A UAV Flight Control Method Based on Target Recognition

Abstract

This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology and discloses a UAV flight control method based on target recognition. The method includes: a ground control station generating and issuing a strategy switching command containing a strategy identifier; the UAV receiving the command and its local flight strategy state machine autonomously and continuously generating control signals to execute the corresponding flight strategy; the UAV also calculating a timeout threshold based on its own health status and using this threshold to monitor the communication heartbeat; and forcibly switching to a safety strategy when the conditions are met. This invention separates flight control decision-making from execution, so that the continuity of UAV mission execution is no longer limited by the real-time integrity of the communication link. Simultaneously, the triggering conditions of its safety strategy can be adaptively adjusted according to the UAV's health status, improving the system's mission resilience in complex environments and its operational reliability throughout its entire lifecycle.

Description

Technical Field

[0001] This invention relates to a flight control method for unmanned aerial vehicles (UAVs) based on target recognition, belonging to the field of UAV flight control technology. Background Technology

[0002] Currently, a common technical approach involves a ground control station continuously locating the target and sending specific flight commands to the UAV in real time. The UAV then acts as the execution end to complete the flight actions defined by the commands. This approach concentrates decision-making and perception at the ground station and can complete control tasks under the conditions of stable communication links and continuous target behavior. Its direct control logic and functional division represent the basic technical practice in this field.

[0003] When this type of drone is applied to complex and dynamic scenarios such as urban environments, each flight action of the drone depends on the instructions received at the current moment. The closure of the entire control loop depends on the continuous and unobstructed communication link. Once this premise is temporarily interrupted due to physical obstruction or signal interference, the drone will perform preset safety actions such as hovering due to the loss of instruction input, and its core tracking task will also be interrupted.

[0004] To address the above situation, a direct approach is to improve the performance of communication technology. However, this approach does not change the system architecture's dependence on communication continuity. In practical applications, intermittent signal interruptions are an objective physical limitation. Improving communication hardware performance can only reduce the probability of interruptions, but cannot solve the inevitable task interruption problem that will occur during foreseeable communication interruptions, such as when the target enters a tunnel or behind a large building. Specifically, existing technologies have the following operational limitations: 1. The continuity of flight control and the integrity of the communication link are strictly bound together in the architecture, resulting in a lack of resilience in task execution when facing signal interruptions; 2. The instructions issued by the ground station only define the instantaneous physical state target, without including strategic intentions to deal with the future behavior trends of the target, leaving the UAV without the basis to perform any forward-looking autonomous maneuvers after losing communication. Therefore, how to establish a flight control method that separates the continuity of UAV flight control from the real-time integrity of its communication link, so that the UAV can still maintain the continuous execution of the task autonomously according to the preset strategic intentions when communication is intermittently interrupted, becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a target recognition-based UAV flight control method, the main purpose of which is to solve the problem in the prior art that the flight control method relies on the real-time integrity of the communication link, which causes the UAV to be unable to maintain mission continuity when faced with unstable communication or foreseeable signal interruption.

[0006] To achieve the above objectives, the present invention provides a flight control method for unmanned aerial vehicles (UAVs) based on target recognition, comprising the following steps:

[0007] Step a: The ground control station generates a policy switching instruction based on the target identification information and environmental context information. The policy switching instruction includes a policy identifier, context parameters required to execute the policy, and effective time parameters.

[0008] Step b: The ground control station sends the strategy switching command to the UAV via the communication link;

[0009] Step c: The UAV receives a strategy switching instruction and, based on the instruction, switches its locally running flight strategy state machine to the flight strategy specified by the strategy identifier.

[0010] Step d: Under the guidance of the flight strategy, the UAV autonomously and continuously generates flight control signals to adjust its own flight status.

[0011] Step e: During the execution of a flight strategy, the UAV monitors the energy consumption data associated with that strategy, associates the energy consumption data with the strategy identifier, and sends it to the ground control station. The ground control station references the energy consumption data when generating subsequent strategy switching commands.

[0012] Step f: The UAV monitors the communication heartbeat with the ground control station and calculates a timeout threshold in each risk assessment cycle based on a system health index that represents its own health status and a preset threshold function. When the interruption duration of the communication heartbeat meets the timeout threshold, or the current time exceeds the valid time parameter, the flight strategy state machine is forced to switch to the preset safety strategy.

