Fire extinguishing unmanned aerial vehicle intelligent obstacle avoidance method and system

Abstract

This invention provides an intelligent obstacle avoidance method and system for firefighting drones. The method includes: acquiring perception information of the environment surrounding the aircraft and quantifying the degree of identification of potential threats in the perception information; adjusting the safety margin range around the aircraft based on the quantification result of the potential threat identification degree and the current operating state of the aircraft; when a potential threat enters the safety margin range, selecting a rapid avoidance strategy from a preset set of avoidance actions based on the safety margin range; maintaining the execution of the rapid avoidance strategy for a preset time period during execution, and making local adjustments to the strategy using new environmental information; and deactivating the strategy under specific conditions. This application can effectively address obstacle avoidance challenges in complex environments such as fires.

Description

Technical Field

[0001] This application relates to the field of drone obstacle avoidance technology, and more specifically, to an intelligent obstacle avoidance method and system for fire-fighting drones. Background Technology

[0002] Existing firefighting drones face significant challenges in their intelligent obstacle avoidance systems when operating in hazardous environments such as fires. Particularly in particle clouds formed by large amounts of burning particles and high-temperature gases, the data quality of traditional sensing systems such as lidar and infrared thermal imaging deteriorates severely, leading to high uncertainty in obstacle identification, localization, and threat assessment. This uncertainty makes it difficult for the drone's flight controller to decisively select the optimal obstacle avoidance strategy, resulting in frequent reassessments of the decision-making module and delays in path planning calculations. Ultimately, this leads to delayed obstacle avoidance response, increasing the risk of mission failure and drone damage. Summary of the Invention

[0003] This application discloses an intelligent obstacle avoidance method and system for fire-fighting drones, which aims to solve the problems of delayed obstacle avoidance response and decision-making errors caused by the decline in the quality of environmental perception information and decision uncertainty in traditional obstacle avoidance systems in complex fire scenarios.

[0004] The technical solution of this application is as follows: In a first aspect, this application discloses an intelligent obstacle avoidance method for fire-fighting drones, comprising: acquiring perception information of the environment surrounding the aircraft and quantifying the degree of identification of potential threats in the perception information; adjusting the safety margin range around the aircraft based on the quantification result of the degree of identification of potential threats and the current operating state of the aircraft; when a potential threat enters the safety margin range, selecting a rapid avoidance strategy from a preset set of avoidance actions based on the safety margin range; maintaining the execution of the rapid avoidance strategy for a preset time period during the execution of the strategy and making local adjustments to the strategy using new environmental information; and deactivating the maintained execution state of the strategy under specific conditions.

[0005] Secondly, this application also discloses an intelligent obstacle avoidance system for fire-fighting drones. The system includes: a perception information acquisition module for acquiring perception information about the environment surrounding the aircraft and quantifying the degree of identification of potential threats in the perception information; a safety margin adjustment module for adjusting the safety margin range around the aircraft based on the quantification result of the potential threat identification degree and the current operating state of the aircraft; a strategy selection module for selecting a rapid avoidance strategy from a preset set of avoidance actions based on the safety margin range when a potential threat enters the safety margin range; a strategy maintenance and adjustment module for maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information; and a strategy deactivation module for deactivating the strategy under specific conditions.

[0006] Compared with the prior art, this application has at least the following beneficial effects: This application can effectively address obstacle avoidance challenges in complex environments such as fires. By quantifying potential threats, dynamically adjusting safety margins, selecting rapid avoidance strategies, and making local adjustments during strategy execution, it significantly improves the efficiency and safety of obstacle avoidance decisions for UAVs under extreme conditions, and solves the problems of decision lag and errors in traditional obstacle avoidance systems when information uncertainty is high. Attached Figure Description

[0007] Figure 1 This is a flowchart illustrating an intelligent obstacle avoidance method for firefighting drones provided in this application.

[0008] Figure 2 This is a structural diagram of an intelligent obstacle avoidance system for firefighting drones provided in this application. Detailed implementation method. The technical solutions in this application will now be clearly and completely described in conjunction with the accompanying drawings.

[0009] This application proposes an intelligent obstacle avoidance method for firefighting drones, such as... Figure 1 As shown, it includes the following steps: Acquire sensory information about the environment surrounding the aircraft and quantify the degree of identification of potential threats in the sensory information; Based on the quantification of the degree of identification of potential threats and the current operational status of the aircraft, adjust the safety margin range around the aircraft; When a potential threat enters the safety margin range, a rapid avoidance strategy is selected from the preset set of avoidance actions based on the safety margin range. During the execution of the rapid evasion strategy, the strategy is maintained within a preset time period, and local adjustments are made to the strategy using new environmental information; and Under certain conditions, the strategy's maintenance execution status is lifted.

[0010] This application aims to significantly improve the obstacle avoidance capabilities and decision-making efficiency of UAVs in complex fire environments by introducing mechanisms for quantifying the degree of potential threat identification, dynamically adjusting the safety margin range, and making local adjustments during strategy execution, thereby effectively addressing the challenges faced by traditional obstacle avoidance systems.

[0011] To better understand the intelligent obstacle avoidance method for firefighting drones proposed in this application, the key terms and implementation environment involved will be explained in detail below.

[0012] The aircraft typically refers to firefighting drones, which possess the ability to fly autonomously, perceive their environment, perform tasks, and avoid obstacles. These aircraft are equipped with various sensors, such as lidar, infrared thermal imagers, visual sensors, and inertial measurement units (IMUs), to acquire information about their surroundings and their own operational status.

[0013] Perception information refers to the data about the surrounding environment obtained by an aircraft through its various onboard sensors, including but not limited to the distance, shape, temperature, and motion status of obstacles, as well as environmental characteristics such as the density and distribution of smoke and particulate clouds.

[0014] Potential threats refer to any object or environmental factor that may pose a risk to the safe flight of an aircraft, such as building debris, falling objects, high-temperature areas, dense smoke, particulate clouds, other aircraft, etc.

[0015] Quantifying the degree of recognition refers to the process of numerically evaluating the clarity, confidence level, and completeness of potential threats in perceived information. For example, when lidar data is interfered with by particulate clouds, the sparsity and noise level of the obstacle point cloud can be used as quantitative indicators of the degree of recognition.

[0016] A safety margin range refers to a virtual protection zone established around an aircraft. When a potential threat enters this range, the aircraft needs to take evasive action. The size of this range can be dynamically adjusted based on the level of threat identification and the aircraft's operational status.

[0017] A rapid avoidance strategy refers to an emergency obstacle avoidance maneuver that an aircraft selects and executes from a pre-set set of actions when it detects a potential threat, such as emergency climb, descent, left turn, right turn, hovering, or a combination of maneuvers.

[0018] A preset time period refers to a set duration for which the strategy is maintained during the execution of a rapid avoidance strategy to ensure its effectiveness and stability. During this period, the main execution of the strategy is not easily interrupted, but partial adjustments are permitted.

[0019] Local adjustments refer to minor modifications made to parameters such as trajectory, attitude, or speed of the strategy during the execution of a rapid evasion strategy, based on new environmental information, to adapt to dynamic changes in the environment without deviating from the main strategy's evasion objective.

[0020] Specific conditions refer to the triggering conditions for disabling the strategy to maintain execution, such as the potential threat has been successfully avoided, the aircraft has left the danger zone, or a better strategy has been identified and needs to be switched.

[0021] The implementation environment of this application is usually a fire scene, which may contain complex and dynamically changing factors such as high temperature, dense smoke, particulate clouds, and structural collapse, which places extremely high demands on the perception and obstacle avoidance capabilities of UAVs.

[0022] The core of the intelligent obstacle avoidance method for firefighting drones proposed in this application lies in ensuring that drones can perform their tasks efficiently and safely in complex and ever-changing fire environments through a series of refined steps.

[0023] First, the method involves acquiring perception information about the aircraft's surrounding environment and quantifying the degree of identification of potential threats within that information. In practice, the aircraft can utilize various onboard sensors to acquire perception information. For example, it can use lidar to acquire point cloud data of the surrounding environment, infrared thermal imagers to acquire temperature distribution images of the environment, and visible light cameras to acquire visual images of the environment. This sensor data can be transmitted in real time to the aircraft's onboard processing unit. Quantifying the degree of identification of potential threats can be achieved in several ways. For example, for lidar data, the density, sparsity, noise level, and confidence of point cloud clustering can be calculated to quantify the degree of obstacle identification. When point cloud data is interfered with by smoke or particulate clouds, the point cloud density decreases, the noise level increases, and the clustering confidence decreases; these changes reflect a reduction in the degree of identification. Similarly, for infrared thermal imaging data, the contrast between the target and background, edge sharpness, and the uniformity of the temperature gradient in the image can be analyzed to quantify the degree of identification. When a target is obscured by high-temperature smoke or particulate clouds, the contrast will decrease, the edges will become blurred, and the temperature gradient will become uneven. These can all be used as a basis for quantifying the degree of recognition.

[0024] Secondly, based on the quantification of the potential threat identification level and the aircraft's current operating status, the safety margin range around the aircraft is adjusted. The aircraft's current operating status includes, but is not limited to, flight speed, altitude, attitude, remaining battery power, and mission priority. For example, when the potential threat identification level is low (e.g., the obstacle outline is blurred in dense smoke), to ensure safety, the safety margin range can be increased, allowing the aircraft to initiate evasive maneuvers at a greater distance from the obstacle. Conversely, when the identification level is high, the safety margin range can be appropriately reduced to improve flight efficiency. Simultaneously, the aircraft's current operating status also influences the adjustment of the safety margin range. For example, when the aircraft is flying at high speed, a larger safety margin range is needed to allow sufficient reaction time; when the aircraft's battery power is low or the mission priority is high, it may be necessary to appropriately reduce the safety margin range to complete the mission as quickly as possible, while ensuring safety. The adjustment of the safety margin range can be achieved through a pre-defined lookup table, a fuzzy logic controller, or a machine learning-based model.

[0025] Secondly, when a potential threat enters the safety margin range, a rapid avoidance strategy is selected from a pre-defined set of avoidance actions based on the safety margin. This set includes various predefined obstacle avoidance actions, such as emergency climb, descent, left turn, right turn, hovering, deceleration, or combinations thereof. Selecting a rapid avoidance strategy requires comprehensive consideration of the size of the safety margin, the type, location, and speed of the potential threat, as well as the aircraft's current operational state. For example, if the potential threat is a rapidly approaching falling object and the safety margin is large, the aircraft might choose an emergency climb or lateral maneuver. If the potential threat is a stationary obstacle and the safety margin is small, the aircraft might choose to decelerate and make minor attitude adjustments. Strategy selection can be achieved through decision trees, state machines, or reinforcement learning-based policy selectors.

[0026] Next, during the execution of the rapid avoidance strategy, the strategy is maintained for a preset time period, and local adjustments are made to the strategy using new environmental information. The preset time period is set to ensure that the selected strategy can be executed stably for a certain period, avoiding frequent strategy switching due to minor environmental fluctuations, which could lead to flight instability. During this preset time period, the aircraft continuously receives new environmental information. This new environmental information may include updates to sensor data, minor changes in the location of potential threats, changes in wind speed and direction, etc. Based on this new environmental information, the aircraft can make local adjustments to the rapidly avoidance strategy being executed. For example, if the strategy is to maneuver to the left, but new environmental information indicates the appearance of a new, smaller obstacle on the left, the aircraft can fine-tune its flight trajectory or attitude to avoid this new obstacle without changing its overall leftward maneuver direction. This local adjustment can be a small-scale attitude correction, speed adjustment, or trajectory fine-tuning, aimed at improving the adaptability and accuracy of the strategy while maintaining its overall stability.

[0027] Finally, under specific conditions, the strategy's sustained execution state is deactivated. These conditions can include various scenarios. For example, when the aircraft successfully avoids a potential threat and has left the danger zone, the current strategy's sustained execution state can be deactivated, and the aircraft can return to normal mission flight mode. Another example is when, during the execution of the current strategy, the aircraft detects a more serious threat or a better avoidance path, the current strategy can also be deactivated, and a new strategy can be switched to. Furthermore, when a preset time period ends and the environmental assessment determines that the current strategy is no longer the optimal choice, the strategy's sustained execution state can also be deactivated. After deactivating the strategy, the aircraft can reassess the environment and select a new flight strategy or resume normal flight.

[0028] The intelligent obstacle avoidance method for firefighting drones proposed in this application works by constructing a dynamic and adaptive obstacle avoidance decision-making loop. This method first acquires environmental perception information through multi-source sensors and quantifies the degree of potential threat identification. This solves the problem of identification uncertainty caused by the deterioration of perception data quality in complex fire environments using traditional methods. For example, in particulate cloud areas, lidar and infrared thermal imaging data may be blurry; this method, by quantifying the degree of identification, can objectively assess the reliability of the current perception information.

[0029] Subsequently, based on the quantification results and the aircraft's current operating status, the safety margin range is dynamically adjusted. This step is one of the core innovations of this method, enabling the obstacle avoidance strategy to be no longer static but adaptively adjusted according to environmental uncertainties and the aircraft's real-time status. For example, when the recognition level is low, the safety margin range is expanded to allow the aircraft more reaction time, thereby compensating for the risks brought about by perception uncertainty; when the aircraft is flying at high speed, the safety margin range will also increase accordingly to ensure sufficient braking and avoidance space. This dynamic adjustment mechanism effectively avoids the misjudgment or insufficient response caused by fixed safety thresholds in traditional methods.

[0030] When a potential threat enters the dynamically adjusted safety margin range, the system selects a rapid evasion strategy from a pre-set set of evasion actions. This selection process comprehensively considers the nature of the threat, the aircraft's status, and the safety margin range to ensure the timeliness and effectiveness of the selected strategy.

[0031] During the execution of the rapid evasion strategy, this method introduces a mechanism to maintain strategy execution within a preset time period and make local adjustments using new environmental information. This mechanism solves the problems of high-frequency strategy switching and computational delays caused by decision uncertainty in traditional methods. By setting a preset time period, the strategy can be executed stably, avoiding frequent path planning interruptions and restarts. At the same time, the local adjustment mechanism allows the aircraft to make fine-tuning corrections based on minor changes in the environment without interrupting the main strategy, such as fine-tuning attitude or trajectory, thereby improving the adaptability and accuracy of the strategy and ensuring the continuity and stability of evasion actions.

[0032] Finally, under specific conditions, the maintenance execution status of the strategy is deactivated, allowing the aircraft to flexibly exit the current evasion strategy and choose subsequent actions based on the new environmental assessment, whether to resume normal flight or switch to a new evasion strategy.

[0033] Through the close coordination of the above-mentioned steps, the method of this application can effectively address the problems of perception uncertainty, decision ambiguity, and response lag caused by extreme environments such as particulate clouds in fire scenes. By dynamically adjusting the safety margin and combining strategy maintenance with local adjustments, it significantly improves the intelligent obstacle avoidance capability and mission execution efficiency of firefighting drones in complex environments, thereby reducing the risk of mission failure and drone damage.

