CN-121999642-A - Base station integrating sensing and calculation and low-altitude space domain unmanned aerial vehicle control method based on base station

CN121999642ACN 121999642 ACN121999642 ACN 121999642ACN-121999642-A

Abstract

The invention discloses a base station integrating perception and calculation and a low-altitude space-domain unmanned aerial vehicle management and control method based on the base station. The base station comprises a communication base station body, an airspace sensing module and an edge computing module, wherein the airspace sensing module and the edge computing module are integrated on the communication base station body. The airspace sensing module collects airspace situation data in the coverage area of the base station in real time. The method comprises the steps of establishing a multi-factor dynamic risk potential field model based on airspace situation data and running state parameters by an edge computing force module at a base station, processing the situation data by a track prediction model to generate a predictive risk model, planning a flight path for actively avoiding future risks for an unmanned plane through a path optimization algorithm based on the predictive risk model, and generating a management and control instruction. The invention sinks the perception and calculation to the network edge, obviously reduces the end-to-end delay, improves the communication reliability and the system robustness, and realizes the distributed intelligent management and control of the unmanned aerial vehicle with high real-time, high safety and high reliability.

Inventors

CHENG CHENGQI
WU XUEMIN
HU XUELIAN

Assignees

北斗伏羲信息技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260104

Claims (20)

1. A perception and computing integrated base station, comprising: A communication base station body; A airspace sensing module integrated with the communication base station body for collecting airspace situation data in the coverage area of the base station in real time, and An edge computing module integrated with the communication base station body and connected with the airspace perception module, the edge computing module being configured to: Constructing a space-time risk assessment model for representing the safety state of the airspace based on the airspace situation data and at least one running state parameter; Processing the airspace situation data through a track prediction model to predict the future track of the dynamic target in the airspace situation data so as to generate a predictive risk model, and And planning a flight path for the unmanned aerial vehicle through a path optimization algorithm based on the predictive risk model, and generating a corresponding management and control instruction.
2. The base station of claim 1, wherein the space-time risk assessment model is a multi-factor dynamic risk potential field model, and wherein the at least one operational state parameter comprises a quality of a communication link between the drone and the base station and a quality of global navigation satellite system, GNSS, signal.
3. The base station of claim 2, wherein the multi-factor dynamic risk potential model is a unified four-dimensional risk potential that quantifies spatial security states into continuous risk levels that are a function of a threat level of the dynamic target, the communication link quality, the GNSS signal quality, and static obstacle data.
4. The base station of claim 1, wherein the trajectory prediction model is a kalman filter.
5. The base station of claim 4, wherein the edge computing module is configured to predict a trajectory of the dynamic object within a predetermined time window in the future based on current and historical state data of the dynamic object using the kalman filter, the predetermined time window being 3 to 5 seconds.
6. The base station of claim 1, wherein the path optimization algorithm is a modified a-algorithm, and wherein a cost function of the a-algorithm is defined as an accumulated potential energy integral of the unmanned aerial vehicle flying in the predictive risk model.
7. The base station of claim 6, wherein the edge computing power module is configured to determine the flight path by solving for a minimum of the accumulated potential energy integral, thereby enabling the flight path to actively avoid a high risk region to be formed characterized by the predictive risk model.
8. The base station of claim 1, wherein the airspace sensing module comprises millimeter wave radar and a high-definition photoelectric sensor, and wherein the edge computing module is configured to perform a cooperative sensing mechanism of "radar search-photoelectric confirmation" to improve the accuracy of dynamic target recognition and tracking.
9. The base station of claim 1, wherein the edge computing power module comprises an edge computing chip supporting artificial intelligence reasoning for executing the trajectory prediction model and the path optimization algorithm, and a local storage module for caching Beidou grid data and historical flight trajectory data.
10. The base station of claim 9, wherein the edge computing chip is an artificial intelligence inference chip supporting a general parallel computing architecture, the local storage module is a high-speed solid state memory, and the edge computing power module is further configured to utilize cached historical flight trajectory data to perform offline training and iteratively optimize internal parameters of the trajectory prediction model in a low-task load period through cooperation with other base stations by a federal learning method, so that the prediction accuracy of the model on a dynamic target trajectory is continuously improved on the premise of not sharing original sensitive data.
11. A low-altitude space domain unmanned aerial vehicle control method based on an integrated base station is characterized by comprising the following steps: Acquiring airspace situation data in the coverage area of a communication base station in real time through an airspace sensing module deployed on the base station; an edge computing power module deployed on the communication base station is used for constructing a space-time risk assessment model representing the space security state locally on the basis of the space situation data and at least one running state parameter; The edge computing power module processes the airspace situation data by utilizing a track prediction model to predict the future track of the dynamic target in the airspace situation data so as to generate a predictive risk model, and The edge calculation module plans a flight path for the unmanned aerial vehicle through a path optimization algorithm based on the predictive risk model, and generates and issues corresponding management and control instructions.
12. The method of claim 11, wherein the step of constructing a spatiotemporal risk assessment model comprises constructing a multi-factor dynamic risk potential field model, wherein the at least one operational state parameter comprises communication link quality and GNSS signal quality.
13. The method of claim 12, wherein the step of constructing a multi-factor dynamic risk potential field model further comprises uniformly mapping discrete risk factors, including dynamic target threats, communication link quality, GNSS signal quality, and static obstructions, into a four-dimensional space-time grid, thereby converting the spatial security state into a continuous and quantifiable risk potential field.
14. The method of claim 11, wherein the trajectory prediction model is a kalman filter.
15. The method of claim 14, wherein the step of generating a predictive risk model comprises outputting predicted state vectors for the dynamic object at a plurality of discrete time steps in the future using the Kalman filter, and constructing a series of risk potential field snapshots at future times based on the predicted state vectors, thereby forming the predictive risk model.
16. The method of claim 11, wherein the step of planning the flight path employs a modified a-algorithm, and wherein a cost function of the a-algorithm is defined as an accumulated potential energy integral of the unmanned aerial vehicle flying in the predictive risk model.
17. The method of claim 16 wherein the step of planning the flight path using the modified a-algorithm specifically includes initializing a graph search space based on the predictive risk model, setting a path cost function F (n) for any node n in the search space to G (n) +h (n), where G (n) is an actual accumulated potential energy integral from a starting node to node n and H (n) is a predicted accumulated potential energy integral from node n to a target node, and wherein the graph search is constrained to extend only between nodes that have predicted communication link quality above a first preset threshold and predicted GNSS signal quality above a second preset threshold to ensure that the planned flight path combines low risk characteristics with communication robustness.
18. The method of claim 11, further comprising the step of issuing the control instruction by the edge computing module through a double-link communication mechanism consisting of a main link based on a 5.8GHz ISM band and a standby link based on Beidou short messages, wherein the issuing process adopts a national cipher SM4 algorithm for end-to-end encryption.
19. The method of claim 11, further comprising synchronizing data between the edge computing module and an edge computing module of an adjacent integrated base station within a predetermined synchronization period, wherein the edge computing module shares grid data of a boundary area of a respective coverage area to ensure seamless engagement of management and control information when the unmanned aerial vehicle flies across the base stations.
20. The method of claim 11, further comprising implementing a three-level fault tolerance architecture of "base station level hardware redundancy, regional level cooperative fault tolerance, central level monitoring takeover", when a single base station hardware fault is monitored, the neighboring base stations cooperatively take over their management and control regions to ensure continuity of the regional management and control tasks.