[0013] Preferably, the flight strategy included in the flight strategy state machine includes at least: a stable escort strategy, under which the UAV maintains synchronous flight with the target based on the relative position offset defined by the context parameters; and a predictive crossing strategy, under which the UAV autonomously plans and executes a route to a predetermined location near the target exit point based on the target entry point, target exit point and estimated travel time defined by the context parameters.

[0014] Preferably, the strategy switching instruction is structured data, and the context parameters include at least: the target relative position offset required when executing the stable escort strategy, or the target entry point coordinates, target exit point coordinates, and estimated passage time required when executing the predictive crossing strategy.

[0015] Preferably, the safety strategy in step f is a high-level hovering observation strategy or a return-to-home strategy.

[0016] Preferably, when performing step e, the UAV will also monitor the sensor noise level or actuator response delay associated with the flight strategy and send the monitoring results to the ground control station.

[0017] Preferred system health index A standardized value calculated by the drone based on energy consumption data; the timeout threshold calculated by the threshold function. With system health index It exhibits a monotonically decreasing relationship.

[0018] Preferred timeout threshold The following function is used to calculate: in, This is the base timeout threshold corresponding to the drone under a preset baseline health condition; This is the minimum timeout threshold corresponding to the drone under the preset minimum health condition; This is the system health index, and its value ranges from 0 to 1.

[0019] Preferably, before executing step a to generate the strategy switching command, the ground control station further includes: determining, based on environmental context information, that the target is about to enter a communication signal obstruction zone; and when the determination result is yes, generating and sending a strategy switching command with the predicted crossing strategy content in advance.

[0020] Preferably, the energy consumption data includes at least the average current and average voltage of the UAV during the execution of different flight strategies; in step e, the UAV associates the average current and average voltage with the corresponding strategy identifier to form a strategy energy consumption mapping table, and sends the strategy energy consumption mapping table to the ground control station.

[0021] Preferably, the flight strategy state machine is a finite state machine, whose state set consists of a stable escort strategy, a predictive crossing strategy, and a high-altitude hovering observation strategy. The state machine is in only one of the strategy states at any given time.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. By generating and sending strategy switching commands from the ground control station, and setting up a flight strategy state machine on the UAV to receive and execute these commands, the process of issuing and executing control commands has been reconstructed. Under this process, the ground control station focuses on making advanced strategy judgments and decisions based on target information and environmental context, while the UAV autonomously and continuously generates and adjusts flight control signals locally according to the received strategy commands. This arrangement logically separates the UAV's tactical maneuver execution from the real-time continuity of the communication link with the ground station. When the UAV enters a foreseeable area of ​​unstable or interrupted communication, the continuity of its mission execution no longer depends on the instantaneous fluctuations in communication quality, but is guaranteed by the flight strategy that it has already received and executed locally.

[0024] 2. By establishing an internal dynamic collaborative adjustment mechanism, the triggering conditions of flight safety strategies can be adaptively adjusted according to the health status of the UAV itself. During the execution of flight strategies, the UAV monitors and records energy consumption data related to the strategy. This data is used to assess the current system health status of the UAV. At the same time, the UAV monitors the communication heartbeat with the ground station and uses the system health status as a direct basis for adjusting the communication heartbeat interruption judgment threshold. In this way, when the UAV is in good health, the system has a high tolerance for communication latency, avoiding interruption of normal tasks due to temporary network fluctuations. When its health status declines, the system will automatically tighten the safety boundary and make a more sensitive protective response to the same communication latency.

[0025] 3. By monitoring and correlating flight strategies with actual energy consumption data at the UAV end, and then sending this correlated data to the ground control station, an information feedback path is established from the tactical execution end to the strategic decision-making end. After receiving the energy consumption data, when the ground control station generates subsequent strategy switching instructions, its decision-making basis is no longer just external target and environmental information, but also includes the real physical cost of the UAV executing different flight strategies in the current environment. This makes every strategic decision of the ground control station based on the current actual capabilities of the front-end execution unit, avoiding the potential mission failure risk caused by issuing flight instructions that exceed the current endurance of the UAV, and making the entire control and decision-making loop a sustainable system with resource awareness capabilities. Attached Figure Description

[0026] Figure 1 This is a flowchart of the target identification and flight control decision-making process of the present invention;

[0027] Figure 2 This is a timing diagram of the strategy switching for dealing with signal interruptions in this invention;

