Unmanned aerial vehicle adaptive obstacle avoidance control method based on task context awareness

Abstract

The invention discloses an unmanned aerial vehicle adaptive obstacle avoidance control method based on task context awareness. The method comprises the steps of collecting and analyzing context information of a task currently executed by an unmanned aerial vehicle; presetting an obstacle avoidance parameter template library corresponding to various task types; according to task context information acquired in real time, selecting a most matched obstacle avoidance strategy template in the parameter library; or, generating a dynamic obstacle avoidance parameter group through a reasoning / interpolation mode based on the context features, and completing the self-adaptive configuration of the obstacle avoidance strategy. According to the invention, the task context is introduced as an obstacle avoidance strategy driving factor, the most suitable obstacle avoidance parameter group can be matched or generated in real time according to different task types, the fundamental conversion from fixed parameters to dynamic scheduling is realized, and the generalization ability and operation safety of the unmanned aerial vehicle in complex and diverse task environments are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicles (UAVs), and specifically relates to an adaptive obstacle avoidance control method for UAVs based on task context awareness. Background Technology

[0002] With the rapid development of drone technology, its application scenarios are constantly expanding, widely involving multiple fields such as security inspection, emergency response, and logistics transportation. During the process of drones achieving autonomous flight, obstacle avoidance control is one of the core technologies for ensuring flight safety and improving mission completion efficiency.

[0003] Existing drone obstacle avoidance systems generally employ a "fixed-distance obstacle avoidance" strategy. This involves configuring a fixed distance threshold (such as 1.5 meters or 2 meters) as the trigger condition for obstacle detection and avoidance. When the drone detects an obstacle in its surroundings using sensors such as lidar, binocular vision, or ultrasound, and the obstacle is within the set threshold, the flight control system immediately triggers actions such as deceleration, hovering, or obstacle avoidance to avoid collision risks. Simultaneously, the flight control system works to adjust attitude and replan the path.

[0004] However, this fixed threshold obstacle avoidance strategy has many obvious drawbacks in practical applications: First, there is a lack of task context awareness. Existing systems do not take into account the different obstacle avoidance strategy requirements of different task types. For example, in low-speed nighttime inspection missions, UAVs have extremely high requirements for the accuracy of environmental perception and extremely low fault tolerance, requiring more cautious obstacle avoidance methods. However, in high-speed daytime transportation missions, if a conservative fixed-distance obstacle avoidance strategy is still used, it will cause the UAV to frequently decelerate or even make misjudgments, seriously affecting the execution efficiency of the mission.

[0005] Secondly, the obstacle avoidance strategy is rigid and has poor adaptability. The fixed distance strategy cannot adjust the obstacle avoidance parameters according to the dynamic factors such as the drone's flight status, speed, altitude, mission urgency, and environmental complexity. This leads to the problem of "avoiding too early" or "avoiding too late" in some mission scenarios.

[0006] Third, strategy optimization relies on experience and has poor versatility. Most of the fixed thresholds in the current system are set manually based on experience and lack task migration capabilities. When changing task scenarios, recalibration and debugging are often required, which undoubtedly increases deployment costs and time.

[0007] In summary, existing UAV obstacle avoidance control methods lack task-driven adaptive capabilities, making it difficult to balance flight safety and mission efficiency in real-world operating environments with multiple scenarios and tasks. Therefore, there is an urgent need for a control method with task context awareness and the ability to dynamically adjust obstacle avoidance strategies. Summary of the Invention

[0008] To address the aforementioned issues, this invention provides a task context-aware adaptive obstacle avoidance control method for unmanned aerial vehicles (UAVs). This method can dynamically adjust obstacle avoidance parameters based on the type of the current flight mission and environmental requirements, thereby improving the safety and mission execution efficiency of UAVs in various scenarios.

[0009] Therefore, the technical solution of the present invention is: an adaptive obstacle avoidance control method for unmanned aerial vehicles based on task context awareness, comprising the following steps: S1. Collect and parse the context information of the task currently being performed by the UAV; S2. Pre-set obstacle avoidance parameter template library for various task types; S3. Based on the real-time acquired task context information, select the most matching obstacle avoidance strategy template from the parameter library; or, based on context features, generate a dynamic obstacle avoidance parameter set through inference / interpolation to complete the adaptive configuration of the obstacle avoidance strategy.

