Multi-machine collaborative operation anti-collision monitoring system and method

Abstract

The invention belongs to the technical field of intelligent control, and discloses a multi-machine collaborative operation anti-collision monitoring system and method, and the system comprises a distributed sensing layer which is used for obtaining local sensing data of current equipment and neighborhood sensing data of other equipment in a preset communication range, a local situation awareness result is formed after space-time alignment and data integration; the collaborative cognitive layer performs data registration on a local situation awareness result, eliminates coordinate deviation and time error across equipment, constructs a global environment map, and establishes a group motion intention model according to the operation task type, the motion mode and the historical track of each piece of equipment; the collision prediction layer is used for predicting an equipment motion track cluster in a future time window based on the current motion state of each piece of equipment and a group motion intention model, and identifying track crossing, operation area overlapping and speed conflict conditions through space-time occupation modeling; dependence on manual intervention is reduced, and the autonomous cooperation capability of operation is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and more specifically, to a multi-machine collaborative operation anti-collision monitoring system and method. Background Technology

[0002] With the increasing application of drone swarms, mobile robot swarms, and multi-autonomous device collaborative operations, it has become commonplace for multiple devices to simultaneously perform inspection, transportation, testing, or construction tasks within a shared space. However, due to the large number of devices and their complex movement patterns, existing multi-machine collaborative operation collision avoidance technologies still have the following problems: Lack of global situational awareness: Existing systems mostly rely on local sensors of single devices for environmental perception. The lack of data sharing and neighborhood collaboration mechanisms between devices leads to fragmented perception information, limited field of view, and an inability to form a consistent global situational awareness, affecting the accuracy and reliability of overall collaboration.

[0003] Insufficient ability to predict future trajectories: Traditional prediction methods cannot comprehensively consider behavioral factors such as task, historical trajectory, and movement pattern, and also lack the ability to jointly extrapolate the future state of multiple devices, resulting in large deviations in the prediction of potential conflict locations, conflict times, and conflict probabilities.

[0004] Avoidance decisions lack global coordination: Existing technologies mostly adopt local trigger-based avoidance strategies, which cannot prioritize the conflict risks of multiple devices as a whole, nor can they reasonably allocate avoidance responsibilities. This can easily lead to simultaneous avoidance, mutual interference, or secondary conflicts, reducing collaboration efficiency.

[0005] In view of this, the present invention proposes a multi-machine collaborative operation collision avoidance monitoring system and method to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a multi-machine collaborative operation anti-collision monitoring system, comprising: The distributed perception layer is used to acquire local perception data of the current device and neighboring perception data of other devices within a preset communication range. After spatiotemporal alignment and data integration, it forms a local situational awareness result. The collaborative cognition layer performs data registration on the local situational awareness results, eliminates coordinate deviations and time errors between devices, constructs a global environment map, and establishes a group movement intention model based on the task type, movement mode, and historical trajectory of each device. The collision prediction layer predicts the cluster of device motion trajectories within a future time window based on the current motion state of each device and the group motion intention model. It identifies trajectory intersections, overlapping work areas, and speed conflicts through spatiotemporal occupancy modeling, calculates the collision probability between devices and the time difference to reach the conflict point, and constructs a global collision risk matrix. The negotiation and decision-making layer determines the master-slave relationship of avoidance based on the global collision risk matrix and generates a collaborative avoidance plan; The monitoring layer executes avoidance commands to the equipment according to the collaborative avoidance scheme and verifies the trajectory of the equipment after executing the avoidance commands in real time. When an execution deviation is detected, online replanning is triggered to drive the effective execution of the avoidance commands.

[0007] Preferably, the method for acquiring local sensing data of the current device and neighborhood sensing data of other devices within a preset communication range includes: The device collects local perception data in real time through its onboard positioning device, motion state acquisition device, and environmental perception sensor. The local perception data includes position coordinates, velocity vector, acceleration vector, attitude angle information, and obstacle point cloud information. These data are packaged into a perception data package in a unified format and include a collection timestamp. Neighborhood sensing data is provided by neighboring devices within the current device's preset communication range, and its data type is consistent with that of local sensing data.

[0008] Preferably, the method for obtaining the local situational awareness results includes: The location coordinates, attitude angle information and obstacle point cloud information contained in the neighborhood perception data are converted into the spatial coordinate system used by the current device, and the velocity vector and acceleration vector are simultaneously corrected for attitude consistency, thereby achieving spatial alignment of the neighborhood perception data. Based on the acquisition timestamps attached to the local sensing data, time correction is performed on the spatially aligned neighboring sensing data. Time interpolation is used to compensate for the sampling frequency differences between different devices, so that the reference time of the local sensing data and the neighboring sensing data remains consistent. The spatially aligned and temporally corrected neighborhood sensing data is integrated with the local sensing data to form a local situational awareness result.