[0028] Figure 3 This is a diagram of the separate collaborative control architecture of the ground station and the UAV in this invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0030] This invention provides a target recognition-based UAV flight control method. Its architecture is built upon a control system where a ground control station and a UAV work collaboratively. The ground control station, as the decision-making unit, analyzes target state and environmental information to generate strategy commands, while the UAV, as the execution unit, autonomously adjusts its own flight state (a non-electrical variable) upon receiving the strategy commands. The two interact via a wireless communication link. In a specific application scenario, such as using a UAV to continuously track vehicles in an urban environment, the method is implemented through the following steps: The ground control station executes step a, generating a strategy switching command based on target recognition information and environmental context information. When facing situations where the target may be temporarily obscured (e.g., the vehicle is about to enter a tunnel), continuously issuing coordinate commands will cause task interruption due to command interruption. To address this, the strategy switching command generated by this method is a structured data packet containing a strategy identifier, context parameters required to execute the strategy, and effective time parameters. For example, when the target vehicle is traveling on an open road, the strategy switching command generated and issued by the ground control station could be: {Strategy identifier: Stable} Escort, context parameter: {offset: [10, 10, 20]}, effective time parameter: a preset instruction expiration time. Here, the strategy identifier Stable Escort corresponds to the stable escort strategy. The context parameter defines the relative position offset between the UAV and the target. When the environmental context information indicates that there is a tunnel in front of the target vehicle, the ground control station generates a strategy switching instruction in advance with the content of the predicted crossing strategy. Its content can be: {strategy identifier: Predictive Transit, context parameter: {entry point: [lat1, lon1], exit point: [lat2, lon2], estimated time: 15}, effective time parameter: an instruction expiration time estimated based on the passage time. The strategy identifier Predictive Transit corresponds to the predicted crossing strategy. The context parameter includes the tunnel entrance point coordinates, exit point coordinates, and the estimated passage time of 15 seconds based on the vehicle's current speed. In this way, the content of the instruction issued by the ground control station changes from a single physical location information to strategy information containing tactical intentions.

[0031] Subsequently, in step b, the ground control station sends the strategy switching command to the UAV via the communication link. After receiving the strategy switching command, the UAV executes steps c and d. The core of this method lies in using a locally running flight strategy state machine to parse and execute the command. To address the limitation of only being able to perform preset actions such as hovering after losing the command, this method sets up a finite state machine in the UAV flight control software. Each state of this state machine corresponds to a flight strategy, such as a stable escort strategy, a predictive crossing strategy, or a high-altitude hovering observation strategy. After receiving the strategy switching command, the UAV parses the strategy identifier and switches its local flight strategy state machine to the flight strategy state specified by the identifier. Then, under the rules of this strategy, the UAV autonomously and continuously generates flight control signals to adjust its flight state. For example, when the state machine enters the stable escort strategy, the flight control system uses onboard sensor data and the offset parameter in the command to continuously calculate and generate control signals to send to each motor through a PID controller to maintain alignment with the target. Relative position; when the state machine switches to the predictive crossing strategy, the UAV autonomously plans and executes a route to a predetermined location near the tunnel exit point based on the entrance / exit coordinates and estimated travel time in the command. This mechanism closes the high-frequency flight attitude adjustment task locally on the UAV, so that the continuity of task execution is no longer strictly limited by the real-time performance of the communication link. In order to enable the ground control station to make decisions based on the actual operating status of the UAV, this method also includes step e, establishing a state information feedback path from the execution end to the decision end. During the execution of the flight strategy, the UAV will monitor the energy consumption data associated with the strategy, such as average current and average voltage, and associate these energy consumption data with the identifier of the current strategy to form a strategy energy consumption correspondence table, which is sent back to the ground control station through the communication link. After receiving the data, when the ground control station generates subsequent strategy switching commands, its decision model can refer to the actual energy consumption cost of the UAV executing different strategies, thereby helping to avoid the mission risks caused by issuing commands that exceed the current endurance of the UAV.