[0010] Based on the above scheme and as a preferred embodiment of the above scheme, the specific steps for obtaining the obstacle avoidance parameter set in step S3 are as follows: S3.1 The system periodically or event-triggeredly receives the task context state vector; S3.2. Based on the current task type, search the preset strategy parameter library for the most matching strategy template. If the matching degree between the current task and a certain standard template is greater than or equal to the threshold, then directly load the corresponding obstacle avoidance parameter group as the current obstacle avoidance control strategy. S3.3 If the current task context fails to fully match a standard template, the system will trigger an interpolation generation mechanism or a strategy fusion mechanism to dynamically calculate the obstacle avoidance parameter set that best suits the current state.

[0011] Based on the above scheme and as a preferred embodiment of the above scheme: In step S3.3, the dynamic calculation steps of the obstacle avoidance parameter group are as follows: S3.3.1 Adjacent Template Selection: Assume the flight speed in the current task context is... The system first selects and Two adjacent policy templates, A and B, are used as interpolation bases, i.e., the flight speed of template A is... The flight speed of template B is , < < ; The parameter set of template A is { , , , , , }; The parameter set of template B is { , , , , , }; The parameter set of the strategy corresponding to the current task is { , , , , , }; S3.3.2, Let the velocity interpolation weighting coefficient be... The calculation method is as follows: ; S3.3.3 Obstacle avoidance parameter interpolation calculation: ; ; = ; , , Then, a hybrid strategy is generated using soft mapping or conditional selection.

[0012] Based on the above scheme and as a preferred option: In step S3, after obtaining the initial obstacle avoidance parameter set, the system also needs to continuously monitor the dynamic changes of the task context, and combine the parameters defined in each template to determine whether the current context has changed significantly. If the change exceeds the threshold, the system immediately re-executes template matching or interpolation inference to update the obstacle avoidance parameters. If it is in a stable range, the current obstacle avoidance strategy is maintained to avoid instability caused by frequent switching.

[0013] Based on the above scheme and as the preferred scheme: the final generated obstacle avoidance parameter set is transmitted in real time to the flight controller and obstacle avoidance module, serving as the core control basis for path planning, speed adjustment and action execution, to achieve closed-loop control.

[0014] Based on the above scheme and as a preferred option: in step S1, the mission context information includes mission type, flight speed, flight altitude, ambient brightness, and weather conditions.

[0015] Based on the above scheme and as a preferred embodiment of the above scheme: In step S1, the acquisition and modeling of task context information includes the following steps: S1.1 Obtain the type identifier of the task currently being performed by the UAV through the task planning system or task scheduling platform, namely: A: low-speed fine inspection; B: medium-speed routine monitoring; C: high-speed material transportation; D: danger zone reconnaissance. S1.2. Real-time acquisition of the UAV's current flight speed, flight altitude, heading angle, and GPS position to quantify the risk level and operational window of the flight status; S1.3. Combine sensor systems to obtain external environmental state parameters, including: ambient brightness, meteorological conditions, and airspace congestion. S1.4. The above multi-source information is linearly normalized and encoded to form a task context state vector C = [T,v, h, L, W], where T is the task type code, v is the flight speed, h is the flight altitude, L is the ambient brightness index, and W is the weather category.

[0016] Based on the above scheme and as a preferred embodiment of the above scheme: the obstacle avoidance parameter template in step S2 consists of a set of obstacle avoidance control parameters, and each set of parameters includes at least the following dimensions: Obstacle avoidance trigger distance The minimum permissible distance between an obstacle and the drone; if the distance is less than this value, obstacle avoidance will be triggered. Safety Buffer Increased warning distance beyond the edge of the obstacle to avoid delayed response; Path replanning weights Prioritize the replanning module in the trade-off between local obstacle avoidance and global path optimization; Obstacle avoidance strategy types This includes deceleration and avoidance, obstacle avoidance, hovering and waiting, and emergency climbing; Perceived credibility factor Used to adjust the level of trust in data from different sensors; Dynamically adjust threshold : The threshold for the magnitude of context fluctuations that trigger policy updates; The above parameters can be stored as structured data tables, strategy graphs, or configuration files, supporting fast lookup and matching.

[0017] Based on the above scheme and as a preferred option: in step S2, it is necessary to construct a mapping rule from task type to obstacle avoidance strategy and set an initial matching relationship between task semantic tags and corresponding strategy templates.