[0009] Preferably, the method for constructing a global environment map includes: Static reference features are extracted from local situational awareness results to establish a unified reference feature set across devices. Based on the spatial coverage of local situational awareness results, the overlapping parts of the observation areas between different devices are identified. Feature matching is performed on the reference feature set within the overlapping areas to obtain the observation bias. Based on the observation bias, registration residual optimization is performed, and the static reference features optimized by registration residual are overlaid from multiple sources to generate a global environment map.

[0010] Preferably, the method for establishing a group movement intention model includes: Collect the task type, motion mode and historical trajectory information of each device, and perform feature extraction and normalization on the motion trajectory and motion mode of each device. Based on the equipment's historical trajectory, movement patterns, and task types, we infer the equipment's possible movement targets and behavior patterns within future time windows, forming a movement intention label for each piece of equipment; we then aggregate the movement intention labels of all equipment to construct a group movement intention model.

[0011] Preferably, the method for obtaining the device motion trajectory cluster includes: The current position coordinates, velocity vector, acceleration vector and attitude angle information of each device are extracted from the situational awareness results to form the motion state vector of the device, and then the current motion state of each device is obtained. Starting from the current moment, initialize the prediction time series; use the current motion state vector of the device as the initial input value of the group motion intention model, and obtain the motion intention label corresponding to the device; Based on the current motion status and motion intention labels of each device, the prediction model is automatically selected from the constant speed model, constant acceleration model and interactive multi-model algorithm; Based on the selected prediction model, the current motion state vector is used as the initial state. The predicted position points at each time within the future time window are calculated iteratively according to the prediction time series to form the predicted motion trajectory. The predicted motion trajectories of all devices are collected to form a cluster of device motion trajectories.

[0012] Preferably, the method for constructing the global collision risk matrix includes: Based on a unified coordinate system for the equipment's operating area, the spatial region is divided into grid cells of fixed size, and sampling intervals for the equipment's motion trajectory are applied in the time dimension. Discrete processing is performed to construct a spatiotemporal occupancy grid map containing spatial and temporal information; the predicted location points contained in the predicted motion trajectory of each device are mapped to the corresponding grid, and the spatial occupancy information of the device in different time slices in the future is recorded; The predicted motion trajectories of all equipment are analyzed pairwise to identify trajectory intersections, overlapping work areas, and speed conflicts. The collision probability of each pair of equipment is calculated, and a global collision risk matrix is ​​constructed.

[0013] Preferably, the method for generating a cooperative avoidance scheme includes: The collision probabilities in the global collision risk matrix are sorted by risk level to determine the pairs of conflicting equipment that need to be dealt with first; the current motion state and historical motion trajectory of each equipment are combined to determine the maneuverability constraint range of the equipment in order to judge the feasibility of the equipment performing avoidance operations. Determine the master-slave relationship of conflicting equipment based on the mobility constraint range and path maintenance requirements, thereby determining the master and slave equipment, so that the master equipment maintains its original working path or task status, and the slave equipment performs avoidance operations. For conflict types such as trajectory intersection, work area overlap, and speed conflict, a collaborative avoidance scheme is generated from the equipment.

[0014] Preferably, the method for effectively executing the drive avoidance command includes: The collaborative avoidance scheme issues avoidance commands to the equipment, which include heading adjustments, speed changes, or sets of detour paths. The equipment then performs maneuvers based on the avoidance commands. The system continuously collects the equipment's motion status data and compares it in real time with the target's expected trajectory corresponding to the avoidance commands. If a positional deviation or heading deviation occurs, and the heading deviation exceeds a preset heading deviation threshold, the system determines that the current avoidance maneuver is unattainable or that there is a new collision risk. The monitoring layer transmits deviation information back to the collaborative cognition layer, the collision prediction layer recalculates the motion trajectory clusters and collision risks within the future time window, and the negotiation decision layer generates a new collaborative avoidance scheme and dynamically updates the execution content of the avoidance command, driving the effective execution of the avoidance command.