[0034] The intelligent obstacle avoidance method for firefighting drones proposed in this application demonstrates significant advantages and innovation compared to existing technologies in dealing with obstacle avoidance challenges in complex fire environments, especially under extreme conditions such as particulate clouds.

[0035] The core innovation of this application lies in the introduction of quantification of the degree of potential threat identification and the dynamic adjustment of the safety margin range based on this. For example, when particulate clouds cause sparse lidar point clouds and high noise, the system can quantify this reduction in identification level and expand the safety margin range accordingly. This means that the drone will initiate evasive maneuvers when it is further away from the potential threat, thereby gaining valuable time for decision-making and execution, and effectively compensating for the risks brought about by perception uncertainty.

[0036] Furthermore, this application introduces a mechanism to maintain strategy execution within a preset time period and make local adjustments using new environmental information during the execution of a rapid obstacle avoidance strategy. This mechanism is key to solving the problems of decision uncertainty and computational latency in traditional obstacle avoidance systems. In existing technologies, due to low confidence in obstacle threat assessment, the decision module frequently re-evaluates multiple potential actions, leading to repeated interruptions and restarts of path planning calculations, resulting in severe computational latency and obstacle avoidance lag. This application, by setting a preset time period to maintain strategy execution, effectively avoids such high-frequency strategy switching and computational interruptions, ensuring the continuity and stability of avoidance actions. Simultaneously, the local adjustment mechanism allows the aircraft to make fine-tuning corrections based on minor environmental changes without interrupting the main strategy, such as fine-tuning attitude or trajectory, thereby improving the adaptability and accuracy of the strategy and ensuring the continuity and stability of avoidance actions. This combination of strategy maintenance and local adjustment significantly improves the decision-making efficiency and obstacle avoidance response speed of UAVs in complex environments.

[0037] In summary, this application effectively addresses the challenges faced by existing firefighting drones in extreme fire environments due to perception uncertainty, decision-making ambiguity, and response lag, through innovative mechanisms such as quantifying the degree of potential threat identification, dynamically adjusting the safety margin range, and making local adjustments during strategy execution. The method described in this application enables drones to achieve more intelligent, stable, and efficient obstacle avoidance in complex and ever-changing environments, thereby significantly improving mission success rate and flight safety.

[0038] However, in actual firefighting scenarios, drones performing rapid evasive maneuvers may encounter sudden local environmental anomalies, such as strong airflow disturbances, unexpected minor collisions, or particulate impacts. These anomalies can cause significant physical impacts on the aircraft, leading to unexpected deviations in flight attitude and trajectory. Failure to respond to these physical impacts promptly and accurately may affect the effectiveness of rapid evasive strategies and even endanger the safety of the aircraft.

[0039] In this regard, this application further proposes that the steps of maintaining the execution of the strategy within a preset time period during the execution of the rapid avoidance strategy, and making local adjustments to the strategy using new environmental information, include: During the execution of the rapid evasion strategy, monitor the inertial measurement unit data of the aircraft itself; Monitor the actual response data of the motors on the aircraft itself; The deviation between the inertial measurement unit data and the expected inertial measurement unit data is compared to obtain the first deviation information; By comparing the deviation between the actual motor response data and the expected motor response data, a second deviation information is obtained; Based on the first deviation information and the second deviation information, determine the intensity of the physical impact of local environmental anomalies on the aircraft; When the physical impact intensity exceeds a preset threshold, select an attitude adjustment action from a preset local attitude correction library; Execute attitude adjustment actions and superimpose the attitude adjustment action instructions with the instructions for the rapid avoidance strategy; and After the attitude adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

[0040] Specifically, monitoring the inertial measurement unit (IMU) data of the aircraft body refers to acquiring real-time data such as acceleration, angular velocity, and attitude angles output by the aircraft's internal IMU. This data accurately reflects the aircraft's motion state and attitude changes in three-dimensional space. Simultaneously, monitoring the actual response data of the aircraft's motors refers to acquiring real-time operating parameters such as speed, current, and torque of each motor (e.g., the individual rotor motors of a multi-rotor UAV). This data directly reflects the actual output of the aircraft's power system.

[0041] The expected inertial measurement unit (IMU) data can be understood as the IMU data that the aircraft should have when executing the current rapid avoidance strategy under ideal conditions, i.e., without external physical impact interference. Similarly, the expected motor response data refers to the response data that the aircraft's motor system should have when executing the current rapid avoidance strategy under ideal conditions. By comparing the real-time monitored IMU data with the expected IMU data, the first deviation information can be obtained, which quantifies the difference between the aircraft's actual motion state and its expected motion state. Likewise, by comparing the real-time monitored actual motor response data with the expected motor response data, the second deviation information can be obtained, which reflects the difference between the actual motor output and the strategy command output.

[0042] In practical applications, the intensity of the physical impact on the aircraft caused by local environmental anomalies can be determined based on the first and second deviation information. For example, when the first deviation information (such as a sudden change in attitude angle or acceleration) and the second deviation information (such as abnormal fluctuations in motor speed) occur simultaneously and exceed a certain range, it indicates that the aircraft may have been subjected to an external physical impact. The intensity of the physical impact can be quantified using a preset algorithm, for example, by combining the magnitude and duration of the deviation for a comprehensive evaluation. When the intensity of the physical impact exceeds a preset threshold, it indicates that the aircraft is experiencing significant external disturbance and requires immediate attitude correction. At this time, the system will select one or more attitude adjustment actions from a preset local attitude correction library. The local attitude correction library stores various fine-tuning attitude actions for different impact intensities and directions, aiming to quickly restore the stability of the aircraft.

[0043] Furthermore, the selected attitude adjustment maneuver command will be executed and superimposed on the command of the currently executing rapid avoidance strategy. This superposition mechanism ensures that while making local attitude corrections, the aircraft can still roughly follow its main avoidance path, avoiding complete deviation from the avoidance target due to attitude corrections. After the attitude adjustment maneuver is completed, the aircraft will be guided back to the expected trajectory of the rapid avoidance strategy to ensure the successful completion of the obstacle avoidance mission.

[0044] This application's solution, through real-time monitoring of the aircraft's inertial measurement unit data and actual motor response data, can accurately capture physical impacts caused by any local environmental anomalies that the aircraft may encounter during the execution of a rapid avoidance strategy. It is precisely this continuous monitoring of these key operating parameters that enables the system to promptly detect deviations between the aircraft's actual and expected states. By comprehensively analyzing the first and second deviation information, the intensity of the physical impact can be accurately determined, thereby avoiding overreaction to minor disturbances while ensuring a rapid response to severe impacts. When the intensity of the physical impact reaches a level requiring intervention, the system can intelligently select the most suitable attitude adjustment action from a pre-set local attitude correction library and, through command superposition, quickly restore the aircraft's flight stability without interrupting the main avoidance strategy. This mechanism ensures that the aircraft maintains its attitude stability and trajectory accuracy when performing highly dynamic avoidance tasks in complex and changing environments.

[0045] Through the above technical solution, this application can significantly improve the flight stability and anti-interference capability of firefighting drones during the execution of rapid evasion strategies. Especially in complex environments such as fire scenes, when faced with sudden strong airflow, particulate impacts, or minor collisions, this solution can sense and quantify these physical impacts in real time and quickly take local attitude adjustment actions to counteract them. This effectively avoids the risk of the aircraft deviating from the predetermined evasion trajectory or becoming out of control due to external interference, thereby ensuring the effective execution of the rapid evasion strategy and greatly improving the operational safety and mission success rate of drones in hazardous environments.

[0046] However, in actual firefighting scenarios, the environment is complex and ever-changing. In particular, there may be a large number of particulate clouds in the fire scene, whose density, distribution, movement speed, and local temperature characteristics change dynamically over time. If only general local adjustments are made, it may not be able to adequately cope with these dynamic and cumulative environmental factors, resulting in insufficient precision and foresight in local adjustments, thereby affecting the overall effectiveness of obstacle avoidance strategies and the safety of the aircraft.

[0047] In this regard, this application further proposes the following steps for maintaining the execution of the aforementioned rapid avoidance strategy within a preset time period and making local adjustments to the strategy using new environmental information: While executing a rapid evasion strategy and remaining under strategy lock-in, continuously receive environmental data; Analyze the density, distribution, movement speed, and gradual temperature variation trends of particulate clouds in environmental data; Calculate the cumulative risk index based on the gradual change trend; When the cumulative risk index exceeds a preset threshold, the magnitude and direction of the local adjustments are dynamically corrected. Predict the aircraft's expected position and relative distance to the potential threat at the end of the lock-on, under the revised local adjustments; and Based on the prediction results, further fine-tune the parameters for local adjustments.

[0048] Specifically, during the execution of a rapid evasion strategy and while under strategic lock-on, the aircraft continuously receives environmental data. This environmental data can originate from various sensors onboard the aircraft, such as lidar, visual sensors, infrared sensors, and gas sensors, used to monitor the fire environment around the aircraft in real time. Particulate clouds refer to the aggregates of tiny particles in the air formed by smoke, ash, and combustion products generated by the fire.

[0049] Furthermore, the received environmental data will be used to analyze the density, distribution, movement speed, and gradual trends in local temperature of the particulate cloud. Density reflects the concentration of the particulate cloud, distribution indicates its spatial occupancy, movement speed reveals its dynamic evolution, and gradual trends in local temperature may indicate the spread of fire or the direction of thermal currents. Gradual trends refer to the continuous or periodic changes of these parameters over a period of time, rather than instantaneous values.

[0050] Based on this, a cumulative risk index is calculated according to the gradual changing trend of the particulate cloud. This index aims to quantify the potential cumulative threat of particulate clouds to the safe flight of aircraft. For example, high-density, fast-moving particulate clouds with continuously rising temperatures will generate a higher cumulative risk index. The calculation of this index can be based on a preset weighting model or machine learning algorithm, comprehensively considering the impact of various parameters on aircraft safety.

[0051] When the cumulative risk index exceeds a preset threshold, it indicates that the cumulative risk posed by the current environment to the aircraft has reached a level requiring intervention. At this point, dynamic adjustments to the magnitude and direction of local modifications are necessary. This correction is not a simple attitude adjustment, but a predictive adjustment based on cumulative risk, aimed at more effectively mitigating potential threats.

[0052] Subsequently, with the corrected local adjustments, the expected position of the aircraft and its relative distance to potential threats at the end of the strategy lockout are predicted. This prediction can be simulated using the aircraft's dynamics model, environmental model, and the corrected local adjustment commands. The prediction results can be used to evaluate the effectiveness of the current correction scheme.

[0053] Finally, based on the prediction results, the parameters of the local adjustments are further fine-tuned. If the prediction shows that the aircraft may still be too close to a potential threat at the end of the lock-on, or that the obstacle avoidance efficiency is poor, the parameters of the local adjustments (such as the magnitude, duration, or direction of the adjustment) can be optimized more precisely to ensure that the aircraft can safely complete the avoidance maneuver.

[0054] This application's solution achieves dynamic correction of local adjustments to rapid evasion strategies by introducing continuous monitoring and analysis of the dynamic characteristics of particulate clouds in the fire environment and calculating a cumulative risk index based on this. Traditional local adjustments may only be reactive adjustments based on instantaneous environmental information, making it difficult to effectively address dynamic threats with cumulative effects, such as particulate clouds. This application, by analyzing the density, distribution, movement speed, and gradual temperature changes of particulate clouds, can gain a more comprehensive and in-depth understanding of the environmental evolution patterns, thereby calculating a cumulative risk index that reflects long-term risks. It is precisely this forward-looking risk assessment that allows the aircraft to predictively and dynamically correct the magnitude and direction of local adjustments before the risk accumulates to a critical point. Furthermore, by predicting the aircraft's expected position and relative distance to potential threats after correction, and making fine-tuning accordingly, the accuracy and effectiveness of local adjustments are further ensured. This ensures that during strategy lockout, the aircraft can continuously and safely execute evasion maneuvers, avoiding accidents caused by dynamic environmental changes.

[0055] Through the aforementioned technical solution, this application significantly improves the adaptability and safety of firefighting drones when executing rapid obstacle avoidance strategies in complex fire environments. By deeply analyzing the dynamic characteristics of particulate clouds and calculating the cumulative risk index, the aircraft can identify and quantify potential cumulative threats earlier and more accurately, thereby achieving predictive and dynamic correction of local adjustments. This correction mechanism enables the aircraft to perform more refined and forward-looking trajectory optimization when facing dynamic environments such as dense smoke and surging hot air currents, effectively avoiding obstacle avoidance failures caused by sudden environmental changes or cumulative effects. Compared to solutions that adjust based solely on instantaneous environmental information, the solution presented in this application provides more stable and reliable obstacle avoidance performance. Especially during the lock-on period of long-term execution of the avoidance strategy, it effectively reduces the risk of secondary contact between the aircraft and potential threats, greatly enhancing the survivability and mission success rate of the drone in extreme environments.

[0056] In some embodiments, the steps of maintaining the execution of the rapid evasion strategy for a preset time period and making local adjustments to the strategy using new environmental information during the execution of the rapid evasion strategy include: continuously receiving environmental data during the execution of the rapid evasion strategy and while the strategy is locked; analyzing changes in the local turbulence intensity, temperature gradient, or illumination conditions of the particle cloud in the environmental data; calculating an environmental disturbance index based on the changes in the local turbulence intensity, temperature gradient, or illumination conditions of the particle cloud; correcting the magnitude and direction of the local adjustment when the environmental disturbance index exceeds a preset threshold; predicting the expected flight stability of the aircraft at the end of the lock under the corrected local adjustment; and fine-tuning the parameters of the local adjustment based on the prediction results.

[0057] Specifically, during the execution of a rapid evasion strategy and while in a strategy lock-on state, continuous reception of environmental data refers to the aircraft acquiring various physical parameters of the fire scene environment in real time through its onboard sensors, such as lidar, thermal imagers, visible light cameras, and anemometers. The strategy lock-on period can be understood as the state in which the aircraft maintains its main flight commands for a preset time period to ensure the continuity and effectiveness of evasion maneuvers while executing the rapid evasion strategy.

[0058] Analyzing changes in local turbulence intensity, temperature gradient, or illumination conditions of particulate clouds in environmental data involves processing and parsing received environmental data to identify and quantify local airflow instabilities (turbulence intensity), spatial rate of temperature change (temperature gradient), and fluctuations in ambient illumination levels caused by particulate clouds (such as smoke, dust, and combustion products) in a fire scene. These factors can directly impact the aerodynamic performance, sensor perception capabilities, and flight control of aircraft.

[0059] In practical applications, the environmental disturbance index is calculated based on changes in the local turbulence intensity, temperature gradient, or illumination conditions of particulate clouds. This involves comprehensively considering the degree of change in these environmental factors and their potential impact on aircraft stability. A pre-defined algorithm model (such as weighted average, fuzzy logic, or machine learning models) is used to calculate a comprehensive numerical index that quantifies the degree of disturbance the current environment causes to the aircraft's flight state. Its purpose is to provide a quantitative basis for subsequent local adjustments.