[0018] Based on the above scheme and as a preferred embodiment of the above scheme: In step S2, the obstacle avoidance parameter template library can be obtained by combining rule setting and machine learning training methods: Rule setting refers to the process by which experts set parameter templates based on experience values, mission scenario simulation results, and flight safety standards. Machine learning training involves recording task performance, obstacle avoidance effectiveness, and context changes during system operation, and adjusting existing parameter templates based on feedback signals, or generating new templates to be added to the strategy library.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention introduces "task context" as a driving factor for obstacle avoidance strategy, which can match or generate the most suitable obstacle avoidance parameter set in real time according to different task types, realizing a fundamental transformation from "fixed parameters" to "dynamic scheduling", effectively improving the generalization ability and operational safety of UAVs in complex and diverse task environments.

[0020] It has a preset obstacle avoidance parameter template library for different mission types and can generate suitable obstacle avoidance strategies through matching or interpolation. This avoids frequent deceleration and misjudgment caused by fixed distance obstacle avoidance strategies in high-speed transportation and other missions, reduces unnecessary flight interruptions and path deviations, and significantly improves the smoothness and efficiency of mission execution.

[0021] By employing mechanisms such as linear interpolation, the obstacle avoidance parameters are made to transition smoothly and continuously, avoiding the control abrupt changes or system oscillations that are prone to occur when switching traditional fixed threshold strategies. This ensures that the UAV has good controllability and stability during flight, and is especially suitable for high-dynamic flight missions.

[0022] Context modeling, policy interpolation, and control mechanisms have good scalability, can be flexibly adapted to different types of UAV platforms and diverse mission requirements, and can be integrated into existing flight control systems without large-scale structural reconstruction, possessing strong engineering implementation capabilities and commercialization potential.

[0023] The constructed multi-task type standard obstacle avoidance strategy parameter template library supports both rule setting and data-driven approaches. The system can directly match highly adaptable strategies based on task tags, improving the efficiency and applicability of strategy generation. The system can continuously monitor the dynamic changes in the task context. When the change exceeds the threshold, the obstacle avoidance parameters are updated in a timely manner. When the change is in a stable range, the current strategy is maintained. This ensures the timeliness of the strategy and avoids instability caused by frequent switching, realizing dynamic optimization and efficient updating of obstacle avoidance strategies. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0025] The task context-aware UAV adaptive obstacle avoidance control method described in this embodiment includes the following steps: S1. Collect and parse the context information of the task currently being performed by the UAV; the task context includes, but is not limited to, parameters such as task type (e.g., low-speed inspection, high-speed transportation, emergency response), flight speed, flight altitude, ambient brightness, and weather conditions.

[0026] Specifically, the acquisition and modeling of task context information includes the following steps: S1.1 Task type identification: The task type identifier currently being performed by the UAV is obtained through a task planning system or task scheduling platform, namely: Category A: Low-speed precision inspection (such as nighttime pipeline inspection, power line patrol); Category B: Medium-speed routine monitoring (such as daytime area patrol); Category C: High-speed material transportation (such as emergency relief package delivery); and Category D: Danger zone reconnaissance (such as emergency monitoring in disaster areas). The task type, as a core semantic tag, is the dominant factor in the task context awareness of this invention.

[0027] S1.2 Flight status parameter acquisition: The system collects dynamic physical quantities such as the drone's current flight speed (v), flight altitude (h), heading angle, and GPS position in real time. These parameters can be directly obtained through the flight control system interface and are used to quantify the risk level and operable window of the flight status.

[0028] S1.3, Environmental perception parameter fusion: External environmental parameters are obtained by combining sensor systems (such as light sensors, barometers, temperature and humidity sensors), including: ambient brightness (used to assess the availability of visual sensors), meteorological conditions (such as wind speed and direction, rain and fog intensity, which affect flight stability and sensor reliability), and airspace congestion (combined with flight control systems or radar feedback to assess surrounding flight risks).

[0029] S1.4, Context state vector construction: The aforementioned multi-source information is linearly normalized and encoded to form a task context state vector C = [T, v, h, L, W], where T is the task type encoding, v is the speed, h is the altitude, L is the brightness index, and W is the weather category. This state vector can be used as input to an obstacle avoidance strategy selector to achieve strategy mapping under multiple tasks and scenarios.