[0015] A multi-machine collaborative operation collision avoidance monitoring method includes: S1. Acquire the local sensing data of the current device and the neighboring sensing data of other devices within the preset communication range, and form a local situational awareness result after spatiotemporal alignment and data integration; S2. Perform data registration on the local situational awareness results to eliminate coordinate deviations and time errors between devices, construct a global environment map, and establish a group movement intention model based on the task type, movement mode, and historical trajectory of each device. S3. Based on the current motion state of each device and the group motion intention model, predict the device motion trajectory clusters within the future time window. Identify trajectory intersections, overlapping work areas, and speed conflicts through spatiotemporal occupancy modeling. Calculate the collision probability between devices and the time difference to reach the conflict point to construct a global collision risk matrix. S4. Determine the master-slave relationship of avoidance based on the global collision risk matrix and generate a cooperative avoidance scheme; S5. Issue avoidance commands to the equipment according to the collaborative avoidance scheme, and verify the trajectory of the equipment after executing the avoidance commands in real time. When an execution deviation is detected, trigger online replanning to drive the effective execution of the avoidance commands.

[0016] Compared with the prior art, the present invention has the following beneficial effects: By collecting local and neighboring sensing data and performing spatiotemporal alignment, a consistent local situational awareness across devices can be constructed, providing a data foundation for subsequent collaborative cognition. Data registration eliminates coordinate and time discrepancies, enabling cross-device data fusion. Furthermore, based on task type, motion pattern, and historical trajectories, a group motion intention model is established, allowing the system to infer future device behavior trends and achieve higher-precision trajectory prediction. By analyzing motion trajectory clusters within future time windows, future trajectory intersections, overlapping work areas, and speed conflicts can be accurately identified, and conflict probabilities and time differences can be further calculated, achieving a quantifiable global safety risk characterization.

[0017] The avoidance master-slave relationship is rationally determined based on risk level, equipment maneuverability, and path-keeping requirements. This ensures the master equipment maintains operational continuity while the slave equipment executes targeted collaborative avoidance plans, preventing mutual interference and improving the rationality and feasibility of avoidance decisions. The actual trajectory is compared with the expected trajectory in real time. When positional deviations, heading deviations, or new risks are detected, online replanning can be triggered promptly. The decision-making level generates new avoidance plans, achieving closed-loop control of the entire avoidance process and improving the stability and reliability of avoidance behavior. The system can automatically identify conflicts, generate avoidance plans, and dynamically modify and execute them without human intervention, reducing manual intervention, mitigating collaborative operation risks, and improving operational efficiency and safety. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a multi-machine collaborative operation collision avoidance monitoring system according to the present invention; Figure 2 This is a schematic diagram of a multi-machine collaborative operation anti-collision monitoring method according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 Please see Figure 1 As shown, this embodiment provides a multi-machine collaborative operation collision avoidance monitoring system, which specifically includes the following steps: The distributed perception layer is used to acquire local perception data of the current device and neighboring perception data of other devices within a preset communication range. After spatiotemporal alignment and data integration, it forms a local situational awareness result. The collaborative cognition layer performs data registration on the local situational awareness results, eliminates coordinate deviations and time errors between devices, constructs a global environment map, and establishes a group movement intention model based on the task type, movement mode, and historical trajectory of each device. The collision prediction layer predicts the cluster of device motion trajectories within a future time window based on the current motion state of each device and the group motion intention model. It identifies trajectory intersections, overlapping work areas, and speed conflicts through spatiotemporal occupancy modeling, calculates the collision probability between devices and the time difference to reach the conflict point, and constructs a global collision risk matrix. The negotiation and decision-making layer determines the master-slave relationship of avoidance based on the global collision risk matrix and generates a collaborative avoidance plan; The monitoring layer executes avoidance commands to the equipment according to the collaborative avoidance scheme and verifies the trajectory of the equipment after executing the avoidance commands in real time. When an execution deviation is detected, online replanning is triggered to drive the effective execution of the avoidance commands.

[0021] Methods for obtaining local sensing data of the current device and neighborhood sensing data of other devices within a preset communication range include: The device collects local perception data in real time through its onboard positioning device, motion state acquisition device, and environmental perception sensor. The local perception data includes position coordinates, velocity vector, acceleration vector, attitude angle information, and obstacle point cloud information (obstacle point cloud acquired by LiDAR or camera). The data is packaged into a perception data package in a unified format and includes a collection timestamp. The neighborhood sensing data is provided by neighboring devices within the current device's preset communication range. It is consistent with the local sensing data in terms of data type. Specifically, after each device obtains its own sensing data packet, it periodically broadcasts it within the preset communication range. After receiving sensing data packets broadcast by other neighboring devices, the current device parses and integrates the contents into neighborhood sensing data.