[0060] When the environmental disturbance index exceeds the preset threshold, the magnitude and direction of the local adjustment are corrected. This means that once the environmental disturbance index reaches or exceeds the preset safety threshold, it indicates that the current environment has reached a level that requires active intervention on the stability of the aircraft. At this time, the flight control system will correct the local adjustment parameters of the fast avoidance strategy being executed according to the magnitude and direction of the disturbance index. For example, it may increase or decrease the magnitude of the attitude adjustment or change the direction of the adjustment to counteract the impact of the environmental disturbance.

[0061] Furthermore, predicting the aircraft's expected flight stability at the end of the lock-up under the corrected local adjustments refers to simulating and predicting the stability of parameters such as flight attitude, speed, and position of the aircraft at the end of the strategy lock-up (i.e., when the rapid evasion strategy is completed) using the aircraft's dynamics and environmental models after correcting the magnitude and direction of the local adjustments. The purpose is to evaluate the effectiveness of the correction scheme and provide a basis for further fine-tuning.

[0062] Therefore, fine-tuning the parameters of local adjustments based on the prediction results refers to making precise adjustments to the parameters of local adjustments based on the predicted flight stability. For example, if the prediction shows that stability is still insufficient, the magnitude or duration of the adjustment may be further fine-tuned to ensure that the aircraft can maintain optimal flight status after completing the evasive maneuver.

[0063] This application's solution, by continuously monitoring and meticulously analyzing key disturbance factors in the fire environment, such as local turbulence intensity, temperature gradient, or illumination conditions of particulate clouds, can more accurately assess the impact of the environment on aircraft flight stability. It is precisely the introduction of the environmental disturbance index, a quantitative indicator, that enables the flight control system to promptly identify and respond to environmental disturbances exceeding preset thresholds. By correcting the magnitude and direction of local adjustments and combining this with predictions of future flight stability, this solution can proactively offset or mitigate the impact of environmental disturbances on aircraft attitude and trajectory, thereby ensuring the stability and safety of the aircraft when executing rapid avoidance strategies. This local adjustment mechanism based on environmental disturbance prediction effectively compensates for the limitations of basic solutions in responding to complex environmental disturbances.

[0064] Through the aforementioned technical solutions, firefighting drones can more effectively cope with complex and ever-changing environmental disturbances in fire scenes when executing rapid obstacle avoidance strategies, significantly improving the flight stability of the aircraft. This solution, through refined analysis and prediction of specific environmental factors such as local turbulence intensity, temperature gradients, or lighting conditions, makes local adjustments no longer simple reactive corrections, but rather forward-looking and adaptable. This ensures that the aircraft maintains precise attitude control and trajectory tracking even in extreme environments, greatly improving obstacle avoidance success rates and mission reliability.

[0065] In some embodiments, the steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information during the execution of the rapid avoidance strategy include: continuously receiving environmental data during the execution of the rapid avoidance strategy and while the strategy is locked; identifying changes in the clarity of potential threat features in the environmental data; calculating an information utilization efficiency index based on the changes in clarity; correcting the response parameters of the local adjustment when the information utilization efficiency index is lower than a preset threshold; predicting the relative distance between the aircraft and the potential threat and the obstacle avoidance efficiency at the end of the lock under the corrected local adjustment; and fine-tuning the parameters of the local adjustment based on the prediction results.

[0066] Specifically, during the execution of a rapid evasion strategy and while under strategic lock-on, the aircraft continuously receives environmental data from various sensors (such as visual sensors, lidar, and infrared sensors). Identifying changes in the sharpness of potential threat features within the environmental data refers to performing image processing or signal analysis on the received environmental data to assess the sharpness or identifiability of features such as the boundaries, textures, and shapes of potential threats (e.g., obstacles, fire sources, smoke plumes). For example, sharpness can be quantified by calculating metrics such as edge gradients, contrast, signal-to-noise ratio, or information entropy of the image.

[0067] Furthermore, an information utilization efficiency index is calculated based on changes in clarity. This index can be understood as the effective contribution of potential threat information in the current environmental data to obstacle avoidance decisions. A higher clarity results in a higher information utilization efficiency index; conversely, a lower clarity results in a lower information utilization efficiency index. The calculation of this index can be based on a preset mapping function or machine learning model, converting the clarity index into a value between 0 and 1.

[0068] When the information utilization efficiency index falls below a preset threshold, it indicates that the guidance provided by current environmental data for obstacle avoidance decisions is weakened, and the response parameters of local adjustments need to be corrected. Response parameters may include, but are not limited to, the magnitude, speed, acceleration limit, steering angle limit, or time constant of the local adjustment. For example, when the information utilization efficiency index is low, the magnitude of the local adjustment can be appropriately reduced, or the adjustment speed can be decreased, to avoid excessive or erroneous actions based on uncertain information.

[0069] Based on this, the relative distance between the aircraft and the potential threat at the end of the lock-on, as well as the obstacle avoidance efficiency, are predicted under the corrected local adjustments. This prediction can be calculated through simulation using the aircraft dynamics model, the environmental model, and the corrected local adjustment parameters. Obstacle avoidance efficiency can be quantified as the probability of successfully avoiding the threat, the smoothness of the obstacle avoidance path, or energy consumption, etc.

[0070] Finally, based on the prediction results, fine-tune the parameters of the local adjustments. If the prediction results show that the obstacle avoidance effect is not good or there is a potential risk, the response parameters can be further fine-tuned. For example, while ensuring safety, the adjustment range can be slightly increased to improve the obstacle avoidance speed, or the adjustment direction can be adjusted to optimize the obstacle avoidance path.

[0071] This application addresses the limitation of traditional methods in terms of the effectiveness of local adjustments when environmental information quality fluctuates by introducing an assessment of the clarity of potential threat features in environmental data. Specifically, when an aircraft is executing a rapid avoidance strategy and is in a locked-down state, the clarity of potential threat features in the continuously received environmental data may change due to the complexity of the fire environment. By identifying and quantifying this clarity change, the system can calculate the actual utilization efficiency of the current environmental information for obstacle avoidance decision-making. Due to this real-time assessment of information quality, when the information utilization efficiency index falls below a preset threshold, the system can promptly identify potential decision risks and proactively correct the response parameters of local adjustments. This correction mechanism prevents the aircraft from blindly making large adjustments based on unclear or unreliable information, thus avoiding secondary risks caused by misjudgment. Subsequently, by predicting the corrected local adjustments, the system can assess their impact on the relative distance between the aircraft and potential threats and obstacle avoidance efficiency, and fine-tune the local adjustment parameters based on this prediction result. This closed-loop feedback adjustment mechanism ensures that even under poor environmental information conditions, the aircraft can perform local obstacle avoidance in a more robust and reliable manner, thereby improving the adaptability and safety of the overall obstacle avoidance strategy.

[0072] Through the above technical solution, this application can effectively address the problem of fluctuating environmental data quality in complex environments such as fire scenes, significantly improving the robustness and accuracy of local adjustments made by firefighting drones during the execution of rapid obstacle avoidance strategies. By assessing the clarity of potential threat characteristics in real time and calculating the information utilization efficiency index, the aircraft can avoid making inappropriate or excessive adjustments based on low-quality information, thereby reducing the risk of obstacle avoidance failure. Furthermore, by predicting the effect of corrected local adjustments and making fine-tuning, the obstacle avoidance path and efficiency are further optimized, enabling the drone to complete obstacle avoidance tasks more safely and effectively in complex and changing environments, improving the success rate of the mission and the survivability of the aircraft.

[0073] In some embodiments, during the aforementioned rapid avoidance strategy, the step of maintaining the execution of the strategy for a preset time period and making local adjustments to the strategy using new environmental information includes: During the execution of the rapid evasion strategy and while in a strategy lock-on state, it continuously receives vibration data, material stress data, and sensor performance data of the aircraft body; Continuously receive particle flow characteristic data in the fire environment, including particle size, density, charge, and thermal radiation characteristics; Analyze the correlation between vibration data, material stress data, sensor performance data and particle flow characteristic data of the aircraft body to identify whether there is abnormal vibration, material stress accumulation or sensor function degradation caused by specific physical characteristics of particle flow. The intensity of the effects of specific interactions is quantified based on the degree of abnormal vibrations, material stress accumulation, or sensor function degradation identified. When the intensity of the influence exceeds a preset threshold, select one or more local adjustment actions from the preset interaction cancellation action library to cancel or mitigate the specific interaction; Execute local adjustment actions and superimpose the instructions for these local adjustment actions with the instructions for the rapid evasion strategy; and After the local adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

[0074] Specifically, during strategy lockout, vibration data of the aircraft body can be continuously collected by accelerometers or vibration sensors installed on the aircraft structure to monitor the dynamic response of the aircraft structural components; material stress data can be acquired by strain gauges or fiber optic sensors to assess the stress and fatigue accumulation of key structural components; sensor performance data can include information such as the output stability, noise level, or calibration deviation of key sensors for aircraft navigation and obstacle avoidance, for example, by monitoring the effective detection range of lidar and the temperature drift of infrared sensors. This data is continuously received to comprehensively reflect the real-time health status of the aircraft body in complex environments.

[0075] In fire scene environments, particle flow characteristics data refer to the fluid formed by the movement of solid or liquid particles in an airflow generated by events such as combustion, explosion, or structural collapse at a fire site. Specifically, particle flow characteristics data can include particle size, such as particle size distribution obtained through laser diffraction or image analysis techniques; particle density, such as estimation using mass flow meters or optical density sensors; particle charge, such as detecting the amount of charge carried by particles using electrostatic sensors; and particle thermal radiation characteristics, such as measuring the surface temperature or radiation intensity of particles using infrared thermal imagers or radiometers. These characteristics are crucial for understanding the physical interactions between particle flows and aircraft.

[0076] In practical applications, analyzing the correlation between vibration data, material stress data, sensor performance data, and particle flow characteristic data of the aircraft body aims to identify whether there are anomalies caused by specific physical characteristics of the particle flow. For example, when the particle flow has pulses of a specific frequency, it may resonate with the aircraft structure, leading to abnormal vibrations; the impact of high-density particle flow may cause material stress accumulation; and a particle flow carrying a strong charge may interfere with the normal operation of the aircraft's electronic equipment or sensors. This correlation analysis can be achieved through time-frequency analysis, correlation analysis, or machine learning models to accurately identify potential harmful interactions.

[0077] Furthermore, quantifying the impact intensity of a specific interaction refers to calculating a quantitative index based on the degree of identified abnormal vibrations, material stress accumulation, or sensor performance degradation, combined with a pre-defined hazard assessment model. For example, the amplitude of abnormal vibrations, the rate of stress accumulation, or the percentage of sensor performance degradation can be mapped to a risk score from 0 to 100; this score represents the impact intensity. When the impact intensity exceeds a preset threshold, it indicates that the aircraft is experiencing significant negative impacts, requiring immediate action.

[0078] As a preferred implementation, the preset interaction cancellation action library can include a variety of local adjustment actions. For example, to address abnormal vibrations caused by resonance, the local attitude or power output frequency of the aircraft can be adjusted to avoid the resonance point; to address the impact of high-density particle flow, the flight trajectory can be fine-tuned to reduce the frontal area or the flight attitude can be changed to disperse the impact force; to address charge interference, the electrostatic discharge device on the aircraft can be activated or the operating parameters of the sensors can be adjusted. These actions are designed to directly cancel or mitigate the specific negative impacts of particle flow on the aircraft.

[0079] The superposition of local adjustment commands with rapid avoidance strategy commands can be understood as adding local adjustment commands as fine-tuning or correction commands to the aircraft's control system in real time, while keeping the main rapid avoidance strategy command unchanged. For example, if the rapid avoidance strategy command is to translate to the left, and the local adjustment command is to slightly raise the nose to reduce particle impact, the final aircraft control command will be to translate to the left while simultaneously slightly raising the nose. This superposition ensures that while dealing with local interactions, the aircraft does not deviate from the main obstacle avoidance objective.

[0080] Therefore, guiding the aircraft back to the expected trajectory of the rapid avoidance strategy after the local adjustment action is completed means that after the local adjustment action is completed and the expected effect is achieved, the aircraft control system will gradually eliminate the impact of the local adjustment action, so that the aircraft can smoothly return to the original flight path of the rapid avoidance strategy to continue to complete the obstacle avoidance mission.

[0081] This application's solution, by continuously monitoring the aircraft's vibration data, material stress data, and sensor performance data during the execution of a rapid obstacle avoidance strategy, and combining this with specific physical characteristics data of particle flow in the fire environment, enables in-depth analysis of the microscopic physical interactions between the aircraft and the complex environment. It is precisely this meticulous monitoring and correlation analysis that allows the system to identify potential hazards such as abnormal vibrations, material stress accumulation, or sensor malfunction degradation caused by particle flow. By quantifying the intensity of these interactions, the system can accurately determine when intervention is necessary. When the intensity reaches a preset threshold, the system can select targeted local adjustment actions from a pre-set library of countermeasures, such as adjusting flight attitude or power output, to directly counteract or mitigate the negative impact of particle flow. This local adjustment action, superimposed on the rapid obstacle avoidance strategy commands, ensures that the aircraft's anti-interference capability and stability in complex fire environments are effectively improved without interrupting the main obstacle avoidance mission. Finally, after the local adjustment action is completed, the aircraft is guided back to the expected trajectory, ensuring the continuity and effectiveness of the obstacle avoidance mission.

[0082] Through the above technical solution, this application effectively addresses the problem of poor obstacle avoidance or damage to aircraft performance caused by traditional obstacle avoidance methods in complex fire environments due to insufficient consideration of the specific physical interactions between particle flow and the aircraft body. This solution, by real-time monitoring of the aircraft's health data and particle flow characteristic data, and performing deep correlation analysis, enables the system to accurately identify and quantify anomalies caused by particle flow. Consequently, the aircraft can promptly take targeted local adjustments to effectively offset or mitigate the negative impacts of particle flow, significantly improving the flight stability, safety, and mission reliability of firefighting drones in fire environments filled with smoke, ash, and other particulate matter. This not only avoids aircraft instability, structural damage, or sensor failure caused by microscopic physical interactions but also ensures the smooth execution of rapid obstacle avoidance strategies, thereby improving overall obstacle avoidance efficiency and the success rate of firefighting missions.

[0083] In some embodiments, the step of continuously receiving vibration data, material stress data, and sensor performance data of the aircraft body during the execution of a rapid evasion strategy and while in a strategy lock-in state further includes: Time-frequency domain analysis was performed on the vibration data, material stress data, and sensor performance data of the aircraft body to extract its energy distribution and instantaneous phase changes within a specific frequency range; Real-time analysis of particle flow characteristic data can identify whether there are pulse flows of specific frequencies, high-density agglomerate flows, or strongly charged particle flows. The time-frequency domain analysis results of the aircraft body are correlated with the real-time analysis results in the particle flow characteristic data to identify whether there are characteristics such as resonant frequency, energy absorption or charge interference. Based on the identified resonant frequency, energy absorption, or charge interference characteristics, determine whether there is intermittent or nonlinear coupling between the particle flow and the aircraft body. When a coupling effect is detected, one or more local adjustment actions are selected from the preset coupling effect cancellation action library according to the type and strength of the coupling effect. The local adjustment actions include adjusting the local attitude or power output of the aircraft. Execute local adjustment actions and superimpose the instructions for these local adjustment actions with the instructions for the rapid evasion strategy; and After the local adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

[0084] Specifically, time-frequency domain analysis of vibration data, material stress data, and sensor performance data of the aircraft body refers to transforming the signal from the time domain to the time-frequency domain using signal processing methods such as Fourier transform and wavelet transform, so as to observe the changes of the signal in time and frequency simultaneously. This aims to reveal the transient and nonlinear characteristics of the signal, such as energy concentration or dispersion within a specific frequency range, and rapid instantaneous phase drift. These characteristics are often direct manifestations of complex interactions. Among them, the energy distribution and instantaneous phase change within a specific frequency range can be understood as the energy intensity and vibration mode exhibited by the aircraft structure or sensor response at certain frequencies at a specific point in time, with the purpose of identifying potential resonance or anomalous responses.