[0032] To address the issue that the communication interruption judgment threshold is difficult to adapt to the changing states of a UAV throughout its entire lifecycle, this method introduces step f, an adaptive safety protection mechanism. During flight, the UAV calculates a standardized system health index that characterizes its own health status based on monitored data such as energy consumption. Its value ranges from 0 to 1, with 1 representing the optimal state. When the UAV continuously monitors the communication heartbeat with the ground control station, it does not use a fixed timeout threshold, but rather a preset threshold function based on the current system health index. A dynamic timeout threshold is calculated in real time. The threshold function is designed to correlate with the system health index. The relationship is monotonically decreasing, and a specific calculation formula is as follows: In the formula, This is a dynamically calculated timeout threshold, in seconds. For drones in baseline health condition ( The corresponding base timeout threshold can be set to 3.0 seconds; To ensure the drone is in a preset minimum health condition ( The minimum timeout threshold corresponding to (close to 0) can be set to 0.8 seconds; This is the currently calculated system health index, and it is a dimensionless value; for example, when the drone is in a brand new state, At that time, the calculated The time is 3.0 seconds; after prolonged use, the drone's health index drops to... At that time, the calculated The timeout period is automatically adjusted to 1.9 seconds. When the UAV detects that the interruption duration of the communication heartbeat meets the currently calculated timeout threshold, or the current time exceeds the valid time parameter in the strategy instruction, its flight strategy state machine is forcibly switched to the preset safety strategy, such as the high-level hovering observation strategy or the return-to-home strategy. This mechanism enables the system's safety response threshold to be dynamically adjusted according to the health status of the UAV itself. When the UAV is in good condition, the system has a high tolerance for communication delay. When its condition deteriorates, the safety boundary is automatically tightened, and a more sensitive protective response is made to the same communication delay.

[0033] Example 1: In an urban security tracking scenario, a UAV using the aforementioned flight control method is monitoring a high-speed target vehicle. The vehicle is traveling along a complex elevated bridge area and is about to enter a 200-meter-long underpass. This area suffers from visual obstruction and wireless signal attenuation. Initially, the ground control station sends a strategy switching command with the strategy identifier "Stable Escort" to the UAV based on continuous target identification information. After receiving the command, the UAV's local flight strategy state machine maintains a stable escort strategy and maintains a fixed escort flight state with the target vehicle based on the relative position offset parameters included in the command. When the environmental context information indicates that the target vehicle will enter the tunnel entrance in 10 seconds at the current speed, if the real-time command is not relied upon, the interruption of visual and positional information after the vehicle enters the tunnel will cause the UAV to switch to hovering due to loss of input, thus interrupting the tracking task. To cope with this foreseeable task interruption, the ground control station makes a decision based on the environmental context information 8 seconds before the target vehicle enters the tunnel, generating a strategy in advance. A strategy switching command was issued, with the strategy identifier PredictiveTransit. Its context parameters included the coordinates of the tunnel entrance and exit points, as well as the estimated travel time of 8 seconds calculated based on the vehicle's current speed. Upon receiving the command, the UAV's flight strategy state machine switched from a stable escort strategy to a predictive transit strategy. At this point, the UAV's flight control changed from relying on continuous external commands to local autonomous execution. The UAV did not hover at the tunnel entrance, but instead autonomously planned and executed a flight path that flew over the tunnel and towards a predetermined airspace near the exit point with low power consumption, based on the received entrance and exit coordinates and travel time.

[0034] During this autonomous flight, there was a 2.5-second delay in the communication heartbeat between the UAV and the ground station. At this time, the UAV's internal adaptive safety protection mechanism activated, first obtaining the system health index that represents the current health status of the UAV. It is 0.9, and based on a preset threshold function. The calculation yields the current dynamic timeout threshold as follows: Because the 2.5-second interruption duration did not meet the 2.78-second timeout threshold, the safety policy was not triggered, allowing the predictive traversal strategy to continue. This forward-looking maneuver was not interrupted by temporary network fluctuations. When the target vehicle exited the tunnel, the drone had already reached the predetermined position and immediately reacquired the target. The ground control station then resumed transmitting Stable data. The Escort strategy switching command maintained continuous tracking of the target vehicle even after a complete signal and visual interruption. After the UAV re-established stable escort of the target vehicle, the ground control station received an updated strategy energy consumption table reported by the UAV. This table showed that the actual average power consumption of the previous predicted crossing strategy was 20% higher than the baseline value. Based on this, it was calculated that the UAV's remaining flight time was lower than the preset return-to-home safety threshold. Therefore, the ground control station did not instruct the UAV to continue high-energy close tracking. Instead, it generated and issued a strategy switching command with the strategy identifier "high-level hovering observation strategy" and set the effective time parameter of the command to 120 seconds. After the UAV received and executed the command, at the high-level hovering observation point, the communication link with the ground station was completely blocked due to electromagnetic interference, and no new strategy switching commands could be received. After 120 seconds, the UAV's local timer determined that the current time had exceeded the effective time parameter of the executed command, and then forcibly switched its flight strategy state machine to the preset return-to-home strategy and autonomously started the return-to-home procedure.