[0022] Methods for obtaining local situational awareness results include: The location coordinates, attitude angle information and obstacle point cloud information contained in the neighborhood perception data are converted into the spatial coordinate system used by the current device, and the velocity vector and acceleration vector are simultaneously corrected for attitude consistency, thereby achieving spatial alignment of the neighborhood perception data. It should be noted that in order to ensure that the motion states of different devices can be compared under the same attitude reference, it is necessary to perform "attitude consistency correction" on the velocity vectors and acceleration vectors of neighboring devices. The core is to project the velocity vectors and acceleration vectors of neighboring devices onto the reference coordinate system used by the current device. Assume the velocity vector of neighboring device A ,in, This indicates that neighboring device A is moving forward 2 meters per second; at this time... In the reference coordinate system used by the neighboring devices, the position of the neighboring devices themselves is taken as the origin, and there is only translation and no rotation between the reference coordinate system used by the neighboring devices and the reference coordinate system used by the current device; If the attitude angle information of device A at this time is: (yaw represents the yaw angle, pointing towards the left side of the current device) ( (Indicates pitch angle) ( (Indicates the roll angle), then the attitude rotation matrix of device A can be simplified to rotation around the Z-axis. : ;in, This represents the attitude rotation matrix of device A; Will Transform from the reference coordinate system used by neighboring devices to the reference coordinate system used by the current device: ;in, The velocity vector of the neighboring device A after attitude uniformity correction reflects that the neighboring device A moves to the left at a speed of 2 m / s relative to the current device; Based on the acquisition timestamps attached to the local sensing data, time correction is performed on the spatially aligned neighboring sensing data. Time interpolation is used to compensate for the sampling frequency differences between different devices, so that the reference time of the local sensing data and the neighboring sensing data remains consistent. The neighborhood sensing data, after spatial alignment and time correction, is integrated with the local sensing data to form a local situational awareness result.

[0023] Methods for constructing a global environment map include: Static reference features that reflect the environmental structure are extracted from local situational awareness results, and a unified reference feature set across devices is established. Static reference features include the geometric shape descriptions of ground edge lines, obstacle outlines, road boundaries, as well as sparse corner points, edge points, etc. These static reference features all have the attribute of being jointly observed across devices. Based on the spatial coverage of the local situational awareness results, the system identifies the overlapping areas of the observation regions between different devices. Within the overlapping areas, feature matching is performed on the reference feature set to obtain the observation bias. This includes geometric distance matching of point cloud features, calculation of the overlap rate of linear structures, and consistency comparison of obstacle contours. Through the above matching operations, the system can quantify the observation bias of different devices in the same environmental region, such as spatial offset between point clouds and angular deviation of structural line segments. For dynamic objects appearing in the overlapping areas, in order to avoid them interfering with the static environment registration, dynamic objects are separately classified as dynamic target features, and consistency evaluation is performed based on the direction, speed change, and temporal continuity of the trajectory to ensure that the dynamic part does not affect the consistency calibration of the static environment. Based on the observation bias, registration residual optimization is performed, and the static reference features after registration residual optimization are overlaid from multiple sources to generate a global environment map.

[0024] Specifically, during the registration residual optimization process, the positional residuals of static reference features, the shape deviations of environmental boundaries, and the relative geometric structures of feature points are constrained respectively, so that the environmental descriptions formed by each device gradually converge toward a unified optimal environmental model. Through continuous iterative optimization, the problem of cross-device observation bias caused by sensing noise, sensor installation errors, or local mapping differences is effectively eliminated.

[0025] Methods for establishing group movement intention models include: Collect the task type, motion mode, and historical trajectory information of each device. The task type indicates the type of task performed by the device and the corresponding work space and constraints. The motion mode includes state modes such as uniform speed, uniform acceleration, curvilinear motion, or random motion, as well as speed, acceleration, and attitude change characteristics. The historical trajectory includes the time-series position, speed, and acceleration information of the device during the execution of previous tasks. Perform feature extraction and normalization on the motion trajectory and motion mode of each device. Based on the equipment's historical trajectory, movement pattern, and task type, the possible movement targets and behavior patterns of the equipment in future time windows are inferred, forming a movement intention label for each piece of equipment, such as stable and uniform speed operation, acceleration / deceleration, temporary change of direction / job switching, and job area switching; the movement intention labels of all equipment are aggregated to construct a group movement intention model.