[0085] Meanwhile, real-time analysis of particle flow characteristic data refers to the continuous monitoring and rapid analysis of particle flows (including particle size, density, charge, and thermal radiation characteristics) in the fire environment using high-speed sensors and data processing units. The aim is to promptly identify specific types of particle flows with potential hazards, such as specific-frequency pulse flows that may induce resonance, high-density clump flows that may cause structural impact or blockage, and strongly charged particle flows that may generate electromagnetic interference or electrostatic adsorption.

[0086] Furthermore, correlating the time-frequency domain analysis results of the aircraft body with the real-time analysis results in the particle flow characteristic data involves using algorithms to compare and find the correspondence between the two in terms of time, frequency, or physical characteristics. For example, when the aircraft body exhibits abnormal vibration at a specific frequency, and the particle flow data also shows the presence of pulse flow at the same frequency, a resonant frequency characteristic may exist. Energy absorption characteristics may manifest as a sudden decrease in the vibration energy of the aircraft at a specific frequency, while the particle flow data shows an increase in the particle flow energy at that frequency. Charge interference characteristics may be reflected in abnormal fluctuations in sensor performance data, while the particle flow data indicates the presence of a strong charged particle flow. The aim is to accurately diagnose the physical mechanisms of the interactions.

[0087] Based on the identified resonant frequencies, energy absorption, or charge interference characteristics, determining whether intermittent, nonlinear coupling exists between the particle flow and the aircraft body involves assessing, based on the results of the aforementioned correlation analysis, whether this interaction is a continuous, linear, simple impact or a more complex, time-varying, and non-proportional dynamic coupling. For example, resonance is a typical nonlinear coupling, and its effects may be suddenly amplified under specific conditions. Intermittent coupling may manifest as an interaction that only occurs within a specific time period. The aim is to gain a deeper understanding of the nature of the interaction, providing a basis for subsequent precise intervention.

[0088] When coupling is detected, one or more local adjustment actions are selected from a pre-defined coupling cancellation action library based on the type and intensity of the coupling. This library contains refined adjustment schemes for different coupling types (such as resonance and charge interference) and intensities (such as slight, moderate, and severe). For example, for resonance, fine-tuning the aircraft's local attitude to change aerodynamic loads or adjusting power output to change the vibration frequency might be chosen. For charge interference, adjusting the aircraft's attitude to change charge distribution or activating partial discharge devices might be chosen. The aim is to provide targeted and effective intervention measures.

[0089] This application's solution, by introducing time-frequency domain analysis of aircraft vibration data, material stress data, and sensor performance data, along with real-time analysis of particle flow characteristic data, enables a more refined capture of the complex dynamic interactions between the aircraft and the fire-affected particle flow. Traditional methods may only identify anomalies but struggle to understand their underlying physical mechanisms. This solution, through time-frequency domain analysis, reveals transient and nonlinear vibration modes and energy changes. Combined with real-time identification of specific frequency pulse flows, high-density agglomerate flows, or strongly charged particle flows within the particle flow, the system can precisely correlate the aircraft's abnormal responses with the specific physical characteristics of the particle flow. This allows for the identification of deeper coupling characteristics such as resonant frequencies, energy absorption, or charge interference, thereby determining the presence of intermittent or nonlinear coupling effects. This precise assessment of the type and intensity of coupling effects enables the system to select more accurate and effective local adjustment actions from a pre-defined library of coupling effect cancellation actions, such as adjusting the aircraft's local attitude or power output. This directly and specifically cancels or mitigates specific interactions, avoiding blind adjustments or ineffective interventions, and significantly improving the effectiveness of obstacle avoidance strategies and the aircraft's stability.

[0090] Through the above technical solution, this application overcomes the limitations of existing technologies in identifying the deep-level interactions between aircraft and particle flows in complex fire environments. By introducing time-frequency domain analysis and real-time particle flow characteristic analysis, and performing precise correlation, this solution can identify intermittent and nonlinear coupling effects such as resonance, energy absorption, and charge interference, thereby achieving a more comprehensive and in-depth understanding of potential threat mechanisms. This accurate judgment of the type and intensity of coupling effects makes the selected local adjustment actions more targeted and effective, enabling more precise counteraction or mitigation of specific interactions. This significantly improves the flight stability, obstacle avoidance efficiency, and mission reliability of firefighting drones in complex and dynamic fire environments, effectively reducing the risk of damage or loss of control of the aircraft due to complex physical coupling effects.

[0091] In some embodiments, the steps of performing time-frequency domain analysis on the vibration data, material stress data, and sensor performance data of the aircraft body to extract its energy distribution and instantaneous phase changes within a specific frequency range further include: Short-time Fourier transform was performed on the vibration data, material stress data, and sensor performance data of the aircraft body to obtain the time-frequency diagram; From the time-frequency diagram, extract the energy distribution and instantaneous phase changes within the preset frequency range; Continuously receive the aircraft's own attitude adjustment data and motor load data; Time-frequency domain analysis was performed on the attitude adjustment data and motor load data to obtain time-frequency characteristics; The time-frequency diagram of the aircraft body is compared with the time-frequency characteristics of attitude adjustment data and motor load data; Identify anomalous energy distributions and instantaneous phase changes in the time-frequency plot of the aircraft itself, and exclude time-frequency features related to the aircraft's own attitude adjustments or motor load changes; Based on the excluded abnormal energy distribution and instantaneous phase changes, it can be determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of the particle flow.

[0092] Specifically, short-time Fourier transform (STFT) is performed on the vibration data, material stress data, and sensor performance data of the aircraft body. The purpose is to convert the time-domain signal into a time-frequency domain representation, thereby revealing the energy distribution and phase information of the signal at different times and frequencies. Through STFT, the dynamic response of the aircraft body at a specific time point and frequency range can be obtained, forming a time-frequency diagram. Extracting the energy distribution and instantaneous phase changes within a predetermined frequency range from the time-frequency diagram refers to quantitatively analyzing the energy intensity and phase changes within the frequency range of interest (e.g., the resonant frequency range that may be generated by interaction with particle flow) after obtaining the time-frequency diagram. This helps to focus on potential signals related to external disturbances.

[0093] In practical applications, continuously receiving the aircraft's own attitude adjustment data and motor load data refers to acquiring, in real time, the attitude commands generated by the aircraft's internal control system, actual attitude feedback, and load information such as the speed, current, and torque of each motor (e.g., rotor motor) during the aircraft's mission. This data reflects the aircraft's motion state and power output. Furthermore, time-frequency domain analysis is performed on the attitude adjustment data and motor load data to obtain time-frequency characteristics. The purpose of this analysis is to identify the vibration and stress characteristics caused by the aircraft's own operations (such as steering, ascent, descent, and hovering). These internally generated time-frequency characteristics can be used as baselines or noise sources for identification.

[0094] Therefore, comparing the time-frequency diagram of the aircraft body with the time-frequency characteristics of attitude adjustment data and motor load data aims to distinguish between the influence of external particulate flow and internal operational influence. This comparison identifies which features in the aircraft body's time-frequency diagram are caused by its own attitude adjustments or motor load changes. Specifically, identifying anomalous energy distributions and instantaneous phase changes in the aircraft body's time-frequency diagram and excluding time-frequency features related to the aircraft's own attitude adjustments or motor load changes means, based on the comparison, removing time-frequency features related to the aircraft's own operations from the overall time-frequency diagram of the aircraft body. The remaining anomalous energy distributions and instantaneous phase changes after this exclusion are more likely to represent real anomalies caused by specific physical characteristics of external particulate flow (such as impact, resonance, charge interference, etc.). Finally, based on the excluded anomalous energy distributions and instantaneous phase changes, it is determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of particulate flow, aiming to improve the accuracy of the judgment. By removing internal interference, the potential impact of external particulate flow on the aircraft's structure and performance can be assessed more reliably.

[0095] This application's solution introduces time-frequency domain analysis of the aircraft's own attitude adjustment data and motor load data, and compares this with time-frequency graphs of the aircraft's vibration, material stress, and sensor performance data. This effectively separates the vibration and stress characteristics caused by the aircraft's internal operations from anomalous signals induced by external particulate flow. This separation allows for more accurate identification of anomalous energy distribution and instantaneous phase changes in the aircraft's time-frequency graph, preventing the misinterpretation of signals generated during normal aircraft operation as external threats. By eliminating these internal interferences, it is possible to more accurately determine whether there are anomalous vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of the particulate flow, thus providing a more reliable basis for subsequent coupling effect assessment and selection of local adjustment actions.

[0096] In some embodiments, the step of performing short-time Fourier transform on the vibration data, material stress data, and sensor performance data of the aircraft body to obtain a time-frequency diagram further includes: In the time-frequency plot, transient and nonlinear time-frequency characteristics are identified and quantified. These characteristics include rapid instantaneous changes in energy within a specific frequency range and nonlinear drift of the instantaneous phase. Based on the identified transient and nonlinear time-frequency characteristics, it is determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical properties of the particle flow.

[0097] Specifically, transient and nonlinear time-frequency characteristics refer to energy distributions and phase behaviors in time-frequency graphs that are short in duration, exhibit drastic intensity changes, and do not follow simple linear laws. Rapid instantaneous changes in energy within a specific frequency range can be understood as significant, abrupt increases or decreases in the energy levels of one or more frequency components within an extremely short time window, such as the impact response caused by particle impact. These changes typically have high-frequency components and extremely short durations, making them difficult to capture effectively through long-term averaging or smoothing. Nonlinear drift in instantaneous phase refers to the fact that at a specific frequency, the phase of a signal no longer changes linearly over time, but exhibits jumps, distortions, or irregular fluctuations, usually due to nonlinear coupling or impact responses. In practical applications, identifying and quantifying these transient and nonlinear time-frequency characteristics can be achieved using advanced signal processing techniques, such as wavelet transform, Hilbert-Huang transform (HHT), or machine learning-based time-frequency feature extraction algorithms. These methods offer higher time resolution or stronger nonlinear feature capture capabilities than short-time Fourier transform. For example, wavelet transform can effectively capture transient events by analyzing signals at different scales through the selection of appropriate wavelet basis functions. Hilbert-Huang transform, on the other hand, can decompose nonstationary and nonlinear signals into a series of eigenmode functions, enabling analysis of instantaneous frequency and amplitude to reveal the nonlinear characteristics of the signal. By quantifying these characteristics, such as calculating instantaneous energy peaks, phase shift rates, or specific nonlinear indices, a numerical evaluation result can be obtained.

[0098] This application's solution, by identifying and quantifying transient and nonlinear time-frequency characteristics in time-frequency plots, can more precisely capture the complex, transient, and nonlinear interactions that may exist between particle flows and the aircraft body. Traditional time-frequency analysis methods, when processing such signals, may smooth out or ignore this crucial information due to limitations in time-frequency resolution or insufficient sensitivity to nonlinear features. However, by focusing on these transient and nonlinear features, such as rapid instantaneous changes in energy and nonlinear drift in instantaneous phase, abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical properties of particle flows (such as high-speed impacts, charge accumulation, or thermal shock) can be identified more accurately. This deeper analysis allows even weak or transient interactions to be effectively detected, thus avoiding misjudgments or delayed responses due to information omissions.

[0099] Based on the same inventive concept, this application also discloses an intelligent obstacle avoidance system for fire-fighting drones, such as... Figure 2 As shown, the system includes: The perception information acquisition module 1 is used to acquire perception information of the environment around the aircraft and quantify the degree of identification of potential threats in the perception information. Safety margin adjustment module 2 is used to adjust the safety margin range around the aircraft based on the quantitative results of the identification of potential threats and the current operating status of the aircraft. The strategy selection module 3 is used to select a quick avoidance strategy from a set of preset avoidance actions based on the safety margin when a potential threat enters the safety margin range. The strategy maintenance and adjustment module 4 is used to maintain the execution of the rapid avoidance strategy within a preset time period and to make local adjustments to the strategy using new environmental information; and The policy release module 5 is used to release the policy from its execution state under specific conditions.

[0100] The system provided in this application, through the close collaboration of its various modules, can significantly improve the intelligent obstacle avoidance capability and mission execution efficiency of UAVs in extreme environments.

[0101] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Detailed Implementation

[0102] The technical solutions in this application will now be clearly and completely described in conjunction with the accompanying drawings.

[0103] This application proposes an intelligent obstacle avoidance method for firefighting drones, such as... Figure 1 As shown, it includes the following steps: Acquire sensory information about the environment surrounding the aircraft and quantify the degree of identification of potential threats in the sensory information; Based on the quantification of the degree of identification of potential threats and the current operational status of the aircraft, adjust the safety margin range around the aircraft; When a potential threat enters the safety margin range, a rapid avoidance strategy is selected from the preset set of avoidance actions based on the safety margin range. During the execution of the rapid evasion strategy, the strategy is maintained within a preset time period, and local adjustments are made to the strategy using new environmental information; and Under certain conditions, the strategy's maintenance execution status is lifted.

[0104] This application aims to significantly improve the obstacle avoidance capabilities and decision-making efficiency of UAVs in complex fire environments by introducing mechanisms for quantifying the degree of potential threat identification, dynamically adjusting the safety margin range, and making local adjustments during strategy execution, thereby effectively addressing the challenges faced by traditional obstacle avoidance systems.

[0105] To better understand the intelligent obstacle avoidance method for firefighting drones proposed in this application, the key terms and implementation environment involved will be explained in detail below.

[0106] The aircraft typically refers to firefighting drones, which possess the ability to fly autonomously, perceive their environment, perform tasks, and avoid obstacles. These aircraft are equipped with various sensors, such as lidar, infrared thermal imagers, visual sensors, and inertial measurement units (IMUs), to acquire information about their surroundings and their own operational status.

[0107] Perception information refers to the data about the surrounding environment obtained by an aircraft through its various onboard sensors, including but not limited to the distance, shape, temperature, and motion status of obstacles, as well as environmental characteristics such as the density and distribution of smoke and particulate clouds.

[0108] Potential threats refer to any object or environmental factor that may pose a risk to the safe flight of an aircraft, such as building debris, falling objects, high-temperature areas, dense smoke, particulate clouds, other aircraft, etc.