[0030] S2. A pre-defined obstacle avoidance parameter template library corresponding to various task types; the templates include obstacle avoidance trigger distance, safety buffer, path replanning weight coefficient, etc. This parameter library can be obtained by combining rule setting and machine learning training methods.

[0031] S2.1 Parameter Template Structure Design Each policy template in the policy library corresponds to a typical task type, and its core consists of a set of obstacle avoidance control parameters. Each set of parameters includes at least the following dimensions (which can be expanded according to specific applications): Obstacle avoidance trigger distance ( ): The minimum permissible distance between the obstacle and the drone; if it is less than this value, obstacle avoidance action will be triggered. Safety buffer ( ): An increased warning distance beyond the edge of an obstacle to avoid delayed response; Path replanning weights ( Prioritize the replanning module in the trade-off between local obstacle avoidance and global path selection; Obstacle avoidance strategy types ( Examples include: deceleration and avoidance, obstacle avoidance, hovering and waiting, and emergency climbing. Perceived credibility factor ( ): Used to adjust the level of trust in data from different sensors (e.g., downweighted visual data when night vision deteriorates). Dynamically adjust threshold ( ): The threshold for the context fluctuation amplitude that triggers a policy update. These parameters can be stored as structured data tables, policy graphs, or configuration files, supporting fast lookup and matching.

[0032] S2.2, Task-Strategy Mapping Relationship Setting Based on typical task types, construct task-to-policy mapping rules and set the initial matching relationship between task semantic tags (such as "nighttime low-speed inspection" and "daytime high-speed logistics") and corresponding policy templates. S2.3 Strategy Library Construction Method The strategy parameter library is based on a rule-based manual setting method: experts set parameter templates based on experience values, mission scenario simulation results, and flight safety standards, which is suitable for early deployment and critical scenario support.

[0033] S2.4 Strategy Template Extension and Update Mechanism To enhance the applicability and self-evolution capabilities of the strategy library, this embodiment supports online expansion and dynamic updates of templates. During operation, the system can record task performance, obstacle avoidance effectiveness, and contextual changes, and adjust existing parameter templates based on feedback signals, or generate new templates to add to the strategy library.

[0034] By constructing a task-aware obstacle avoidance strategy parameter library, this embodiment achieves rapid mapping and flexible adjustment between the three-layer structure of "task-strategy-parameter", solving the technical bottleneck that the traditional fixed obstacle avoidance threshold method cannot adapt to multi-task scenarios.

[0035] S3. Based on the real-time acquired task context information, select the most matching obstacle avoidance strategy template from the parameter library; or, based on context features, generate a dynamic obstacle avoidance parameter set through inference / interpolation to complete the adaptive configuration of the obstacle avoidance strategy.

[0036] The adaptive scheduling mechanism includes the following core steps: S3.1, Context State Vector Input The system periodically or event-triggeredly receives the task context state vector C = [T, v, h, L, W]. This vector consists of task type (T), flight speed (v), flight altitude (h), ambient light (L), weather information (W), etc., and is the core input of the scheduling engine.

[0037] S3.2 Strategy Template Matching Mechanism The scheduling module first searches for the most matching policy template in the preset policy parameter library based on the current task type (T). If the current task is highly consistent with a certain standard template (e.g., matching degree ≥ threshold), then the corresponding obstacle avoidance parameter group is directly loaded. , , , , , This serves as the current obstacle avoidance control strategy. For example, when the task type is "nighttime low-speed inspection" and the brightness is low, the speed is slow, and the visual confidence is low, the system will directly select the high-redundancy obstacle avoidance strategy from the corresponding template.

[0038] S3.3 Context-driven parameter interpolation / inference mechanism If the current task context does not perfectly match a standard template (e.g., flight speed or altitude falls between multiple templates), the system triggers an interpolation generation mechanism or a policy fusion mechanism to dynamically calculate the obstacle avoidance parameter set best suited to the current state. The process based on linear interpolation is as follows: S3.3.1, Selection of Adjacent Templates: Let the flight speed in the current mission context be... The system first selects and Two adjacent policy templates, A (medium speed) and B (high speed), are used as interpolation benchmarks, meaning the flight speed of template A is... The flight speed of template B is , < < ; The parameter set of template A is { , , , , ,}; The parameter set of template B is { , , , , , }; The parameter set of the strategy corresponding to the current task is { , , , , , }; S3.3.2, Let the velocity interpolation weighting coefficient be... The calculation method is as follows: ; S3.3.3 Obstacle avoidance parameter interpolation calculation: For continuous parameters (e.g.) , , Interpolate as follows: ; ; = ; For discrete parameters (such as) , , Then, a soft mapping or conditional selection method is used to generate a hybrid strategy. For example: If α∈ [0, 0.33] → = Low; if α∈ (0.33, 0.66] → = in; if α∈(0.66, 1] → = High.