[0026] It should be noted that the methods for inferring the possible motion targets and behavior patterns of the device within a future time window include: the system first starts from the perspective of a single device, and then uses: historical trajectory statistical analysis to identify target areas that the device has stayed in or visited multiple times, and extract potential target points; real-time motion trend extrapolation to infer short-term direction and approach / departure trends based on velocity vectors, acceleration, and attitude changes. Task constraint filtering and matching: Based on the current task type (inspection, transportation, following, etc.), the target candidate area is restricted; thus, multiple candidate motion intention labels for the equipment within a future time window are obtained, and corresponding probability scores are given. The system further analyzes the evolution trend of equipment behavior: if the speed is stable → predicts continued straight movement in the current direction; if acceleration and path curvature increase → infers target approach behavior; if there is lateral displacement and frequent obstacle avoidance actions → determine environmental response or detour behavior; if it maintains a relative structure with other equipment → determine formation following or cooperative operation mode.

[0027] Methods for obtaining equipment motion trajectory clusters include: The current position coordinates, velocity vector, acceleration vector and attitude angle information of each device are extracted from the situational awareness results to form the motion state vector of the device, and then the current motion state of each device is obtained. Initialize the predicted time series starting from the current time. ,in, The sampling interval for the device's motion trajectory is initially set to 0.1 seconds, which can be adjusted by staff based on actual needs or expert experience; the current motion state vector of the device is used as the initial input value for the group motion intention model to obtain the motion intention label corresponding to the device; Based on the current motion status and motion intention labels of each device, the prediction model is automatically selected from the constant speed model, constant acceleration model and interactive multi-model algorithm; Specifically, when the equipment maintains a stable and uniform speed (e.g., the absolute value of the current equipment acceleration is less than or equal to 0.2 m / s², and the speed fluctuation coefficient is less than or equal to 5%), a constant speed (CV) model is used; when the equipment is accelerating / decelerating and the direction is fixed, a constant acceleration (CA) model is used; when the equipment is undergoing temporary direction changes / operation switching and operation area switching (e.g., the rate of change of the equipment's direction angle is greater than 10 degrees per second, and the speed fluctuation coefficient is greater than 15%), an interactive multi-model algorithm is used to simultaneously consider the prediction results of multiple models. Specifically, for complex motion modes, the interactive multi-model algorithm constructs a model set containing "uniform speed sub-model, constant acceleration sub-model, and turning sub-model", and the weights of each sub-model can be dynamically adjusted through the Markov switching probability matrix. Based on the selected prediction model, the motion state vector at the current moment is used as the initial state. The predicted position points at each moment in the future time window are calculated iteratively according to the prediction time series to form the predicted motion trajectory. The predicted motion trajectories of all devices are collected to form a motion trajectory cluster.

[0028] Methods for constructing a global collision risk matrix include: Based on a unified coordinate system for the equipment's operating area, the spatial region is divided into grid cells of fixed size, and sampling intervals for the equipment's motion trajectory are applied in the time dimension. Discrete processing is performed to construct a spatiotemporal occupancy grid map containing spatial and temporal information; the predicted location points contained in the predicted motion trajectory of each device are mapped to the corresponding grid, and the spatial occupancy information of the device in different time slices in the future is recorded; The predicted motion trajectories of all equipment are analyzed pairwise to identify trajectory intersections, overlapping work areas, and speed conflicts. The collision probability of each pair of equipment is calculated, and a global collision risk matrix is ​​constructed.

[0029] It should be noted that by comparing the spatial occupancy information of different devices in the same or adjacent time slices, identifying trajectory intersections, overlapping work areas, and speed direction conflicts, it is possible to reflect various risk exposure scenarios that the equipment may encounter during future operation. When two devices occupy the same or adjacent grid cells in the same or adjacent time slices and the spatial distance is less than a preset spatial distance threshold, it is determined that there is a potential risk of trajectory intersection, and the trajectory intersection point and its probability weight are identified. Based on the probability weight, the probability of the two trajectory pairs that have intersected is accumulated to obtain the collision probability caused by the trajectory intersection. When the grid sets occupied by two devices have obvious spatial overlap, it is determined that there is overlapping of work areas; the area overlap ratio is calculated based on the grid overlap area occupied by the two devices to obtain the collision probability caused by the overlapping of work areas. For situations where there is no trajectory intersection but there is a clear trend of convergence, when the speed directions between devices are opposite or the speed angle is greater than the preset speed angle threshold, and the distance between devices shows a continuous decreasing trend, it is determined that there is a risk of speed vector conflict, and the occurrence time and predicted minimum distance are recorded. The collision probability caused by the speed direction conflict of the devices is calculated using the distance decay function.