[0109] Quantifying the degree of recognition refers to the process of numerically evaluating the clarity, confidence level, and completeness of potential threats in perceived information. For example, when lidar data is interfered with by particulate clouds, the sparsity and noise level of the obstacle point cloud can be used as quantitative indicators of the degree of recognition.

[0110] A safety margin range refers to a virtual protection zone established around an aircraft. When a potential threat enters this range, the aircraft needs to take evasive action. The size of this range can be dynamically adjusted based on the level of threat identification and the aircraft's operational status.

[0111] A rapid avoidance strategy refers to an emergency obstacle avoidance maneuver that an aircraft selects and executes from a pre-set set of actions when it detects a potential threat, such as emergency climb, descent, left turn, right turn, hovering, or a combination of maneuvers.

[0112] A preset time period refers to a set duration for which the strategy is maintained during the execution of a rapid avoidance strategy to ensure its effectiveness and stability. During this period, the main execution of the strategy is not easily interrupted, but partial adjustments are permitted.

[0113] Local adjustments refer to minor modifications made to parameters such as trajectory, attitude, or speed of the strategy during the execution of a rapid evasion strategy, based on new environmental information, to adapt to dynamic changes in the environment without deviating from the main strategy's evasion objective.

[0114] Specific conditions refer to the triggering conditions for disabling the strategy to maintain execution, such as the potential threat has been successfully avoided, the aircraft has left the danger zone, or a better strategy has been identified and needs to be switched.

[0115] The implementation environment of this application is usually a fire scene, which may contain complex and dynamically changing factors such as high temperature, dense smoke, particulate clouds, and structural collapse, which places extremely high demands on the perception and obstacle avoidance capabilities of UAVs.

[0116] The core of the intelligent obstacle avoidance method for firefighting drones proposed in this application lies in ensuring that drones can perform their tasks efficiently and safely in complex and ever-changing fire environments through a series of refined steps.

[0117] First, the method involves acquiring perception information about the aircraft's surrounding environment and quantifying the degree of identification of potential threats within that information. In practice, the aircraft can utilize various onboard sensors to acquire perception information. For example, it can use lidar to acquire point cloud data of the surrounding environment, infrared thermal imagers to acquire temperature distribution images of the environment, and visible light cameras to acquire visual images of the environment. This sensor data can be transmitted in real time to the aircraft's onboard processing unit. Quantifying the degree of identification of potential threats can be achieved in several ways. For example, for lidar data, the density, sparsity, noise level, and confidence of point cloud clustering can be calculated to quantify the degree of obstacle identification. When point cloud data is interfered with by smoke or particulate clouds, the point cloud density decreases, the noise level increases, and the clustering confidence decreases; these changes reflect a reduction in the degree of identification. Similarly, for infrared thermal imaging data, the contrast between the target and background, edge sharpness, and the uniformity of the temperature gradient in the image can be analyzed to quantify the degree of identification. When a target is obscured by high-temperature smoke or particulate clouds, the contrast will decrease, the edges will become blurred, and the temperature gradient will become uneven. These can all be used as a basis for quantifying the degree of recognition.

[0118] Secondly, based on the quantification of the potential threat identification level and the aircraft's current operating status, the safety margin range around the aircraft is adjusted. The aircraft's current operating status includes, but is not limited to, flight speed, altitude, attitude, remaining battery power, and mission priority. For example, when the potential threat identification level is low (e.g., the obstacle outline is blurred in dense smoke), to ensure safety, the safety margin range can be increased, allowing the aircraft to initiate evasive maneuvers at a greater distance from the obstacle. Conversely, when the identification level is high, the safety margin range can be appropriately reduced to improve flight efficiency. Simultaneously, the aircraft's current operating status also influences the adjustment of the safety margin range. For example, when the aircraft is flying at high speed, a larger safety margin range is needed to allow sufficient reaction time; when the aircraft's battery power is low or the mission priority is high, it may be necessary to appropriately reduce the safety margin range to complete the mission as quickly as possible, while ensuring safety. The adjustment of the safety margin range can be achieved through a pre-defined lookup table, a fuzzy logic controller, or a machine learning-based model.

[0119] Secondly, when a potential threat enters the safety margin range, a rapid avoidance strategy is selected from a pre-defined set of avoidance actions based on the safety margin. This set includes various predefined obstacle avoidance actions, such as emergency climb, descent, left turn, right turn, hovering, deceleration, or combinations thereof. Selecting a rapid avoidance strategy requires comprehensive consideration of the size of the safety margin, the type, location, and speed of the potential threat, as well as the aircraft's current operational state. For example, if the potential threat is a rapidly approaching falling object and the safety margin is large, the aircraft might choose an emergency climb or lateral maneuver. If the potential threat is a stationary obstacle and the safety margin is small, the aircraft might choose to decelerate and make minor attitude adjustments. Strategy selection can be achieved through decision trees, state machines, or reinforcement learning-based policy selectors.

[0120] Next, during the execution of the rapid avoidance strategy, the strategy is maintained for a preset time period, and local adjustments are made to the strategy using new environmental information. The preset time period is set to ensure that the selected strategy can be executed stably for a certain period, avoiding frequent strategy switching due to minor environmental fluctuations, which could lead to flight instability. During this preset time period, the aircraft continuously receives new environmental information. This new environmental information may include updates to sensor data, minor changes in the location of potential threats, changes in wind speed and direction, etc. Based on this new environmental information, the aircraft can make local adjustments to the rapidly avoidance strategy being executed. For example, if the strategy is to maneuver to the left, but new environmental information indicates the appearance of a new, smaller obstacle on the left, the aircraft can fine-tune its flight trajectory or attitude to avoid this new obstacle without changing its overall leftward maneuver direction. This local adjustment can be a small-scale attitude correction, speed adjustment, or trajectory fine-tuning, aimed at improving the adaptability and accuracy of the strategy while maintaining its overall stability.

[0121] Finally, under specific conditions, the strategy's sustained execution state is deactivated. These conditions can include various scenarios. For example, when the aircraft successfully avoids a potential threat and has left the danger zone, the current strategy's sustained execution state can be deactivated, and the aircraft can return to normal mission flight mode. Another example is when, during the execution of the current strategy, the aircraft detects a more serious threat or a better avoidance path, the current strategy can also be deactivated, and a new strategy can be switched to. Furthermore, when a preset time period ends and the environmental assessment determines that the current strategy is no longer the optimal choice, the strategy's sustained execution state can also be deactivated. After deactivating the strategy, the aircraft can reassess the environment and select a new flight strategy or resume normal flight.

[0122] The intelligent obstacle avoidance method for firefighting drones proposed in this application works by constructing a dynamic and adaptive obstacle avoidance decision-making loop. This method first acquires environmental perception information through multi-source sensors and quantifies the degree of potential threat identification. This solves the problem of identification uncertainty caused by the deterioration of perception data quality in complex fire environments using traditional methods. For example, in particulate cloud areas, lidar and infrared thermal imaging data may be blurry; this method, by quantifying the degree of identification, can objectively assess the reliability of the current perception information.

[0123] Subsequently, based on the quantification results and the aircraft's current operating status, the safety margin range is dynamically adjusted. This step is one of the core innovations of this method, enabling the obstacle avoidance strategy to be no longer static but adaptively adjusted according to environmental uncertainties and the aircraft's real-time status. For example, when the recognition level is low, the safety margin range is expanded to allow the aircraft more reaction time, thereby compensating for the risks brought about by perception uncertainty; when the aircraft is flying at high speed, the safety margin range will also increase accordingly to ensure sufficient braking and avoidance space. This dynamic adjustment mechanism effectively avoids the misjudgment or insufficient response caused by fixed safety thresholds in traditional methods.

[0124] When a potential threat enters the dynamically adjusted safety margin range, the system selects a rapid evasion strategy from a pre-set set of evasion actions. This selection process comprehensively considers the nature of the threat, the aircraft's status, and the safety margin range to ensure the timeliness and effectiveness of the selected strategy.

[0125] During the execution of the rapid evasion strategy, this method introduces a mechanism to maintain strategy execution within a preset time period and make local adjustments using new environmental information. This mechanism solves the problems of high-frequency strategy switching and computational delays caused by decision uncertainty in traditional methods. By setting a preset time period, the strategy can be executed stably, avoiding frequent path planning interruptions and restarts. At the same time, the local adjustment mechanism allows the aircraft to make fine-tuning corrections based on minor changes in the environment without interrupting the main strategy, such as fine-tuning attitude or trajectory, thereby improving the adaptability and accuracy of the strategy and ensuring the continuity and stability of evasion actions.

[0126] Finally, under specific conditions, the maintenance execution status of the strategy is deactivated, allowing the aircraft to flexibly exit the current evasion strategy and choose subsequent actions based on the new environmental assessment, whether to resume normal flight or switch to a new evasion strategy.

[0127] Through the close coordination of the above-mentioned steps, the method of this application can effectively address the problems of perception uncertainty, decision ambiguity, and response lag caused by extreme environments such as particulate clouds in fire scenes. By dynamically adjusting the safety margin and combining strategy maintenance with local adjustments, it significantly improves the intelligent obstacle avoidance capability and mission execution efficiency of firefighting drones in complex environments, thereby reducing the risk of mission failure and drone damage.

[0128] The intelligent obstacle avoidance method for firefighting drones proposed in this application demonstrates significant advantages and innovation compared to existing technologies in dealing with obstacle avoidance challenges in complex fire environments, especially under extreme conditions such as particulate clouds.

[0129] The core innovation of this application lies in the introduction of quantification of the degree of potential threat identification and the dynamic adjustment of the safety margin range based on this. For example, when particulate clouds cause sparse lidar point clouds and high noise, the system can quantify this reduction in identification level and expand the safety margin range accordingly. This means that the drone will initiate evasive maneuvers when it is further away from the potential threat, thereby gaining valuable time for decision-making and execution, and effectively compensating for the risks brought about by perception uncertainty.

[0130] Furthermore, this application introduces a mechanism to maintain strategy execution within a preset time period and make local adjustments using new environmental information during the execution of a rapid obstacle avoidance strategy. This mechanism is key to solving the problems of decision uncertainty and computational latency in traditional obstacle avoidance systems. In existing technologies, due to low confidence in obstacle threat assessment, the decision module frequently re-evaluates multiple potential actions, leading to repeated interruptions and restarts of path planning calculations, resulting in severe computational latency and obstacle avoidance lag. This application, by setting a preset time period to maintain strategy execution, effectively avoids such high-frequency strategy switching and computational interruptions, ensuring the continuity and stability of avoidance actions. Simultaneously, the local adjustment mechanism allows the aircraft to make fine-tuning corrections based on minor environmental changes without interrupting the main strategy, such as fine-tuning attitude or trajectory, thereby improving the adaptability and accuracy of the strategy and ensuring the continuity and stability of avoidance actions. This combination of strategy maintenance and local adjustment significantly improves the decision-making efficiency and obstacle avoidance response speed of UAVs in complex environments.

[0131] In summary, this application effectively addresses the challenges faced by existing firefighting drones in extreme fire environments due to perception uncertainty, decision-making ambiguity, and response lag, through innovative mechanisms such as quantifying the degree of potential threat identification, dynamically adjusting the safety margin range, and making local adjustments during strategy execution. The method described in this application enables drones to achieve more intelligent, stable, and efficient obstacle avoidance in complex and ever-changing environments, thereby significantly improving mission success rate and flight safety.

[0132] However, in actual firefighting scenarios, drones performing rapid evasive maneuvers may encounter sudden local environmental anomalies, such as strong airflow disturbances, unexpected minor collisions, or particulate impacts. These anomalies can cause significant physical impacts on the aircraft, leading to unexpected deviations in flight attitude and trajectory. Failure to respond to these physical impacts promptly and accurately may affect the effectiveness of rapid evasive strategies and even endanger the safety of the aircraft.

[0133] In this regard, this application further proposes that the steps of maintaining the execution of the strategy within a preset time period during the execution of the rapid avoidance strategy, and making local adjustments to the strategy using new environmental information, include: During the execution of the rapid evasion strategy, monitor the inertial measurement unit data of the aircraft itself; Monitor the actual response data of the motors on the aircraft itself; The deviation between the inertial measurement unit data and the expected inertial measurement unit data is compared to obtain the first deviation information; By comparing the deviation between the actual motor response data and the expected motor response data, a second deviation information is obtained; Based on the first deviation information and the second deviation information, determine the intensity of the physical impact of local environmental anomalies on the aircraft; When the physical impact intensity exceeds a preset threshold, select an attitude adjustment action from a preset local attitude correction library; Execute attitude adjustment actions and superimpose the attitude adjustment action instructions with the instructions for the rapid avoidance strategy; and After the attitude adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

[0134] Specifically, monitoring the inertial measurement unit (IMU) data of the aircraft body refers to acquiring real-time data such as acceleration, angular velocity, and attitude angles output by the aircraft's internal IMU. This data accurately reflects the aircraft's motion state and attitude changes in three-dimensional space. Simultaneously, monitoring the actual response data of the aircraft's motors refers to acquiring real-time operating parameters such as speed, current, and torque of each motor (e.g., the individual rotor motors of a multi-rotor UAV). This data directly reflects the actual output of the aircraft's power system.

[0135] The expected inertial measurement unit (IMU) data can be understood as the IMU data that the aircraft should have when executing the current rapid avoidance strategy under ideal conditions, i.e., without external physical impact interference. Similarly, the expected motor response data refers to the response data that the aircraft's motor system should have when executing the current rapid avoidance strategy under ideal conditions. By comparing the real-time monitored IMU data with the expected IMU data, the first deviation information can be obtained, which quantifies the difference between the aircraft's actual motion state and its expected motion state. Likewise, by comparing the real-time monitored actual motor response data with the expected motor response data, the second deviation information can be obtained, which reflects the difference between the actual motor output and the strategy command output.

[0136] In practical applications, the intensity of the physical impact on the aircraft caused by local environmental anomalies can be determined based on the first and second deviation information. For example, when the first deviation information (such as a sudden change in attitude angle or acceleration) and the second deviation information (such as abnormal fluctuations in motor speed) occur simultaneously and exceed a certain range, it indicates that the aircraft may have been subjected to an external physical impact. The intensity of the physical impact can be quantified using a preset algorithm, for example, by combining the magnitude and duration of the deviation for a comprehensive evaluation. When the intensity of the physical impact exceeds a preset threshold, it indicates that the aircraft is experiencing significant external disturbance and requires immediate attitude correction. At this time, the system will select one or more attitude adjustment actions from a preset local attitude correction library. The local attitude correction library stores various fine-tuning attitude actions for different impact intensities and directions, aiming to quickly restore the stability of the aircraft.

[0137] Furthermore, the selected attitude adjustment maneuver command will be executed and superimposed on the command of the currently executing rapid avoidance strategy. This superposition mechanism ensures that while making local attitude corrections, the aircraft can still roughly follow its main avoidance path, avoiding complete deviation from the avoidance target due to attitude corrections. After the attitude adjustment maneuver is completed, the aircraft will be guided back to the expected trajectory of the rapid avoidance strategy to ensure the successful completion of the obstacle avoidance mission.