[0039] like < 0.5, Then the inheritance strategy template A ;like ≥ 0.5, Then enable strategy template B. .

[0040] It is determined by L_norm, which is the normalized value of the luminance. If L_norm < 0.2 → = Low; if L_norm ∈ [0.2, 0.6) → = In the middle; if L_norm ≥ 0.6 → = High.

[0041] Example of interpolated output: by Assumption: (High-speed template A); (High-speed template B); , , , calculate: Similarly, other parameter values ​​can be calculated to ultimately generate a dynamic obstacle avoidance parameter set: { , , }

[0042] S3.3.4 Obstacle Avoidance Strategy Update Trigger Mechanism The system continuously monitors dynamic changes in the task context, combining this with the definitions in each template. (Dynamically adjust threshold) parameters to determine whether the current context has changed significantly: if the change exceeds the threshold (e.g., brightness changes by 10%, speed increases by 15%), the system immediately re-executes template matching or interpolation inference and updates the obstacle avoidance parameters; if it is in a stable range, the current obstacle avoidance strategy remains unchanged to avoid instability caused by frequent switching.

[0043] S3.3.5 Strategy Parameter Issuance and Obstacle Avoidance Control Execution The generated obstacle avoidance parameter set is transmitted in real time to the flight controller and obstacle avoidance module, serving as the core control basis for path planning, speed adjustment and action execution, thus achieving closed-loop control.

[0044] Compared with the existing traditional UAV obstacle avoidance control methods that use fixed obstacle avoidance thresholds, this method introduces "task context" as a driving factor for obstacle avoidance strategy for the first time. It can match or generate the most suitable obstacle avoidance parameter set in real time according to different task types (such as low-speed inspection, high-speed transportation, night reconnaissance, etc.), realizing a fundamental transformation from "fixed parameters" to "dynamic scheduling". This effectively improves the generalization ability and operational safety of UAVs in complex and diverse task environments.

[0045] Compared to the control abrupt changes or system oscillations that are easily caused by switching traditional fixed threshold strategies, this embodiment achieves continuous transition and smooth adjustment of obstacle avoidance parameters through mechanisms such as linear interpolation, ensuring that the strategy has good controllability and stability during flight, and is especially suitable for high-dynamic flight missions.

[0046] The context modeling, strategy interpolation, and control mechanisms in this embodiment can be flexibly extended to adapt to different types of UAV platforms and mission requirements. They can also be integrated into existing flight control systems without large-scale structural reconstruction, demonstrating good engineering implementation capabilities and commercial potential.

[0047] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for UAV adaptive obstacle avoidance control based on task context awareness, characterized in that: Includes the following steps: S1. Collect and parse the context information of the task currently being performed by the UAV; S2. Pre-set obstacle avoidance parameter template library for various task types; S3. Based on the real-time acquired task context information, select the most matching obstacle avoidance strategy template from the parameter library; or, based on context features, generate a dynamic obstacle avoidance parameter set through inference / interpolation to complete the adaptive configuration of the obstacle avoidance strategy.

2. The task context-aware adaptive obstacle avoidance control method for unmanned aerial vehicles as described in claim 1, characterized in that: The specific steps for obtaining the obstacle avoidance parameter set in step S3 are as follows: S3.1 The system periodically or event-triggeredly receives the task context state vector; S3.

