completion time is reduced by about 18%.

[0088] (1.3) Verification Example 3: Fault Robustness Verification In the experimental setup, the camera was simulated to be blocked (20s) and the TDOA module was disconnected (intermittently). The PHD filter module automatically activated the historical particle replay mechanism in the event of reception failure, and filled in the missing observation area by predicting the state distribution through Markov; at the same time, the controller introduced an inertial navigation trajectory for transition.

[0089] Real-world testing shows that even in scenarios where the TDOA signal is lost for 5 consecutive seconds or severely obstructed, the system can still maintain a target error of <2m and a recovery time of <3s, demonstrating high robustness and short-term fault tolerance.

[0090] (1.4) Summary When the number of targets is small (Nt = 10), the accuracy of the fusion strategy is significantly better than that of a single sensor; when the targets are dense (Nt = 50) and there are few drones (Nr < 30), TDOA maintains coverage capability, while the fusion mechanism significantly improves estimation stability; in moving target tracking scenarios, the fusion method can reduce the average OSPA error by 30-50%, and has stronger real-time performance and robustness.

[0091] See Figure 3 , Figure 4 and Figure 5 Experimental curves show that as the number of drones increases, the system estimation error rapidly converges to <1 m; when the TDOA coverage increases to 1000 m, the system stability is significantly improved and the target reconstruction capability is enhanced.

[0092] (2) Beneficial effects Simulation results show that the method in this embodiment still has good tracking stability and accuracy under conditions of dense distribution of multiple targets and partial occlusion, and the average OSPA error can be reduced by more than 30%.

[0093] In summary, the method of this embodiment has at least the following advantages and beneficial effects compared to the prior art: 1) Heterogeneous integration and complementary advantages: By integrating visual (high precision) and TDOA (wide area) observation, the system's target observability and anti-interference ability in complex environments are significantly improved.

[0094] 2) Distributed architecture, high efficiency and reliability: The distributed PHD filtering is adopted, and each node processes data locally and only exchanges high-level information (such as density distribution), which greatly reduces the communication bandwidth requirements and enhances the scalability and reliability of the system.

[0095] 3) Adaptive collaboration and coverage optimization: By combining Voronoi diagram partitioning and Lloyd algorithm, the UAV swarm achieves adaptive coverage and focusing on dynamic target distribution, resulting in high resource utilization.

[0096] 4) Strong robustness: The dynamic weighted fusion and fault compensation mechanism ensure that the system can still maintain stable tracking performance in abnormal situations such as partial sensor failure, occlusion or target loss.

[0097] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0098] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0099] Please see Figure 6 , Figure 6 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 601 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 602 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 602 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 602 and is called and executed by the processor 601 using the methods described in the embodiments of this application. The input / output interface 603 is used to implement information input and output; The communication interface 604 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 605 transmits information between various components of the device (e.g., processor 601, memory 602, input / output interface 603, and communication interface 604); The processor 601, memory 602, input / output interface 603, and communication interface 604 are connected to each other within the device via bus 605.

[0100] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0101] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0102] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0103] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0104] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented in the embodiments of this program product are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments. The executable computer program code or "code" used to perform the various embodiments can be written in high-level programming languages ​​such as C, C++, Python, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0105] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0106] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0107] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0108] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0109] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0110] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0111] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0112] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0113] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0114] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0115] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A multi-target cooperative tracking design method, characterized in that, The method includes the following steps: Visual observation data and TDOA observation data of the target are collected collaboratively by multiple UAV nodes; Each UAV node constructs a visual PHD filter and a TDOA PHD filter, and processes the visual observation data and TDOA observation data to obtain the visual target probability density function. and TDOA target probability density function ; Regarding the and Perform weighted fusion to generate the fused target probability density function. ; Based on the above For each UAV node, a Voronoi sub-region is divided, and the target density centroid of each sub-region is calculated. ; Based on the target density centroid The Lloyd algorithm is used to guide the movement of each UAV node in order to achieve collaborative tracking of multiple targets. The formula for calculating the weighted fusion is as follows: in, For fusion weighting coefficients; The fusion weight coefficient It is based on the confidence level of visual observation. and TDOA observation confidence The calculation formula is dynamically adjusted as follows: in, For discrete time steps; The confidence level of visual observation It is calculated based on one or more of the residuals, covariance matrices, or target matching degrees from recent visual observations; Calculate the target density centroid At this time, a gradient suppression term is introduced to achieve a smooth transition, and the calculation formula is as follows: in, For smoothing weighting coefficients, For the first The Voronoi unit corresponding to each UAV within its working area.

2. The method according to claim 1, characterized in that, A proportional-derivative controller is used to control the UAV nodes toward the target density centroid. The movement and control law are: in, To control the quantity, and These are the current position and speed of the drone node, respectively. and These are control parameters.

3. The method according to claim 1, characterized in that, The observation noise covariance of the TDOA PHD filter It is a dynamic estimate, and the estimation formula is: in, Historical weighting coefficients For the current observation value, These are the observed and predicted values.

4. The method according to claim 1, characterized in that, The method also includes a consensus coordination step: minimizing the difference function of the target density centroids among neighboring drone nodes. To optimize motion planning, the difference function is: in, This is a set of indices for all drones. Indicates the first The set of neighboring nodes of a drone node. These are the target density centroids of the two drone nodes.

5. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 4.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 4.

Images