[0035] Example 2: To verify the performance of the flight control method of the present invention in an unstable communication environment, a set of comparative experiments was set up. The experiment used a hardware-in-the-loop simulation platform, which consists of UAV flight control hardware running the control algorithm, a computer simulating UAV dynamics and environment, and a wireless channel simulator to reproduce signal delay and interruption. The experiment set up a control group using the traditional real-time command control method and an experimental group using the method of the present invention. First, to verify the continuous execution capability of the task under foreseeable signal interruption, the experiment simulated a scenario where a vehicle enters a tunnel, causing a fixed interruption of communication and visual signals for 10 seconds. The UAV in the control group hovered after losing the target position information and could not continue tracking after the target left the tunnel. The UAV in the experimental group executed a pre-issued predictive crossing strategy and was already in the monitoring position when the target reappeared. After repeating the experiment 100 times, the task success rate of the control group was 0%, while the task success rate of the experimental group was 100%. Second, to verify the effectiveness of the adaptive safety mechanism, the experiment simulated communication heartbeat interruptions of different durations during the execution of a stable escort strategy. The experimental group was set to two states: State A (system health index) (Set to 1.0) and state B ( (Set to 0.5), the dynamic timeout threshold for state A is calculated according to the formula. The time is 3.0 seconds, calculated by state B. The duration was 1.9 seconds. During the experiment, a 2.5-second communication interruption was applied to both experimental groups in both states. Table 1 records a set of data during the experiment.

[0036] Table 1: Safety mechanism response data of the experimental group under different health conditions.

[0037] Experimental Groups System Health Index (SHI) Dynamic timeout threshold (Second) Communication interruption duration (seconds) System Response Experimental group A 1.0 3.0 2.5 If the interruption duration is less than the threshold, the task will continue to execute. Experimental group B 0.5 1.9 2.5 Interruption duration > threshold, switch to security policy

[0038] Referring to Table 1, for the test group in state A, since the 2.5-second interruption duration is less than its 3.0-second timeout threshold, the system determines it to be a tolerable network fluctuation, and the UAV continues to execute the stable escort flight strategy; for the test group in state B, since the 2.5-second interruption duration is greater than its adjusted 1.9-second timeout threshold, the system determines it to be a communication anomaly, and the UAV is forced to switch to the high-level hovering observation safety strategy.

[0039] Example 3: This example combines Figures 1 to 3 This section describes a flight control method for unmanned aerial vehicles (UAVs) based on target recognition, such as... Figure 1 As shown, the UAV's visual sensor is responsible for capturing real-time high-definition video streams and inputting them into the target recognition and localization module. This module processes the video stream to output the target's coordinates. The output target coordinates are directly supplied to the flight control decision module and also input into the target state estimation and trajectory prediction module. This module estimates the target's speed and direction and predicts its trajectory, providing the predicted state information to the flight control decision module. Meanwhile, the environmental perception and dynamic obstacle avoidance module is responsible for identifying non-target obstacles and planning safe paths. The obstacle avoidance constraints it generates are also incorporated into the flight control decision module. After integrating multi-source information such as target coordinates, predicted state, and obstacle avoidance constraints, the flight control decision module finally generates control commands and issues them to the UAV's power system. This system then executes specific flight attitude and flight path adjustments. The execution results of the UAV's power system, through closed-loop feedback, affect the subsequent inputs of the UAV's visual sensor.

[0040] like Figure 2 As shown in the figure, the horizontal axis represents time in seconds, where 0 is the moment the target enters the tunnel. The vertical axis represents the state. The target vehicle state curve, marked with a solid square line, shows that the target is invisible from 0 to 10 seconds due to entering the tunnel. The UAV strategy curve, marked with a dashed triangle, shows that the UAV switched to a predictive flight strategy 5 seconds before the target disappeared, based on instructions issued by the ground station. After the target reappeared at 10 seconds, the UAV switched back to the escort strategy. The whole process intuitively demonstrates how this method ensures mission continuity during foreseeable signal interruptions by switching strategies in advance.