[0030] Methods for generating cooperative avoidance schemes include: The collision probabilities in the global collision risk matrix are sorted by risk level to determine the pairs of conflicting equipment that need to be dealt with first; the current motion state and historical motion trajectory of each equipment are combined to determine the maneuverability constraint range of the equipment in order to judge the feasibility of the equipment performing avoidance operations. Specifically, this includes calculating the equipment's maneuverability indicators based on its current motion state and historical trajectory changes within a preset time window. These indicators include the maximum permissible heading deflection angle, the maximum permissible speed change, the acceptable trajectory deviation range, and the minimum time required to complete the corresponding maneuver. Based on these maneuverability indicators, a maneuverability constraint range is constructed to reflect the feasibility and extent of the equipment's ability to perform avoidance maneuvers such as detours, deceleration, or heading adjustments while maintaining the continuity of the operational task.

[0031] Determine the master-slave relationship of conflicting equipment based on the mobility constraint range and path maintenance requirements, thereby determining the master and slave equipment, so that the master equipment maintains its original working path or task status, and the slave equipment performs avoidance operations. It should be noted that path-keeping requirements are determined based on the current operational stage of the equipment and the impact of trajectory deviation on operational performance. Specifically, the path-keeping requirement is higher when the equipment is in a precise positioning, critical operation, or trajectory adjustment sensitive stage; and lower when the equipment is in a transitional phase, the task is less affected, or a backup operational path is available. By comprehensively comparing the mobility constraints and path-keeping requirements of conflicting equipment, the equipment with weaker mobility or higher path-keeping requirements is designated as the primary equipment to maintain its original path and perform the task, while the equipment with stronger mobility or lower path-keeping requirements is designated as the secondary equipment to perform avoidance operations.

[0032] For conflict types such as trajectory intersection, overlapping work areas, and speed conflicts, collaborative avoidance schemes are generated from the equipment to include detour, heading adjustment, speed change, or work area segmentation.

[0033] Methods for effectively executing drive avoidance commands include: The collaborative avoidance scheme issues avoidance commands to the equipment, which include heading adjustments, speed changes, or sets of detour paths. The equipment then performs maneuvers based on the avoidance commands. The system continuously collects the equipment's motion status data and compares it in real time with the target's expected trajectory corresponding to the avoidance commands. If a positional deviation or heading deviation occurs, and the heading deviation exceeds a preset heading deviation threshold, the system determines that the current avoidance maneuver is unattainable or that there is a new collision risk. The monitoring layer transmits deviation information back to the collaborative cognition layer, the collision prediction layer recalculates the motion trajectory clusters and collision risks within the future time window, and the negotiation decision layer generates a new collaborative avoidance scheme and dynamically updates the execution content of the avoidance command, driving the effective execution of the avoidance command.

[0034] The preset spatial distance threshold is set by staff based on historical data analysis results or expert experience; similarly, the preset speed angle threshold and preset heading deviation threshold can be adjusted by staff during system operation according to actual conditions.

[0035] In this embodiment, by collecting local and neighboring sensing data and performing spatiotemporal alignment, a consistent local situational awareness across devices can be constructed, providing a data foundation for subsequent collaborative cognition. Data registration eliminates coordinate and time discrepancies, enabling cross-device data fusion. Furthermore, a group movement intention model is established based on task type, motion pattern, and historical trajectories, allowing the system to infer future device behavior trends and achieve higher-precision trajectory prediction. By analyzing motion trajectory clusters within future time windows, future trajectory intersections, overlapping work areas, and speed conflicts can be accurately identified, and conflict probabilities and time differences can be further calculated, achieving a quantifiable global safety risk characterization.

[0036] The avoidance master-slave relationship is rationally determined based on risk level, equipment maneuverability, and path-keeping requirements. This ensures the master equipment maintains operational continuity while the slave equipment executes targeted collaborative avoidance plans, preventing mutual interference and improving the rationality and feasibility of avoidance decisions. The actual trajectory is compared with the expected trajectory in real time. When positional deviations, heading deviations, or new risks are detected, online replanning can be triggered promptly. The decision-making level generates new avoidance plans, achieving closed-loop control of the entire avoidance process and improving the stability and reliability of avoidance behavior. The system can automatically identify conflicts, generate avoidance plans, and dynamically modify and execute them without human intervention, reducing manual intervention, mitigating collaborative operation risks, and improving operational efficiency and safety.