[0138] This application's solution, through real-time monitoring of the aircraft's inertial measurement unit data and actual motor response data, can accurately capture physical impacts caused by any local environmental anomalies that the aircraft may encounter during the execution of a rapid avoidance strategy. It is precisely this continuous monitoring of these key operating parameters that enables the system to promptly detect deviations between the aircraft's actual and expected states. By comprehensively analyzing the first and second deviation information, the intensity of the physical impact can be accurately determined, thereby avoiding overreaction to minor disturbances while ensuring a rapid response to severe impacts. When the intensity of the physical impact reaches a level requiring intervention, the system can intelligently select the most suitable attitude adjustment action from a pre-set local attitude correction library and, through command superposition, quickly restore the aircraft's flight stability without interrupting the main avoidance strategy. This mechanism ensures that the aircraft maintains its attitude stability and trajectory accuracy when performing highly dynamic avoidance tasks in complex and changing environments.

[0139] Through the above technical solution, this application can significantly improve the flight stability and anti-interference capability of firefighting drones during the execution of rapid evasion strategies. Especially in complex environments such as fire scenes, when faced with sudden strong airflow, particulate impacts, or minor collisions, this solution can sense and quantify these physical impacts in real time and quickly take local attitude adjustment actions to counteract them. This effectively avoids the risk of the aircraft deviating from the predetermined evasion trajectory or becoming out of control due to external interference, thereby ensuring the effective execution of the rapid evasion strategy and greatly improving the operational safety and mission success rate of drones in hazardous environments.

[0140] However, in actual firefighting scenarios, the environment is complex and ever-changing. In particular, there may be a large number of particulate clouds in the fire scene, whose density, distribution, movement speed, and local temperature characteristics change dynamically over time. If only general local adjustments are made, it may not be able to adequately cope with these dynamic and cumulative environmental factors, resulting in insufficient precision and foresight in local adjustments, thereby affecting the overall effectiveness of obstacle avoidance strategies and the safety of the aircraft.

[0141] In this regard, this application further proposes the following steps for maintaining the execution of the aforementioned rapid avoidance strategy within a preset time period and making local adjustments to the strategy using new environmental information: While executing a rapid evasion strategy and remaining under strategy lock-in, continuously receive environmental data; Analyze the density, distribution, movement speed, and gradual temperature variation trends of particulate clouds in environmental data; Calculate the cumulative risk index based on the gradual change trend; When the cumulative risk index exceeds a preset threshold, the magnitude and direction of the local adjustments are dynamically corrected. Predict the aircraft's expected position and relative distance to the potential threat at the end of the lock-on, under the revised local adjustments; and Based on the prediction results, further fine-tune the parameters for local adjustments.

[0142] Specifically, during the execution of a rapid evasion strategy and while under strategic lock-on, the aircraft continuously receives environmental data. This environmental data can originate from various sensors onboard the aircraft, such as lidar, visual sensors, infrared sensors, and gas sensors, used to monitor the fire environment around the aircraft in real time. Particulate clouds refer to the aggregates of tiny particles in the air formed by smoke, ash, and combustion products generated by the fire.

[0143] Furthermore, the received environmental data will be used to analyze the density, distribution, movement speed, and gradual trends in local temperature of the particulate cloud. Density reflects the concentration of the particulate cloud, distribution indicates its spatial occupancy, movement speed reveals its dynamic evolution, and gradual trends in local temperature may indicate the spread of fire or the direction of thermal currents. Gradual trends refer to the continuous or periodic changes of these parameters over a period of time, rather than instantaneous values.

[0144] Based on this, a cumulative risk index is calculated according to the gradual changing trend of the particulate cloud. This index aims to quantify the potential cumulative threat of particulate clouds to the safe flight of aircraft. For example, high-density, fast-moving particulate clouds with continuously rising temperatures will generate a higher cumulative risk index. The calculation of this index can be based on a preset weighting model or machine learning algorithm, comprehensively considering the impact of various parameters on aircraft safety.

[0145] When the cumulative risk index exceeds a preset threshold, it indicates that the cumulative risk posed by the current environment to the aircraft has reached a level requiring intervention. At this point, dynamic adjustments to the magnitude and direction of local modifications are necessary. This correction is not a simple attitude adjustment, but a predictive adjustment based on cumulative risk, aimed at more effectively mitigating potential threats.

[0146] Subsequently, with the corrected local adjustments, the expected position of the aircraft and its relative distance to potential threats at the end of the strategy lockout are predicted. This prediction can be simulated using the aircraft's dynamics model, environmental model, and the corrected local adjustment commands. The prediction results can be used to evaluate the effectiveness of the current correction scheme.

[0147] Finally, based on the prediction results, the parameters of the local adjustments are further fine-tuned. If the prediction shows that the aircraft may still be too close to a potential threat at the end of the lock-on, or that the obstacle avoidance efficiency is poor, the parameters of the local adjustments (such as the magnitude, duration, or direction of the adjustment) can be optimized more precisely to ensure that the aircraft can safely complete the avoidance maneuver.

[0148] This application's solution achieves dynamic correction of local adjustments to rapid evasion strategies by introducing continuous monitoring and analysis of the dynamic characteristics of particulate clouds in the fire environment and calculating a cumulative risk index based on this. Traditional local adjustments may only be reactive adjustments based on instantaneous environmental information, making it difficult to effectively address dynamic threats with cumulative effects, such as particulate clouds. This application, by analyzing the density, distribution, movement speed, and gradual temperature changes of particulate clouds, can gain a more comprehensive and in-depth understanding of the environmental evolution patterns, thereby calculating a cumulative risk index that reflects long-term risks. It is precisely this forward-looking risk assessment that allows the aircraft to predictively and dynamically correct the magnitude and direction of local adjustments before the risk accumulates to a critical point. Furthermore, by predicting the aircraft's expected position and relative distance to potential threats after correction, and making fine-tuning accordingly, the accuracy and effectiveness of local adjustments are further ensured. This ensures that during strategy lockout, the aircraft can continuously and safely execute evasion maneuvers, avoiding accidents caused by dynamic environmental changes.

[0149] Through the aforementioned technical solution, this application significantly improves the adaptability and safety of firefighting drones when executing rapid obstacle avoidance strategies in complex fire environments. By deeply analyzing the dynamic characteristics of particulate clouds and calculating the cumulative risk index, the aircraft can identify and quantify potential cumulative threats earlier and more accurately, thereby achieving predictive and dynamic correction of local adjustments. This correction mechanism enables the aircraft to perform more refined and forward-looking trajectory optimization when facing dynamic environments such as dense smoke and surging hot air currents, effectively avoiding obstacle avoidance failures caused by sudden environmental changes or cumulative effects. Compared to solutions that adjust based solely on instantaneous environmental information, the solution presented in this application provides more stable and reliable obstacle avoidance performance. Especially during the lock-on period of long-term execution of the avoidance strategy, it effectively reduces the risk of secondary contact between the aircraft and potential threats, greatly enhancing the survivability and mission success rate of the drone in extreme environments.

[0150] In some embodiments, the steps of maintaining the execution of the rapid evasion strategy for a preset time period and making local adjustments to the strategy using new environmental information during the execution of the rapid evasion strategy include: continuously receiving environmental data during the execution of the rapid evasion strategy and while the strategy is locked; analyzing changes in the local turbulence intensity, temperature gradient, or illumination conditions of the particle cloud in the environmental data; calculating an environmental disturbance index based on the changes in the local turbulence intensity, temperature gradient, or illumination conditions of the particle cloud; correcting the magnitude and direction of the local adjustment when the environmental disturbance index exceeds a preset threshold; predicting the expected flight stability of the aircraft at the end of the lock under the corrected local adjustment; and fine-tuning the parameters of the local adjustment based on the prediction results.

[0151] Specifically, during the execution of a rapid evasion strategy and while in a strategy lock-on state, continuous reception of environmental data refers to the aircraft acquiring various physical parameters of the fire scene environment in real time through its onboard sensors, such as lidar, thermal imagers, visible light cameras, and anemometers. The strategy lock-on period can be understood as the state in which the aircraft maintains its main flight commands for a preset time period to ensure the continuity and effectiveness of evasion maneuvers while executing the rapid evasion strategy.

[0152] Analyzing changes in local turbulence intensity, temperature gradient, or illumination conditions of particulate clouds in environmental data involves processing and parsing received environmental data to identify and quantify local airflow instabilities (turbulence intensity), spatial rate of temperature change (temperature gradient), and fluctuations in ambient illumination levels caused by particulate clouds (such as smoke, dust, and combustion products) in a fire scene. These factors can directly impact the aerodynamic performance, sensor perception capabilities, and flight control of aircraft.

[0153] In practical applications, the environmental disturbance index is calculated based on changes in the local turbulence intensity, temperature gradient, or illumination conditions of particulate clouds. This involves comprehensively considering the degree of change in these environmental factors and their potential impact on aircraft stability. A pre-defined algorithm model (such as weighted average, fuzzy logic, or machine learning models) is used to calculate a comprehensive numerical index that quantifies the degree of disturbance the current environment causes to the aircraft's flight state. Its purpose is to provide a quantitative basis for subsequent local adjustments.

[0154] When the environmental disturbance index exceeds the preset threshold, the magnitude and direction of the local adjustment are corrected. This means that once the environmental disturbance index reaches or exceeds the preset safety threshold, it indicates that the current environment has reached a level that requires active intervention on the stability of the aircraft. At this time, the flight control system will correct the local adjustment parameters of the fast avoidance strategy being executed according to the magnitude and direction of the disturbance index. For example, it may increase or decrease the magnitude of the attitude adjustment or change the direction of the adjustment to counteract the impact of the environmental disturbance.

[0155] Furthermore, predicting the aircraft's expected flight stability at the end of the lock-up under the corrected local adjustments refers to simulating and predicting the stability of parameters such as flight attitude, speed, and position of the aircraft at the end of the strategy lock-up (i.e., when the rapid evasion strategy is completed) using the aircraft's dynamics and environmental models after correcting the magnitude and direction of the local adjustments. The purpose is to evaluate the effectiveness of the correction scheme and provide a basis for further fine-tuning.

[0156] Therefore, fine-tuning the parameters of local adjustments based on the prediction results refers to making precise adjustments to the parameters of local adjustments based on the predicted flight stability. For example, if the prediction shows that stability is still insufficient, the magnitude or duration of the adjustment may be further fine-tuned to ensure that the aircraft can maintain optimal flight status after completing the evasive maneuver.

[0157] This application's solution, by continuously monitoring and meticulously analyzing key disturbance factors in the fire environment, such as local turbulence intensity, temperature gradient, or illumination conditions of particulate clouds, can more accurately assess the impact of the environment on aircraft flight stability. It is precisely the introduction of the environmental disturbance index, a quantitative indicator, that enables the flight control system to promptly identify and respond to environmental disturbances exceeding preset thresholds. By correcting the magnitude and direction of local adjustments and combining this with predictions of future flight stability, this solution can proactively offset or mitigate the impact of environmental disturbances on aircraft attitude and trajectory, thereby ensuring the stability and safety of the aircraft when executing rapid avoidance strategies. This local adjustment mechanism based on environmental disturbance prediction effectively compensates for the limitations of basic solutions in responding to complex environmental disturbances.

[0158] Through the aforementioned technical solutions, firefighting drones can more effectively cope with complex and ever-changing environmental disturbances in fire scenes when executing rapid obstacle avoidance strategies, significantly improving the flight stability of the aircraft. This solution, through refined analysis and prediction of specific environmental factors such as local turbulence intensity, temperature gradients, or lighting conditions, makes local adjustments no longer simple reactive corrections, but rather forward-looking and adaptable. This ensures that the aircraft maintains precise attitude control and trajectory tracking even in extreme environments, greatly improving obstacle avoidance success rates and mission reliability.

[0159] In some embodiments, the steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information during the execution of the rapid avoidance strategy include: continuously receiving environmental data during the execution of the rapid avoidance strategy and while the strategy is locked; identifying changes in the clarity of potential threat features in the environmental data; calculating an information utilization efficiency index based on the changes in clarity; correcting the response parameters of the local adjustment when the information utilization efficiency index is lower than a preset threshold; predicting the relative distance between the aircraft and the potential threat and the obstacle avoidance efficiency at the end of the lock under the corrected local adjustment; and fine-tuning the parameters of the local adjustment based on the prediction results.

[0160] Specifically, during the execution of a rapid evasion strategy and while under strategic lock-on, the aircraft continuously receives environmental data from various sensors (such as visual sensors, lidar, and infrared sensors). Identifying changes in the sharpness of potential threat features within the environmental data refers to performing image processing or signal analysis on the received environmental data to assess the sharpness or identifiability of features such as the boundaries, textures, and shapes of potential threats (e.g., obstacles, fire sources, smoke plumes). For example, sharpness can be quantified by calculating metrics such as edge gradients, contrast, signal-to-noise ratio, or information entropy of the image.

[0161] Furthermore, an information utilization efficiency index is calculated based on changes in clarity. This index can be understood as the effective contribution of potential threat information in the current environmental data to obstacle avoidance decisions. A higher clarity results in a higher information utilization efficiency index; conversely, a lower clarity results in a lower information utilization efficiency index. The calculation of this index can be based on a preset mapping function or machine learning model, converting the clarity index into a value between 0 and 1.

[0162] When the information utilization efficiency index falls below a preset threshold, it indicates that the guidance provided by current environmental data for obstacle avoidance decisions is weakened, and the response parameters of local adjustments need to be corrected. Response parameters may include, but are not limited to, the magnitude, speed, acceleration limit, steering angle limit, or time constant of the local adjustment. For example, when the information utilization efficiency index is low, the magnitude of the local adjustment can be appropriately reduced, or the adjustment speed can be decreased, to avoid excessive or erroneous actions based on uncertain information.

[0163] Based on this, the relative distance between the aircraft and the potential threat at the end of the lock-on, as well as the obstacle avoidance efficiency, are predicted under the corrected local adjustments. This prediction can be calculated through simulation using the aircraft dynamics model, the environmental model, and the corrected local adjustment parameters. Obstacle avoidance efficiency can be quantified as the probability of successfully avoiding the threat, the smoothness of the obstacle avoidance path, or energy consumption, etc.

[0164] Finally, based on the prediction results, fine-tune the parameters of the local adjustments. If the prediction results show that the obstacle avoidance effect is not good or there is a potential risk, the response parameters can be further fine-tuned. For example, while ensuring safety, the adjustment range can be slightly increased to improve the obstacle avoidance speed, or the adjustment direction can be adjusted to optimize the obstacle avoidance path.

[0165] This application addresses the limitation of traditional methods in terms of the effectiveness of local adjustments when environmental information quality fluctuates by introducing an assessment of the clarity of potential threat features in environmental data. Specifically, when an aircraft is executing a rapid avoidance strategy and is in a locked-down state, the clarity of potential threat features in the continuously received environmental data may change due to the complexity of the fire environment. By identifying and quantifying this clarity change, the system can calculate the actual utilization efficiency of the current environmental information for obstacle avoidance decision-making. Due to this real-time assessment of information quality, when the information utilization efficiency index falls below a preset threshold, the system can promptly identify potential decision risks and proactively correct the response parameters of local adjustments. This correction mechanism prevents the aircraft from blindly making large adjustments based on unclear or unreliable information, thus avoiding secondary risks caused by misjudgment. Subsequently, by predicting the corrected local adjustments, the system can assess their impact on the relative distance between the aircraft and potential threats and obstacle avoidance efficiency, and fine-tune the local adjustment parameters based on this prediction result. This closed-loop feedback adjustment mechanism ensures that even under poor environmental information conditions, the aircraft can perform local obstacle avoidance in a more robust and reliable manner, thereby improving the adaptability and safety of the overall obstacle avoidance strategy.