[0041] like Figure 3As shown, the architecture consists of two parts: a ground control station and a drone. The ground control station acts as the decision-making end. Its internal target recognition and environmental perception module is responsible for perceiving the complex dynamic environment and transmitting the perceived information to the strategy decision-making module. This module performs advanced strategy judgment and generation and sends the strategy instructions through the communication module via the wireless communication link. The drone acts as the execution end. Its internal communication and health monitoring module is responsible for receiving instructions and monitoring its own energy consumption and status data, which are then fed back to the ground control station. The strategy instructions received by the drone are parsed and state switched by its core flight strategy state machine, and finally, the onboard flight control system generates the underlying control signals.

[0042] Example 4: This example describes the system health index mentioned in the foregoing specific embodiments. The determination process provides a specific and reproducible calibration and calculation method; to ensure the dynamic timeout threshold upon which the adaptive security mechanism relies. To reflect the actual operational status of a drone, an input parameter that can quantitatively characterize the drone's long-term health status is needed, namely, the system health index. The determination of this index involves two stages: offline calibration and online calculation. The offline calibration stage involves establishing a baseline performance model for a brand-new or recently maintained UAV. This process is performed in a windless indoor environment. Under standard loads, the UAV sequentially executes each flight strategy defined in its flight strategy state machine, such as the stable escort strategy and the predictive fly-through strategy. The duration of each strategy is no less than 120 seconds. During this period, the flight control system records the average power consumption data and sensor (e.g., inertial measurement unit) noise data for each strategy at a sampling rate of 50Hz. The recorded baseline average power consumption for each strategy is as follows: and The recorded standard deviation of the noise from the reference inertial measurement unit accelerometer is: These values ​​are stored in the UAV's non-volatile memory as its performance benchmark. To ensure the reproducibility of the aforementioned offline calibration process, it is conducted in a standardized test environment with controlled environmental parameters: wind speed less than 0.2 m / s and ambient temperature of 25°C ± 2°C. During calibration, the UAV uses a standard payload containing all mission sensors and a fully charged battery of a specified model. During the 120-second period of executing each flight strategy and continuously recording data, the first 10 seconds of startup data and the last 10 seconds of shutdown data are discarded, and only the middle 100 seconds of steady-state operation data are used to calculate the benchmark average power consumption. Standard deviation of noise from the reference inertial measurement unit accelerometer Furthermore, the complete calibration process for each flight strategy is repeated five times, and the final benchmark value stored in the UAV's non-volatile memory is the arithmetic mean of the five independent test results.

[0043] The online calculation phase is executed periodically (e.g., every 60 seconds) in the flight control computer while the UAV is performing a mission. The flight control system first obtains the actual average power consumption of the UAV when executing the current flight strategy in the past calculation cycle. And the standard deviation of the noise of the inertial measurement unit accelerometer during the same time period. Subsequently, the system reads the baseline average power consumption corresponding to the current flight strategy from memory. And calculate a power consumption attenuation factor. The calculation method is as follows This value approaches 1 when the UAV's power system is in good condition, and decreases as the battery ages or the motor efficiency declines; simultaneously, the system calculates a sensor state factor. Its calculation method is as follows This value approaches 1 when the sensor is stable, and decreases as sensor performance deteriorates or body vibration intensifies; ultimately, the system health index... The calculation method is as follows: The above multiple factors are weighted and fused together. In the formula, The final calculated system health index is a dimensionless value. This is the power consumption attenuation factor; For sensor state factors; and These are the weighting coefficients for power consumption and sensor status, respectively, all non-negative and summing to 1. Based on the impact of the power system on flight safety, they can be... Set to 0.7. By setting it to 0.3, the drone can continuously generate a quantified system health index based on its initial performance baseline and current operational data. This value, which serves as a direct input, is used for the dynamic timeout threshold. The calculation provides an operational basis for the operation of the entire adaptive security protection mechanism.