[0037] Example 2 Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A multi-machine collaborative operation collision avoidance monitoring method is provided, including: S1. Acquire the local sensing data of the current device and the neighboring sensing data of other devices within the preset communication range, and form a local situational awareness result after spatiotemporal alignment and data integration; S2. Perform data registration on the local situational awareness results to eliminate coordinate deviations and time errors between devices, construct a global environment map, and establish a group movement intention model based on the task type, movement mode, and historical trajectory of each device. S3. Based on the current motion state of each device and the group motion intention model, predict the device motion trajectory clusters within the future time window. Identify trajectory intersections, overlapping work areas, and speed conflicts through spatiotemporal occupancy modeling. Calculate the collision probability between devices and the time difference to reach the conflict point to construct a global collision risk matrix. S4. Determine the master-slave relationship of avoidance based on the global collision risk matrix and generate a cooperative avoidance scheme; S5. Issue avoidance commands to the equipment according to the collaborative avoidance scheme, and verify the trajectory of the equipment after executing the avoidance commands in real time. When an execution deviation is detected, trigger online replanning to drive the effective execution of the avoidance commands.

[0038] Since the electronic device described in this embodiment is the electronic device used to implement the multi-machine collaborative operation collision avoidance monitoring system and method described in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the multi-machine collaborative operation collision avoidance monitoring system and method described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. As long as those skilled in the art implement the electronic device used in the multi-machine collaborative operation collision avoidance monitoring system and method described in this application embodiment, it falls within the protection scope of this application.

[0039] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0040] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A multi-machine collaborative operation collision avoidance monitoring system, characterized in that, include: The distributed perception layer is used to acquire local perception data of the current device and neighboring perception data of other devices within a preset communication range. After spatiotemporal alignment and data integration, it forms a local situational awareness result. The collaborative cognition layer performs data registration on the local situational awareness results, eliminates coordinate deviations and time errors between devices, constructs a global environment map, and establishes a group movement intention model based on the task type, movement mode, and historical trajectory of each device. The collision prediction layer predicts the cluster of device motion trajectories within a future time window based on the current motion state of each device and the group motion intention model. It identifies trajectory intersections, overlapping work areas, and speed conflicts through spatiotemporal occupancy modeling, calculates the collision probability between devices and the time difference to reach the conflict point, and constructs a global collision risk matrix. The negotiation and decision-making layer determines the master-slave relationship of avoidance based on the global collision risk matrix and generates a collaborative avoidance plan; The monitoring layer executes avoidance commands to the equipment according to the collaborative avoidance scheme and verifies the trajectory of the equipment after executing the avoidance commands in real time. When an execution deviation is detected, online replanning is triggered to drive the effective execution of the avoidance commands.

2. The multi-machine collaborative operation anti-collision monitoring system according to claim 1, characterized in that, The method for acquiring local sensing data of the current device and neighboring sensing data of other devices within a preset communication range includes: The device collects local perception data in real time through its onboard positioning device, motion state acquisition device, and environmental perception sensor. The local perception data includes position coordinates, velocity vector, acceleration vector, attitude angle information, and obstacle point cloud information. These data are packaged into a perception data package in a unified format and include a collection timestamp. Neighborhood sensing data is provided by neighboring devices within the current device's preset communication range, and its data type is consistent with that of local sensing data.

3. The multi-machine collaborative operation anti-collision monitoring system according to claim 2, characterized in that, The method for obtaining the local situational awareness results includes: The location coordinates, attitude angle information and obstacle point cloud information contained in the neighborhood perception data are converted into the spatial coordinate system used by the current device, and the velocity vector and acceleration vector are simultaneously corrected for attitude consistency, thereby achieving spatial alignment of the neighborhood perception data. Based on the acquisition timestamps attached to the local sensing data, time correction is performed on the spatially aligned neighboring sensing data. Time interpolation is used to compensate for the sampling frequency differences between different devices, so that the reference time of the local sensing data and the neighboring sensing data remains consistent. The neighborhood sensing data, after spatial alignment and time correction, is integrated with the local sensing data to form a local situational awareness result.

4. The multi-machine collaborative operation anti-collision monitoring system according to claim 3, characterized in that, The method for constructing a global environment map includes: Static reference features are extracted from local situational awareness results to establish a unified reference feature set across devices. Based on the spatial coverage of local situational awareness results, the overlapping parts of the observation areas between different devices are identified. Feature matching is performed on the reference feature set within the overlapping areas to obtain the observation bias. Based on the observation bias, registration residual optimization is performed, and the static reference features optimized by registration residual are overlaid from multiple sources to generate a global environment map.

5. The multi-machine collaborative operation anti-collision monitoring system according to claim 4, characterized in that, The method for establishing a group movement intention model includes: Collect the task type, motion mode and historical trajectory information of each device, and perform feature extraction and normalization on the motion trajectory and motion mode of each device. Based on the equipment's historical trajectory, movement patterns, and task types, we infer the equipment's possible movement targets and behavior patterns within future time windows, forming a movement intention label for each piece of equipment; we then aggregate the movement intention labels of all equipment to construct a group movement intention model.