[0166] Through the above technical solution, this application can effectively address the problem of fluctuating environmental data quality in complex environments such as fire scenes, significantly improving the robustness and accuracy of local adjustments made by firefighting drones during the execution of rapid obstacle avoidance strategies. By assessing the clarity of potential threat characteristics in real time and calculating the information utilization efficiency index, the aircraft can avoid making inappropriate or excessive adjustments based on low-quality information, thereby reducing the risk of obstacle avoidance failure. Furthermore, by predicting the effect of corrected local adjustments and making fine-tuning, the obstacle avoidance path and efficiency are further optimized, enabling the drone to complete obstacle avoidance tasks more safely and effectively in complex and changing environments, improving the success rate of the mission and the survivability of the aircraft.

[0167] In some embodiments, during the aforementioned rapid avoidance strategy, the step of maintaining the execution of the strategy for a preset time period and making local adjustments to the strategy using new environmental information includes: During the execution of the rapid evasion strategy and while in a strategy lock-on state, it continuously receives vibration data, material stress data, and sensor performance data of the aircraft body; Continuously receive particle flow characteristic data in the fire environment, including particle size, density, charge, and thermal radiation characteristics; Analyze the correlation between vibration data, material stress data, sensor performance data and particle flow characteristic data of the aircraft body to identify whether there is abnormal vibration, material stress accumulation or sensor function degradation caused by specific physical characteristics of particle flow. The intensity of the effects of specific interactions is quantified based on the degree of abnormal vibrations, material stress accumulation, or sensor function degradation identified. When the intensity of the influence exceeds a preset threshold, select one or more local adjustment actions from the preset interaction cancellation action library to cancel or mitigate the specific interaction; Execute local adjustment actions and superimpose the instructions for these local adjustment actions with the instructions for the rapid evasion strategy; and After the local adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

[0168] Specifically, during strategy lockout, vibration data of the aircraft body can be continuously collected by accelerometers or vibration sensors installed on the aircraft structure to monitor the dynamic response of the aircraft structural components; material stress data can be acquired by strain gauges or fiber optic sensors to assess the stress and fatigue accumulation of key structural components; sensor performance data can include information such as the output stability, noise level, or calibration deviation of key sensors for aircraft navigation and obstacle avoidance, for example, by monitoring the effective detection range of lidar and the temperature drift of infrared sensors. This data is continuously received to comprehensively reflect the real-time health status of the aircraft body in complex environments.

[0169] In fire scene environments, particle flow characteristics data refer to the fluid formed by the movement of solid or liquid particles in an airflow generated by events such as combustion, explosion, or structural collapse at a fire site. Specifically, particle flow characteristics data can include particle size, such as particle size distribution obtained through laser diffraction or image analysis techniques; particle density, such as estimation using mass flow meters or optical density sensors; particle charge, such as detecting the amount of charge carried by particles using electrostatic sensors; and particle thermal radiation characteristics, such as measuring the surface temperature or radiation intensity of particles using infrared thermal imagers or radiometers. These characteristics are crucial for understanding the physical interactions between particle flows and aircraft.

[0170] In practical applications, analyzing the correlation between vibration data, material stress data, sensor performance data, and particle flow characteristic data of the aircraft body aims to identify whether there are anomalies caused by specific physical characteristics of the particle flow. For example, when the particle flow has pulses of a specific frequency, it may resonate with the aircraft structure, leading to abnormal vibrations; the impact of high-density particle flow may cause material stress accumulation; and a particle flow carrying a strong charge may interfere with the normal operation of the aircraft's electronic equipment or sensors. This correlation analysis can be achieved through time-frequency analysis, correlation analysis, or machine learning models to accurately identify potential harmful interactions.

[0171] Furthermore, quantifying the impact intensity of a specific interaction refers to calculating a quantitative index based on the degree of identified abnormal vibrations, material stress accumulation, or sensor performance degradation, combined with a pre-defined hazard assessment model. For example, the amplitude of abnormal vibrations, the rate of stress accumulation, or the percentage of sensor performance degradation can be mapped to a risk score from 0 to 100; this score represents the impact intensity. When the impact intensity exceeds a preset threshold, it indicates that the aircraft is experiencing significant negative impacts, requiring immediate action.

[0172] As a preferred implementation, the preset interaction cancellation action library can include a variety of local adjustment actions. For example, to address abnormal vibrations caused by resonance, the local attitude or power output frequency of the aircraft can be adjusted to avoid the resonance point; to address the impact of high-density particle flow, the flight trajectory can be fine-tuned to reduce the frontal area or the flight attitude can be changed to disperse the impact force; to address charge interference, the electrostatic discharge device on the aircraft can be activated or the operating parameters of the sensors can be adjusted. These actions are designed to directly cancel or mitigate the specific negative impacts of particle flow on the aircraft.

[0173] The superposition of local adjustment commands with rapid avoidance strategy commands can be understood as adding local adjustment commands as fine-tuning or correction commands to the aircraft's control system in real time, while keeping the main rapid avoidance strategy command unchanged. For example, if the rapid avoidance strategy command is to translate to the left, and the local adjustment command is to slightly raise the nose to reduce particle impact, the final aircraft control command will be to translate to the left while simultaneously slightly raising the nose. This superposition ensures that while dealing with local interactions, the aircraft does not deviate from the main obstacle avoidance objective.

[0174] Therefore, guiding the aircraft back to the expected trajectory of the rapid avoidance strategy after the local adjustment action is completed means that after the local adjustment action is completed and the expected effect is achieved, the aircraft control system will gradually eliminate the impact of the local adjustment action, so that the aircraft can smoothly return to the original flight path of the rapid avoidance strategy to continue to complete the obstacle avoidance mission.

[0175] This application's solution, by continuously monitoring the aircraft's vibration data, material stress data, and sensor performance data during the execution of a rapid obstacle avoidance strategy, and combining this with specific physical characteristics data of particle flow in the fire environment, enables in-depth analysis of the microscopic physical interactions between the aircraft and the complex environment. It is precisely this meticulous monitoring and correlation analysis that allows the system to identify potential hazards such as abnormal vibrations, material stress accumulation, or sensor malfunction degradation caused by particle flow. By quantifying the intensity of these interactions, the system can accurately determine when intervention is necessary. When the intensity reaches a preset threshold, the system can select targeted local adjustment actions from a pre-set library of countermeasures, such as adjusting flight attitude or power output, to directly counteract or mitigate the negative impact of particle flow. This local adjustment action, superimposed on the rapid obstacle avoidance strategy commands, ensures that the aircraft's anti-interference capability and stability in complex fire environments are effectively improved without interrupting the main obstacle avoidance mission. Finally, after the local adjustment action is completed, the aircraft is guided back to the expected trajectory, ensuring the continuity and effectiveness of the obstacle avoidance mission.

[0176] Through the above technical solution, this application effectively addresses the problem of poor obstacle avoidance or damage to aircraft performance caused by traditional obstacle avoidance methods in complex fire environments due to insufficient consideration of the specific physical interactions between particle flow and the aircraft body. This solution, by real-time monitoring of the aircraft's health data and particle flow characteristic data, and performing deep correlation analysis, enables the system to accurately identify and quantify anomalies caused by particle flow. Consequently, the aircraft can promptly take targeted local adjustments to effectively offset or mitigate the negative impacts of particle flow, significantly improving the flight stability, safety, and mission reliability of firefighting drones in fire environments filled with smoke, ash, and other particulate matter. This not only avoids aircraft instability, structural damage, or sensor failure caused by microscopic physical interactions but also ensures the smooth execution of rapid obstacle avoidance strategies, thereby improving overall obstacle avoidance efficiency and the success rate of firefighting missions.

[0177] In some embodiments, the step of continuously receiving vibration data, material stress data, and sensor performance data of the aircraft body during the execution of a rapid evasion strategy and while in a strategy lock-in state further includes: Time-frequency domain analysis was performed on the vibration data, material stress data, and sensor performance data of the aircraft body to extract its energy distribution and instantaneous phase changes within a specific frequency range; Real-time analysis of particle flow characteristic data can identify whether there are pulse flows of specific frequencies, high-density agglomerate flows, or strongly charged particle flows. The time-frequency domain analysis results of the aircraft body are correlated with the real-time analysis results in the particle flow characteristic data to identify whether there are characteristics such as resonant frequency, energy absorption or charge interference. Based on the identified resonant frequency, energy absorption, or charge interference characteristics, determine whether there is intermittent or nonlinear coupling between the particle flow and the aircraft body. When a coupling effect is detected, one or more local adjustment actions are selected from the preset coupling effect cancellation action library according to the type and strength of the coupling effect. The local adjustment actions include adjusting the local attitude or power output of the aircraft. Execute local adjustment actions and superimpose the instructions for these local adjustment actions with the instructions for the rapid evasion strategy; and After the local adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

[0178] Specifically, time-frequency domain analysis of vibration data, material stress data, and sensor performance data of the aircraft body refers to transforming the signal from the time domain to the time-frequency domain using signal processing methods such as Fourier transform and wavelet transform, so as to observe the changes of the signal in time and frequency simultaneously. This aims to reveal the transient and nonlinear characteristics of the signal, such as energy concentration or dispersion within a specific frequency range, and rapid instantaneous phase drift. These characteristics are often direct manifestations of complex interactions. Among them, the energy distribution and instantaneous phase change within a specific frequency range can be understood as the energy intensity and vibration mode exhibited by the aircraft structure or sensor response at certain frequencies at a specific point in time, with the purpose of identifying potential resonance or anomalous responses.

[0179] Meanwhile, real-time analysis of particle flow characteristic data refers to the continuous monitoring and rapid analysis of particle flows (including particle size, density, charge, and thermal radiation characteristics) in the fire environment using high-speed sensors and data processing units. The aim is to promptly identify specific types of particle flows with potential hazards, such as specific-frequency pulse flows that may induce resonance, high-density clump flows that may cause structural impact or blockage, and strongly charged particle flows that may generate electromagnetic interference or electrostatic adsorption.

[0180] Furthermore, correlating the time-frequency domain analysis results of the aircraft body with the real-time analysis results in the particle flow characteristic data involves using algorithms to compare and find the correspondence between the two in terms of time, frequency, or physical characteristics. For example, when the aircraft body exhibits abnormal vibration at a specific frequency, and the particle flow data also shows the presence of pulse flow at the same frequency, a resonant frequency characteristic may exist. Energy absorption characteristics may manifest as a sudden decrease in the vibration energy of the aircraft at a specific frequency, while the particle flow data shows an increase in the particle flow energy at that frequency. Charge interference characteristics may be reflected in abnormal fluctuations in sensor performance data, while the particle flow data indicates the presence of a strong charged particle flow. The aim is to accurately diagnose the physical mechanisms of the interactions.

[0181] Based on the identified resonant frequencies, energy absorption, or charge interference characteristics, determining whether intermittent, nonlinear coupling exists between the particle flow and the aircraft body involves assessing, based on the results of the aforementioned correlation analysis, whether this interaction is a continuous, linear, simple impact or a more complex, time-varying, and non-proportional dynamic coupling. For example, resonance is a typical nonlinear coupling, and its effects may be suddenly amplified under specific conditions. Intermittent coupling may manifest as an interaction that only occurs within a specific time period. The aim is to gain a deeper understanding of the nature of the interaction, providing a basis for subsequent precise intervention.

[0182] When coupling is detected, one or more local adjustment actions are selected from a pre-defined coupling cancellation action library based on the type and intensity of the coupling. This library contains refined adjustment schemes for different coupling types (such as resonance and charge interference) and intensities (such as slight, moderate, and severe). For example, for resonance, fine-tuning the aircraft's local attitude to change aerodynamic loads or adjusting power output to change the vibration frequency might be chosen. For charge interference, adjusting the aircraft's attitude to change charge distribution or activating partial discharge devices might be chosen. The aim is to provide targeted and effective intervention measures.

[0183] This application's solution, by introducing time-frequency domain analysis of aircraft vibration data, material stress data, and sensor performance data, along with real-time analysis of particle flow characteristic data, enables a more refined capture of the complex dynamic interactions between the aircraft and the fire-affected particle flow. Traditional methods may only identify anomalies but struggle to understand their underlying physical mechanisms. This solution, through time-frequency domain analysis, reveals transient and nonlinear vibration modes and energy changes. Combined with real-time identification of specific frequency pulse flows, high-density agglomerate flows, or strongly charged particle flows within the particle flow, the system can precisely correlate the aircraft's abnormal responses with the specific physical characteristics of the particle flow. This allows for the identification of deeper coupling characteristics such as resonant frequencies, energy absorption, or charge interference, thereby determining the presence of intermittent or nonlinear coupling effects. This precise assessment of the type and intensity of coupling effects enables the system to select more accurate and effective local adjustment actions from a pre-defined library of coupling effect cancellation actions, such as adjusting the aircraft's local attitude or power output. This directly and specifically cancels or mitigates specific interactions, avoiding blind adjustments or ineffective interventions, and significantly improving the effectiveness of obstacle avoidance strategies and the aircraft's stability.

[0184] Through the above technical solution, this application overcomes the limitations of existing technologies in identifying the deep-level interactions between aircraft and particle flows in complex fire environments. By introducing time-frequency domain analysis and real-time particle flow characteristic analysis, and performing precise correlation, this solution can identify intermittent and nonlinear coupling effects such as resonance, energy absorption, and charge interference, thereby achieving a more comprehensive and in-depth understanding of potential threat mechanisms. This accurate judgment of the type and intensity of coupling effects makes the selected local adjustment actions more targeted and effective, enabling more precise counteraction or mitigation of specific interactions. This significantly improves the flight stability, obstacle avoidance efficiency, and mission reliability of firefighting drones in complex and dynamic fire environments, effectively reducing the risk of damage or loss of control of the aircraft due to complex physical coupling effects.

[0185] In some embodiments, the steps of performing time-frequency domain analysis on the vibration data, material stress data, and sensor performance data of the aircraft body to extract its energy distribution and instantaneous phase changes within a specific frequency range further include: Short-time Fourier transform was performed on the vibration data, material stress data, and sensor performance data of the aircraft body to obtain the time-frequency diagram; From the time-frequency diagram, extract the energy distribution and instantaneous phase changes within the preset frequency range; Continuously receive the aircraft's own attitude adjustment data and motor load data; Time-frequency domain analysis was performed on the attitude adjustment data and motor load data to obtain time-frequency characteristics; The time-frequency diagram of the aircraft body is compared with the time-frequency characteristics of attitude adjustment data and motor load data; Identify anomalous energy distributions and instantaneous phase changes in the time-frequency plot of the aircraft itself, and exclude time-frequency features related to the aircraft's own attitude adjustments or motor load changes; Based on the excluded abnormal energy distribution and instantaneous phase changes, it can be determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of the particle flow.