[0044] Example 5: During the execution of the UAV flight control method, in addition to monitoring the communication heartbeat with the ground control station, the flight control system also calculates the system health index. The internal data undergoes periodic validity checks, a procedure that includes determining the calculated validity... Whether the value is within its preset valid range of 0 to 1, and to determine the value used for calculation. The system checks whether the variance of the raw data from the inertial measurement unit exceeds the maximum normal operating variance determined in the sensor's factory calibration. If any of the above verification conditions are met, the system will determine that the health status sensing function is malfunctioning. To avoid errors... Input causing dynamic timeout threshold If the value is set to an inappropriate value, the system will stop using the online calculator. Correspondingly, the flight control system will load a fixed backup system health index preset in the firmware. Its value is 0.7, and this spare value is used for The calculations enable the safety protection mechanism to continue operating at a fixed, conservative threshold, while simultaneously sending telemetry information related to anomalies in the state awareness function to the ground control station.

[0045] Example 6: This example provides an offline optimization parameter-finding procedure for the position controller parameters embedded in the stable escort strategy within the aforementioned flight strategy state machine. To ensure that the stable escort strategy can achieve a stable and rapid response to target position commands on UAVs with different aerodynamic characteristics, the proportional gain of its internal PID controller needs to be optimized. Integral gain With differential gain The optimization process aims to determine a set of gain parameters that optimizes the overall performance of the controller in terms of overshoot and settling time under step response. This process is conducted in a windless indoor environment equipped with a motion capture system. At the start of the procedure, the UAV is placed in the test range, a stable escaping strategy is activated, and a fixed target hovering point is set. Simultaneously, the controller's integral gain is adjusted. With differential gain The initial value is set to zero, and then the proportional gain is gradually increased from zero. The system continues until the motion capture system detects that the drone's position is exhibiting continuous, constant-amplitude oscillations around the target point. This is considered the system's stability boundary, and the proportional gain value at this point is recorded as the limiting gain. And record the oscillation period as the limiting period. Based on experimentally measured system stability boundary parameters and By applying the Ziegler-Nichols rule, a set of gain parameters with excellent dynamic performance can be obtained, and their calculations are as follows: , and Calculated The parameters are written into the UAV's configuration file as the default control parameters for the UAV to execute the stable escort flight strategy, thus providing a reproducible performance optimization basis for the reliable deployment of this flight control method on a specific UAV platform.

[0046] To further verify the method of the present invention, the following comparative examples are set up to demonstrate the technical advantages of the general technical method mentioned in the background art, which concentrates decision-making and perception at the ground station, in dealing with communication interruptions.

[0047] Comparative Example 1: This comparative example aims to simulate the conventional technical path in the background section, where the ground control station continuously locates the target and sends specific flight commands to the UAV in real time, and compares it with the present invention in the embodiment under the exact same test conditions. The test uses the same hardware-in-the-loop simulation platform as Example 2, including UAV flight control hardware running the control algorithm, a computer simulating UAV dynamics and environment, and a wireless channel simulator. The test conditions, UAV model, target vehicle model, and tunnel scene parameters (causing a fixed 10-second interruption of communication and visual signals) are all kept character-level consistent with the settings of Example 2. The UAV flight control software in this comparative example adopts a conventional real-time command receiving and execution architecture without a flight strategy state machine. Under this architecture, the ground control station sends commands containing the three-dimensional coordinates of the specific target to the UAV at a frequency of 5Hz, and the UAV, as a pure execution end, completes the flight actions.

[0048] In the initial tracking phase before the target vehicle enters the tunnel, conventional methods can stably control the UAV to accompany the target. However, when the test scenario simulates the target vehicle entering the tunnel, the ground control station loses visual and positional information about the target and is unable to generate effective real-time coordinate commands. Simultaneously, the wireless communication link between the UAV and the ground station is interrupted due to physical obstruction. After the UAV does not receive any new effective commands for 1.5 seconds (the safety threshold of the general flight control system), it automatically triggers its preset safety action, which is to hover at the last receiving position before the tunnel entrance. During this period, the ground control station logs record continuous communication link interruptions and target loss alarms, requiring manual intervention from the operator. After the 10-second interruption period ends, the target vehicle appears at another tunnel exit, while the UAV is still hovering at the tunnel entrance. The spatial distance between the two is now more than 250 meters, and the UAV cannot recapture the target using its own sensors. The operator attempts to manually control the UAV to fly to the exit position based on map information, but the mission ultimately fails because the target vehicle has merged into the traffic flow. To verify the reliability of this result, the same experiment was repeated 50 times, and the results are shown in Table 2 below.