6. The multi-machine collaborative operation anti-collision monitoring system according to claim 5, characterized in that, The method for obtaining the device motion trajectory cluster includes: The current position coordinates, velocity vector, acceleration vector and attitude angle information of each device are extracted from the situational awareness results to form the motion state vector of the device, and then the current motion state of each device is obtained. Starting from the current moment, initialize the prediction time series; use the current motion state vector of the device as the initial input value of the group motion intention model, and obtain the motion intention label corresponding to the device; Based on the current motion status and motion intention labels of each device, the prediction model is automatically selected from the constant speed model, constant acceleration model and interactive multi-model algorithm; Based on the selected prediction model, the current motion state vector is used as the initial state. The predicted position points at each time within the future time window are calculated iteratively according to the prediction time series to form the predicted motion trajectory. The predicted motion trajectories of all devices are collected to form a cluster of device motion trajectories.

7. A multi-machine collaborative operation collision avoidance monitoring system according to claim 6, characterized in that, The method for constructing the global collision risk matrix includes: Based on a unified coordinate system for the equipment's operating area, the spatial region is divided into grid cells of fixed size, and sampling intervals for the equipment's motion trajectory are applied in the time dimension. Discrete processing is performed to construct a spatiotemporal occupancy grid map containing spatial and temporal information; the predicted location points contained in the predicted motion trajectory of each device are mapped to the corresponding grid, and the spatial occupancy information of the device in different time slices in the future is recorded; The predicted motion trajectories of all equipment are analyzed pairwise to identify trajectory intersections, overlapping work areas, and speed conflicts. The collision probability of each pair of equipment is calculated, and a global collision risk matrix is ​​constructed.

8. The multi-machine collaborative operation anti-collision monitoring system according to claim 7, characterized in that, The method for generating a cooperative avoidance scheme includes: The collision probabilities in the global collision risk matrix are sorted by risk level to determine the pairs of conflicting equipment that need to be dealt with first; the current motion state and historical motion trajectory of each equipment are combined to determine the maneuverability constraint range of the equipment in order to judge the feasibility of the equipment performing avoidance operations. Determine the master-slave relationship of conflicting equipment based on the mobility constraint range and path maintenance requirements, thereby determining the master and slave equipment, so that the master equipment maintains its original working path or task status, and the slave equipment performs avoidance operations. For conflict types such as trajectory intersection, work area overlap, and speed conflict, a collaborative avoidance scheme is generated from the equipment.

9. A multi-machine collaborative operation anti-collision monitoring system according to claim 8, characterized in that, The method for effectively executing the drive avoidance command includes: The collaborative avoidance scheme issues avoidance commands to the equipment, which include heading adjustments, speed changes, or sets of detour paths. The equipment then performs maneuvers based on the avoidance commands. The system continuously collects the equipment's motion status data and compares it in real time with the target's expected trajectory corresponding to the avoidance commands. If a positional deviation or heading deviation occurs, and the heading deviation exceeds a preset heading deviation threshold, the system determines that the current avoidance maneuver is unattainable or that there is a new collision risk. The monitoring layer transmits deviation information back to the collaborative cognition layer, the collision prediction layer recalculates the motion trajectory clusters and collision risks within the future time window, and the negotiation decision layer generates a new collaborative avoidance scheme and dynamically updates the execution content of the avoidance command, driving the effective execution of the avoidance command.

10. A multi-machine collaborative operation collision avoidance monitoring method, implemented by any one of claims 1 to 9, characterized in that, include: S1. Acquire the local sensing data of the current device and the neighboring sensing data of other devices within the preset communication range, and form a local situational awareness result after spatiotemporal alignment and data integration; S2. Perform data registration on the local situational awareness results to eliminate coordinate deviations and time errors between devices, construct a global environment map, and establish a group movement intention model based on the task type, movement mode, and historical trajectory of each device. S3. Based on the current motion state of each device and the group motion intention model, predict the device motion trajectory clusters within the future time window. Identify trajectory intersections, overlapping work areas, and speed conflicts through spatiotemporal occupancy modeling. Calculate the collision probability between devices and the time difference to reach the conflict point to construct a global collision risk matrix. S4. Determine the master-slave relationship of avoidance based on the global collision risk matrix and generate a cooperative avoidance scheme; S5. Issue avoidance commands to the equipment according to the collaborative avoidance scheme, and verify the trajectory of the equipment after executing the avoidance commands in real time. When an execution deviation is detected, trigger online replanning to drive the effective execution of the avoidance commands.

Images