[0186] Specifically, short-time Fourier transform (STFT) is performed on the vibration data, material stress data, and sensor performance data of the aircraft body. The purpose is to convert the time-domain signal into a time-frequency domain representation, thereby revealing the energy distribution and phase information of the signal at different times and frequencies. Through STFT, the dynamic response of the aircraft body at a specific time point and frequency range can be obtained, forming a time-frequency diagram. Extracting the energy distribution and instantaneous phase changes within a predetermined frequency range from the time-frequency diagram refers to quantitatively analyzing the energy intensity and phase changes within the frequency range of interest (e.g., the resonant frequency range that may be generated by interaction with particle flow) after obtaining the time-frequency diagram. This helps to focus on potential signals related to external disturbances.

[0187] In practical applications, continuously receiving the aircraft's own attitude adjustment data and motor load data refers to acquiring, in real time, the attitude commands generated by the aircraft's internal control system, actual attitude feedback, and load information such as the speed, current, and torque of each motor (e.g., rotor motor) during the aircraft's mission. This data reflects the aircraft's motion state and power output. Furthermore, time-frequency domain analysis is performed on the attitude adjustment data and motor load data to obtain time-frequency characteristics. The purpose of this analysis is to identify the vibration and stress characteristics caused by the aircraft's own operations (such as steering, ascent, descent, and hovering). These internally generated time-frequency characteristics can be used as baselines or noise sources for identification.

[0188] Therefore, comparing the time-frequency diagram of the aircraft body with the time-frequency characteristics of attitude adjustment data and motor load data aims to distinguish between the influence of external particulate flow and internal operational influence. This comparison identifies which features in the aircraft body's time-frequency diagram are caused by its own attitude adjustments or motor load changes. Specifically, identifying anomalous energy distributions and instantaneous phase changes in the aircraft body's time-frequency diagram and excluding time-frequency features related to the aircraft's own attitude adjustments or motor load changes means, based on the comparison, removing time-frequency features related to the aircraft's own operations from the overall time-frequency diagram of the aircraft body. The remaining anomalous energy distributions and instantaneous phase changes after this exclusion are more likely to represent real anomalies caused by specific physical characteristics of external particulate flow (such as impact, resonance, charge interference, etc.). Finally, based on the excluded anomalous energy distributions and instantaneous phase changes, it is determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of particulate flow, aiming to improve the accuracy of the judgment. By removing internal interference, the potential impact of external particulate flow on the aircraft's structure and performance can be assessed more reliably.

[0189] This application's solution introduces time-frequency domain analysis of the aircraft's own attitude adjustment data and motor load data, and compares this with time-frequency graphs of the aircraft's vibration, material stress, and sensor performance data. This effectively separates the vibration and stress characteristics caused by the aircraft's internal operations from anomalous signals induced by external particulate flow. This separation allows for more accurate identification of anomalous energy distribution and instantaneous phase changes in the aircraft's time-frequency graph, preventing the misinterpretation of signals generated during normal aircraft operation as external threats. By eliminating these internal interferences, it is possible to more accurately determine whether there are anomalous vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of the particulate flow, thus providing a more reliable basis for subsequent coupling effect assessment and selection of local adjustment actions.

[0190] In some embodiments, the step of performing short-time Fourier transform on the vibration data, material stress data, and sensor performance data of the aircraft body to obtain a time-frequency diagram further includes: In the time-frequency plot, transient and nonlinear time-frequency characteristics are identified and quantified. These characteristics include rapid instantaneous changes in energy within a specific frequency range and nonlinear drift of the instantaneous phase. Based on the identified transient and nonlinear time-frequency characteristics, it is determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical properties of the particle flow.

[0191] Specifically, transient and nonlinear time-frequency characteristics refer to energy distributions and phase behaviors in time-frequency graphs that are short in duration, exhibit drastic intensity changes, and do not follow simple linear laws. Rapid instantaneous changes in energy within a specific frequency range can be understood as significant, abrupt increases or decreases in the energy levels of one or more frequency components within an extremely short time window, such as the impact response caused by particle impact. These changes typically have high-frequency components and extremely short durations, making them difficult to capture effectively through long-term averaging or smoothing. Nonlinear drift in instantaneous phase refers to the fact that at a specific frequency, the phase of a signal no longer changes linearly over time, but exhibits jumps, distortions, or irregular fluctuations, usually due to nonlinear coupling or impact responses. In practical applications, identifying and quantifying these transient and nonlinear time-frequency characteristics can be achieved using advanced signal processing techniques, such as wavelet transform, Hilbert-Huang transform (HHT), or machine learning-based time-frequency feature extraction algorithms. These methods offer higher time resolution or stronger nonlinear feature capture capabilities than short-time Fourier transform. For example, wavelet transform can effectively capture transient events by analyzing signals at different scales through the selection of appropriate wavelet basis functions. Hilbert-Huang transform, on the other hand, can decompose nonstationary and nonlinear signals into a series of eigenmode functions, enabling analysis of instantaneous frequency and amplitude to reveal the nonlinear characteristics of the signal. By quantifying these characteristics, such as calculating instantaneous energy peaks, phase shift rates, or specific nonlinear indices, a numerical evaluation result can be obtained.

[0192] This application's solution, by identifying and quantifying transient and nonlinear time-frequency characteristics in time-frequency plots, can more precisely capture the complex, transient, and nonlinear interactions that may exist between particle flows and the aircraft body. Traditional time-frequency analysis methods, when processing such signals, may smooth out or ignore this crucial information due to limitations in time-frequency resolution or insufficient sensitivity to nonlinear features. However, by focusing on these transient and nonlinear features, such as rapid instantaneous changes in energy and nonlinear drift in instantaneous phase, abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical properties of particle flows (such as high-speed impacts, charge accumulation, or thermal shock) can be identified more accurately. This deeper analysis allows even weak or transient interactions to be effectively detected, thus avoiding misjudgments or delayed responses due to information omissions.

[0193] Based on the same inventive concept, this application also discloses an intelligent obstacle avoidance system for fire-fighting drones, such as... Figure 2 As shown, the system includes: The perception information acquisition module 1 is used to acquire perception information of the environment around the aircraft and quantify the degree of identification of potential threats in the perception information. Safety margin adjustment module 2 is used to adjust the safety margin range around the aircraft based on the quantitative results of the identification of potential threats and the current operating status of the aircraft. The strategy selection module 3 is used to select a quick avoidance strategy from a set of preset avoidance actions based on the safety margin when a potential threat enters the safety margin range. The strategy maintenance and adjustment module 4 is used to maintain the execution of the rapid avoidance strategy within a preset time period and to make local adjustments to the strategy using new environmental information; and The policy release module 5 is used to release the policy from its execution state under specific conditions.

[0194] The system provided in this application, through the close collaboration of its various modules, can significantly improve the intelligent obstacle avoidance capability and mission execution efficiency of UAVs in extreme environments.

[0195] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application.

Claims

1. A method for intelligent obstacle avoidance of firefighting drones, characterized in that, include: Acquire sensory information about the environment surrounding the aircraft and quantify the degree of identification of potential threats in the sensory information; Based on the quantification of the degree of identification of the potential threats and the current operating status of the aircraft, the safety margin range around the aircraft is adjusted. When the potential threat enters the safety margin range, a quick avoidance strategy is selected from a preset set of avoidance actions based on the safety margin range; During the execution of the rapid avoidance strategy, the strategy is maintained for a preset time period, and the strategy is locally adjusted using new environmental information. And under specific conditions, the maintenance execution state of the strategy can be terminated.

2. The intelligent obstacle avoidance method for firefighting drones according to claim 1, characterized in that, The steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information include: During the execution of the aforementioned rapid evasion strategy, inertial measurement unit data of the aircraft body is monitored; Monitor the actual response data of the motors on the aircraft itself; The deviation between the inertial measurement unit data and the expected inertial measurement unit data is compared to obtain the first deviation information; By comparing the deviation between the actual motor response data and the expected motor response data, a second deviation information is obtained; Based on the first deviation information and the second deviation information, determine the intensity of the physical impact of local environmental anomalies on the aircraft; When the physical impact intensity exceeds a preset threshold, an attitude adjustment action is selected from a preset local attitude correction library; Execute the attitude adjustment action and superimpose the instruction of the attitude adjustment action with the instruction of the rapid avoidance strategy; And after the attitude adjustment maneuver is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

3. The intelligent obstacle avoidance method for firefighting drones according to claim 1, characterized in that, The steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information include: During the execution of the rapid evasion strategy and while the strategy is locked, environmental data is continuously received; Analyze the density, distribution, movement speed, and gradual change trends of local temperature of the particulate cloud in the environmental data. Calculate the cumulative risk index based on the described gradual change trend; When the cumulative risk index exceeds a preset threshold, the magnitude and direction of the local adjustment are dynamically corrected. Predict the aircraft's expected position and relative distance to the potential threat at the end of the lock-on, under the revised local adjustments. And based on the prediction results, the parameters of the local adjustment are further fine-tuned.

4. The intelligent obstacle avoidance method for firefighting drones according to claim 1, characterized in that, The steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information include: During the execution of the rapid evasion strategy and while the strategy is locked, environmental data is continuously received; Analyze the changes in local turbulence intensity, temperature gradient, or illumination conditions of the particulate cloud in the environmental data; The environmental disturbance index is calculated based on the changes in local turbulence intensity, temperature gradient, or illumination conditions of the particulate cloud. When the environmental disturbance index exceeds a preset threshold, the magnitude and direction of the local adjustment are corrected. Predict the expected flight stability of the aircraft at the end of the lock-up period under the corrected local adjustment; and fine-tune the parameters of the local adjustment based on the prediction results.

5. The intelligent obstacle avoidance method for firefighting drones according to claim 1, characterized in that, The steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information include: During the execution of the rapid evasion strategy and while the strategy is locked, environmental data is continuously received; Identify changes in the clarity of potential threat features in the environmental data; Calculate the information utilization efficiency index based on the aforementioned clarity change; When the information utilization efficiency index is lower than a preset threshold, the response parameters of the local adjustment are corrected. The predicted relative distance between the aircraft and the potential threat and the obstacle avoidance efficiency at the end of the lock-on is based on the revised local adjustments. And based on the prediction results, fine-tune the parameters of the local adjustment.

6. The intelligent obstacle avoidance method for firefighting drones according to claim 1, characterized in that, The steps of maintaining the execution of the rapid avoidance strategy for a preset time period and making local adjustments to the strategy using new environmental information include: During the execution of the rapid evasion strategy and while in strategy lock-on, vibration data, material stress data, and sensor performance data of the aircraft body are continuously received; Continuously receive particle flow characteristic data in the fire environment, including particle size, density, charge, and thermal radiation characteristics; Analyze the correlation between the vibration data, material stress data, sensor performance data of the aircraft body and the particle flow characteristic data to identify whether there is abnormal vibration, material stress accumulation or sensor function degradation caused by the specific physical characteristics of the particle flow. The intensity of the effect of the specific interaction is quantified based on the degree of abnormal vibration, material stress accumulation, or sensor function degradation identified. When the intensity of the influence exceeds a preset threshold, one or more local adjustment actions to counteract or mitigate the specific interaction are selected from a preset interaction counteraction action library; Execute the local adjustment action, and superimpose the instructions of the local adjustment action with the instructions of the rapid evasion strategy; And after the local adjustment action is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

7. The intelligent obstacle avoidance method for firefighting drones according to claim 6, characterized in that, The step of continuously receiving vibration data, material stress data, and sensor performance data of the aircraft body during the execution of the rapid evasion strategy and while in strategy lock-on state further includes: Time-frequency domain analysis was performed on the vibration data, material stress data, and sensor performance data of the aircraft body to extract its energy distribution and instantaneous phase changes within a specific frequency range; The particle flow characteristic data is analyzed in real time to identify whether there are pulse flows of specific frequencies, high-density agglomerate flows, or particle flows with strong charges. The time-frequency domain analysis results of the aircraft body are correlated with the real-time analysis results in the particle flow characteristic data to identify whether there are features such as resonant frequency, energy absorption or charge interference. Based on the identified resonant frequency, energy absorption, or charge interference characteristics, it is determined whether there is intermittent or nonlinear coupling between the particle flow and the aircraft body. When the coupling effect is determined to exist, one or more local adjustment actions are selected from a preset coupling effect cancellation action library according to the type and intensity of the coupling effect. The local adjustment actions include adjusting the local attitude or power output of the aircraft. Execute the local adjustment action, and superimpose the instructions of the local adjustment action with the instructions of the rapid evasion strategy; And after the local adjustment action is completed, guide the aircraft back to the expected trajectory of the rapid evasion strategy.

8. The intelligent obstacle avoidance method for firefighting drones according to claim 7, characterized in that, The step of performing time-frequency domain analysis on the vibration data, material stress data, and sensor performance data of the aircraft body to extract its energy distribution and instantaneous phase changes within a specific frequency range further includes: Short-time Fourier transform is performed on the vibration data, material stress data, and sensor performance data of the aircraft body to obtain a time-frequency diagram; From the time-frequency diagram, the energy distribution and instantaneous phase changes within the preset frequency range are extracted; Continuously receive the aircraft's own attitude adjustment data and motor load data; Time-frequency domain analysis was performed on the attitude adjustment data and motor load data to obtain time-frequency characteristics; The time-frequency diagram of the aircraft body is compared with the time-frequency characteristics of the attitude adjustment data and motor load data; Identify abnormal energy distributions and instantaneous phase changes in the time-frequency diagram of the aircraft body, and exclude time-frequency features related to the aircraft's own attitude adjustment or motor load changes; Based on the excluded abnormal energy distribution and instantaneous phase changes, it is determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical characteristics of the particle flow.

9. A method for intelligent obstacle avoidance of a fire-fighting drone according to claim 8, characterized in that, The step of performing a short-time Fourier transform on the vibration data, material stress data, and sensor performance data of the aircraft body to obtain a time-frequency diagram further includes: In the time-frequency plot, identify and quantify transient and nonlinear time-frequency characteristics; Based on the identified transient and nonlinear time-frequency characteristics, it is determined whether there are abnormal vibrations, material stress accumulation, or sensor function degradation caused by specific physical properties of particle flow.

10. An intelligent obstacle avoidance system for firefighting drones, characterized in that, The system includes: The perception information acquisition module is used to acquire perception information about the environment around the aircraft and quantify the degree of identification of potential threats in the perception information. The safety margin adjustment module is used to adjust the safety margin range around the aircraft based on the quantification results of the identification degree of the potential threat and the current operating status of the aircraft. The strategy selection module is used to select a quick avoidance strategy from a preset set of avoidance actions based on the security margin range when the potential threat enters the security margin range. A strategy maintenance and adjustment module is used to maintain the execution of the rapid evasion strategy for a preset time period during the execution of the strategy, and to make local adjustments to the strategy using new environmental information; and The policy release module is used to release the policy from its maintained execution state under specific conditions.

Images