2. Based on the current task type, search the preset strategy parameter library for the most matching strategy template. If the matching degree between the current task and a certain standard template is greater than or equal to the threshold, then directly load the corresponding obstacle avoidance parameter group as the current obstacle avoidance control strategy. S3.3 If the current task context fails to fully match a standard template, the system will trigger an interpolation generation mechanism or a strategy fusion mechanism to dynamically calculate the obstacle avoidance parameter set that best suits the current state. 3.The method of claim 2, wherein: In step S3.3, the dynamic calculation steps for the obstacle avoidance parameter set are as follows: S3.3.1, adjacent template selection: assuming the flight speed in the current task context is , the system first selects two strategy templates A and B adjacent to as interpolation reference, i.e. the flight speed of template A is , and the flight speed of template B is , < < ; The parameter set of template A is { , , , , , } ; The parameter set of template B is { , , , , , } ; The parameter group corresponding to the strategy of the current task is { , , , , , }; S3.3.2, set the speed interpolation weight coefficient as The calculation is as follows: ； S3.3.3 Obstacle avoidance parameter interpolation calculation: ； ； = ； , , Soft mapping or conditional selection is used to generate the hybrid strategy.

4. The task context-aware adaptive obstacle avoidance control method for unmanned aerial vehicles as described in claim 2, characterized in that: In step S3, after obtaining the initial obstacle avoidance parameter set, the system also needs to continuously monitor the dynamic changes of the task context and, in conjunction with the parameters defined in each template, determine whether the current context has changed significantly: if the change exceeds the threshold, the system immediately re-executes template matching or interpolation inference to update the obstacle avoidance parameters; if it is in a stable range, the current obstacle avoidance strategy is maintained to avoid instability caused by frequent switching. 5.The method of claim 2, wherein: The final obstacle avoidance parameter set is transmitted in real time to the flight controller and obstacle avoidance module, serving as the core control basis for path planning, speed adjustment and action execution, thus achieving closed-loop control. 6.The method of claim 1, wherein the method further comprises: determining a task context of the UAV; and determining the UAV obstacle avoidance control strategy based on the task context. In step S1, the mission context information includes mission type, flight speed, flight altitude, ambient brightness, and weather conditions.

7. The method of claim 6, wherein the method further comprises: determining a task context of the UAV; and determining the UAV's flight path based on the task context. In step S1, the acquisition and modeling of task context information includes the following steps: S1.1 Obtain the type identifier of the task currently being performed by the UAV through the task planning system or task scheduling platform, namely: A: low-speed fine inspection; B: medium-speed routine monitoring; C: high-speed material transportation; D: danger zone reconnaissance. S1.

2. Real-time acquisition of the UAV's current flight speed, flight altitude, heading angle, and GPS position to quantify the risk level and operational window of the flight status; S1.

3. Combine sensor systems to obtain external environmental state parameters, including: ambient brightness, meteorological conditions, and airspace congestion. S1.

4. The above multi-source information is linearly normalized and encoded to form a task context state vector C = [T, v, h, L, W], where T is the task type code, v is the flight speed, h is the flight altitude, L is the ambient brightness index, and W is the weather category.

8. The task context-aware adaptive obstacle avoidance control method for unmanned aerial vehicles as described in claim 1, characterized in that: The obstacle avoidance parameter template in step S2 consists of a set of obstacle avoidance control parameters, and each set of parameters includes at least the following dimensions: Obstacle avoidance trigger distance : minimum allowed distance between obstacle and UAV, below which obstacle avoidance action is triggered Safety buffer : increased warning distance beyond the edge of the obstacle proximity for avoiding delayed response; path replanning weight : priority given to the replanning module in local obstacle avoidance versus global path tradeoff Obstacle avoidance strategy types This includes deceleration and avoidance, obstacle avoidance, hovering and waiting, and emergency climbing; Perceived credibility factor Used to adjust the level of trust in data from different sensors; Dynamically adjust threshold : The threshold for the magnitude of context fluctuations that trigger policy updates; The above parameters can be stored as structured data tables, strategy graphs, or configuration files, supporting fast lookup and matching.

9. The task context-aware adaptive obstacle avoidance control method for unmanned aerial vehicles as described in claim 8, characterized in that: In step S2, it is necessary to construct the mapping rules from task type to obstacle avoidance strategy and set the initial matching relationship between task semantic tags and corresponding strategy templates.

10. The task context-aware adaptive obstacle avoidance control method for unmanned aerial vehicles as described in claim 8, characterized in that: In step S2, the obstacle avoidance parameter template library can be obtained by combining rule setting and machine learning training methods: Rule setting refers to the process by which experts set parameter templates based on experience values, mission scenario simulation results, and flight safety standards. Machine learning training involves recording task performance, obstacle avoidance effectiveness, and context changes during system operation, and adjusting existing parameter templates based on feedback signals, or generating new templates to be added to the strategy library.

Images