[0049] Table 2: Statistics on the success rate of conventional real-time command control methods in signal interruption scenarios.

[0050] Experimental Groups Control methods Number of trials Number of successful tasks Task success rate Analysis of reasons for failure Comparative Example 1 Conventional real-time command control 50 1 2% The control architecture is designed to be strictly bound to the communication link. Once the command is lost, it can only hover in place and cannot cope with foreseeable task interruptions.

[0051] In 50 repeated trials, only one was successful. Log analysis revealed that this success was due to a random signal instantaneous recovery event with a probability of less than 1% occurring in the channel simulator. This caused the UAV to receive a delayed command after hovering for 2 seconds, which just happened to drift a certain distance towards the tunnel exit, allowing it to barely re-acquire the target when it appeared. This situation further demonstrates the inherent defects of conventional technical approaches in terms of mission resilience.

[0052] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A flight control method for unmanned aerial vehicles (UAVs) based on target recognition, characterized in that, Includes the following steps: Step a: The ground control station generates a policy switching instruction based on the target identification information and environmental context information. The policy switching instruction includes a policy identifier, context parameters required to execute the policy, and effective time parameters. Step b: The ground control station sends the strategy switching command to the UAV via the communication link; Step c: The UAV receives a strategy switching instruction and, based on the instruction, switches its locally running flight strategy state machine to the flight strategy specified by the strategy identifier. Step d: Under the guidance of the flight strategy, the UAV autonomously and continuously generates flight control signals to adjust its own flight status. Step e: During the execution of a flight strategy, the UAV monitors the energy consumption data associated with that strategy, associates the energy consumption data with the strategy identifier, and sends it to the ground control station. The ground control station references the energy consumption data when generating subsequent strategy switching commands. Step f: The UAV monitors the communication heartbeat with the ground control station and calculates a timeout threshold in each risk assessment cycle based on a system health index that represents its own health status and a preset threshold function. When the interruption duration of the communication heartbeat meets the timeout threshold, or the current time exceeds the valid time parameter, the flight strategy state machine is forced to switch to the preset safety strategy. Among them, the system health index A standardized value calculated by the drone based on energy consumption data; the timeout threshold calculated by the threshold function. With system health index They exhibit a monotonically decreasing relationship; Timeout threshold The following function is used to calculate: in, This is the base timeout threshold corresponding to the drone under a preset baseline health condition; This is the minimum timeout threshold corresponding to the drone under the preset minimum health condition; This is a system health index, and its value ranges from 0 to 1. Before executing step a to generate the strategy switching command, the ground control station also includes: determining, based on environmental context information, whether the target is about to enter the communication signal obstruction zone; when the determination result is yes, generating and sending a strategy switching command with the predicted crossing strategy in advance. The flight strategy state machine includes at least: a stable escort strategy, under which the UAV maintains synchronous flight with the target based on the relative position offset defined by the context parameters; and a predictive crossing strategy, under which the UAV autonomously plans and executes a route to a predetermined location near the target exit point based on the target entry point, target exit point and estimated travel time defined by the context parameters.

2. The UAV flight control method based on target recognition according to claim 1, characterized in that, The strategy switching command is structured data, and the context parameters include at least: the target relative position offset required when executing the stable escort strategy, or the target entry point coordinates, target exit point coordinates, and estimated passage time required when executing the predictive crossing strategy.

3. The UAV flight control method based on target recognition according to claim 1, characterized in that, The safety strategy in step f is either a high-level hovering observation strategy or a return-to-home strategy.

4. The UAV flight control method based on target recognition according to claim 1, characterized in that, When performing step e, the UAV will also monitor the sensor noise level or actuator response delay associated with the flight strategy and send the monitoring results to the ground control station.

5. The UAV flight control method based on target recognition according to claim 1, characterized in that, The energy consumption data includes at least the average current and average voltage of the UAV during different flight strategies; in step e, the UAV associates the average current and average voltage with the corresponding strategy identifier to form a strategy energy consumption mapping table, and sends the strategy energy consumption mapping table to the ground control station.

6. The UAV flight control method based on target recognition according to claim 1, characterized in that, The flight strategy state machine is a finite state machine, whose state set consists of the stable escort strategy, the predictive crossing strategy, and the high-altitude hovering observation strategy. At any given time, the state machine is in only one of the strategy states.

Images