Description

Base station integrating sensing and calculation and low-altitude space domain unmanned aerial vehicle control method based on base station Technical Field The invention relates to the technical field of low-altitude airspace management, in particular to the cross fields of unmanned aerial vehicle management and control, edge calculation, beidou navigation application and wireless communication technology. More specifically, the invention relates to a base station integrating perception and calculation and a low-altitude space unmanned aerial vehicle control method based on the base station, aiming at improving the safety, instantaneity and control efficiency of the low-altitude space unmanned aerial vehicle flight. Background In recent years, with rapid progress of unmanned aerial vehicle technology and continuous widening of application scenes, unmanned aerial vehicle activities in a low-altitude airspace (usually referred to as an airspace below 1000 meters above the ground) are increasingly frequent, and the unmanned aerial vehicle is widely applied to a plurality of key fields such as logistics distribution, power inspection, emergency rescue, urban management and the like. However, the environment of the low-altitude airspace is extremely complex, and fills static barriers such as buildings, high-voltage lines, trees and dynamic targets such as birds and other aircrafts. The proliferation of the number of unmanned aerial vehicles and the interleaving of complex environments provide unprecedented challenges for the real-time performance, accuracy and system reliability of air-space management and control. The existing mainstream technical scheme adopts a low-altitude airspace three-dimensional grid division technology based on Beidou grid codes. The technology divides a low-altitude space domain into standardized three-dimensional grid cells according to preset precision (for example, 10 meters multiplied by 5 meters), and endows each cell with a unique Beidou grid code identifier. The method can provide an accurate and uniform reference frame for the spatial position of the unmanned aerial vehicle. In a typical management and control architecture, an unmanned aerial vehicle acquires airspace situation data such as its own position, speed, heading, etc. through airborne equipment, and uploads the data to a remote, centralized management and control center. The center gathers data of all unmanned aerial vehicles in the area, performs computation-intensive tasks such as grid data processing, flight conflict detection and path planning in a unified mode, and finally transmits generated control instructions back to the unmanned aerial vehicles. However, this traditional centralized management architecture exposes several inherent technical bottlenecks when dealing with large-scale, high-density unmanned aerial vehicle operation: First, the centralized power architecture results in processing delays that are too high. All unmanned aerial vehicle situation data and management and control requests are collected to a remote center, and when the unmanned aerial vehicle density in the space exceeds a certain threshold (for example, the urban core area reaches 100 frames per square kilometer) or the grid data volume reaches the GB level, the computational load of a central server is increased sharply. The delay time of data processing may surge from the conventional 50ms to over 200 ms, which is far beyond the safe delay required for real-time obstacle avoidance and decision making of the drone (typically no more than 50ms is required). Such a long delay means that the unmanned aerial vehicle has already flown several meters when the control command is issued, and may have missed the optimal risk avoidance window, thereby causing a collision risk. Second, the reliability of the long-range communication link is inadequate. The communication distance between the drone and the remote center is typically 5 to 20 km, depending on the public network (e.g., 4G/5G) or dedicated links. In urban environments, factors such as shielding of tall buildings, interference of electromagnetic environments and the like can cause serious signal attenuation (up to 30% -50%) and data packet loss. Critical control commands, such as emergency hover or avoidance commands, will directly threaten flight safety once they are not delivered in time due to transmission failure or delay. Meanwhile, long-distance transmission passes through a plurality of network nodes, and network security risk that data are intercepted or tampered maliciously is increased. Third, the separation of perception and computation results in situational information lags. Sensing devices such as radars, photoelectric sensors and the like in the space domain are usually deployed independently, and acquired data can form a complete space situation diagram in a management and control center only through a chain of sensing, transmission and processing. This process has a