Unmanned aerial vehicle dynamic take-off and landing management method and system

Abstract

The application discloses a dynamic take-off and landing management method and system of an unmanned aerial vehicle, relates to the technical field of low-altitude ground traffic cooperative management and control, receives a take-off and landing request sent by an unmanned aerial vehicle, defines the unmanned aerial vehicle, an intersection traffic signal lamp control system and ground traffic flow aggregated by a digital twin model as game participants, and initializes respective strategy spaces; in a digital twin environment, an integrated utility function under different strategy combinations is calculated through iteration to seek a Nash equilibrium strategy combination maximizing the utility of all participants, and the Nash equilibrium strategy combination defines a take-off and landing timing, path and timing coordination scheme of the signal lamp of the unmanned aerial vehicle. The application realizes deep coordination between the take-off and landing of the unmanned aerial vehicle and ground traffic, accurately matches the take-off and landing demand of the unmanned aerial vehicle and the running state of the ground traffic by constructing a dynamic risk field and a game decision mechanism, effectively avoids air-ground traffic conflicts, and improves the safety of the take-off and landing operation of the unmanned aerial vehicle.

Description

Technical Field

[0001] This invention relates to the field of low-altitude ground traffic collaborative management and control technology, and in particular to a method and system for dynamic take-off and landing management of unmanned aerial vehicles (UAVs). Background Technology

[0002] The rapid development of the low-altitude economy has driven the continuous expansion of drone applications in urban infrastructure inspection, traffic emergency response, and public safety, leading to a growing demand for drone control in urban airspace. Currently, drone take-off and landing control in urban environments faces numerous technical challenges. Existing drone take-off and landing sites are scarce, with many relying on independently constructed facilities, resulting in high costs and difficulty in achieving efficient integration with urban road spaces. Furthermore, the coordination between airspace and ground traffic is inefficient, with drone take-off and landing operations lacking effective linkage with ground vehicle and pedestrian traffic, potentially causing traffic conflicts and safety hazards. Electronic fence control often uses fixed boundaries, failing to adapt to dynamic changes in ground traffic flow, resulting in insufficient control precision and flexibility, and making it difficult to match the complex and ever-changing traffic scenarios of urban roads.

[0003] Meanwhile, existing drone take-off and landing control methods lack the ability to accurately predict and assess the trajectories of ground traffic participants. Drone take-off and landing timing and path planning are mostly statically preset, failing to fully consider the real-time status of ground traffic flow. Furthermore, no collaborative decision-making mechanism has been established between drones, traffic signal control systems, and ground traffic flow. This results in drone take-off and landing operations interfering with ground traffic, and the safety and efficiency of the drones themselves are difficult to guarantee. In addition, the existing control systems suffer from insufficient coordination between perception, computation, and execution, and the utilization efficiency of roadside perception data is low. This prevents the realization of dynamic, intelligent, and precise drone take-off and landing control, hindering the large-scale and safe application of drones in urban road scenarios. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for dynamic takeoff and landing control of unmanned aerial vehicles (UAVs). The technical solution adopted is as follows: The dynamic take-off and landing control method for unmanned aerial vehicles (UAVs) includes the following steps: Step 1: Real-time access to multi-source roadside sensing data, construct a real-time four-dimensional digital twin model containing dynamic traffic participants, and construct a dynamic risk field in the spatiotemporal dimension based on the predicted trajectories of each participant. Step 2: Receive take-off and landing requests from the drone, and define the drone, the intersection traffic light control system, and the ground traffic flow aggregated by the digital twin model as game participants, and initialize their respective strategy spaces; Step 3: In the digital twin environment, by iteratively calculating the comprehensive utility function under different strategy combinations, we seek the Nash equilibrium strategy combination that maximizes the utility of all participants. The Nash equilibrium strategy combination defines the take-off and landing timing, path, and timing coordination scheme of the UAV and traffic lights. Step 4: Based on the solved Nash equilibrium strategy combination, control commands are sent to the drone and traffic lights respectively to dynamically adjust the drone's flight path and the electronic fence boundary, thereby achieving coordinated optimization of air and ground resources.

[0005] Optionally, the construction of the dynamic risk field in step 1 specifically includes: using a trajectory prediction model based on social generative adversarial networks or Transformer to predict the trajectories of pedestrians and non-motorized vehicles in the next 3-5 seconds and their probability distribution; based on the predicted trajectory, constructing a Gaussian risk field for each dynamic obstacle, and superimposing all individual risk fields to form a dynamic risk heat map of the intersection that changes continuously with time and space, serving as the underlying mathematical model for the dynamic adjustment of the electronic fence.

[0006] Optional, the drone's strategy space is as follows: Define the policy space of the UAV U as , This indicates that you should immediately descend along the nominal path. Indicates hovering and waiting After time, it descends along the dynamic programming path. This indicates that the aircraft has climbed up and is about to retake the aircraft to the waiting airspace. Define the policy space of traffic light T as , This indicates that the current signal timing scheme will be maintained. This indicates that the red light phase will be extended based on the current cycle. , This indicates that the red light phase of the next cycle will be activated ahead of schedule. ; Define the policy space of ground traffic flow G as , , These correspond to the high-density traffic state, medium-density traffic state, and low-density traffic state output by the trajectory prediction model, respectively.

[0007] Optionally, the formula for calculating the comprehensive utility function in step 3 is: ; in It is the integral risk value of the drone along its take-off and landing path. It is a quantification of the impact of drones on the average speed of ground traffic flow. The extra electrical energy consumed by drones while waiting or taking off again. It is a preset priority coefficient based on task type; The Nash equilibrium solution employs iterative elimination of disadvantageous strategies or optimization algorithms based on swarm intelligence to quickly search for the strategy combination that maximizes the individual utility of all game participants within a limited strategy space, serving as the final collaborative control scheme.

[0008] Optionally, in step 4, the dynamic adjustment of the UAV flight path adopts the fast squared forward algorithm. Using the dynamic risk field generated in step 1 as the cost map, a real-time flight path with the lowest cumulative risk score from the current position of the UAV to the take-off and landing point is planned. The real-time flight path is dynamically replanned as the risk field is updated.

[0009] Optionally, the dynamic adjustment of the UAV flight path in step 4 is specifically as follows: using the fast squared forward algorithm, with the dynamic risk field generated in step 1 as the cost map, a real-time flight path with the lowest cumulative risk score from the current position of the UAV to the take-off and landing point is planned. This path is dynamically replanned as the risk field is updated.

[0010] Optionally, in step 4, the boundary of the electronic fence is a dynamic contour subset of the dynamic risk field, and its spatial morphology is linked to the risk field: in the high-value area of ​​the risk field, the fence boundary contracts inward to avoid high-density traffic flow; in the low-value area of ​​the risk field, the fence boundary expands outward to provide a greater safety margin, thereby realizing the spatiotemporal dynamic adaptation of the electronic fence morphology.

[0011] Optionally, a time-space exchange mechanism can also be included. By solving the Nash equilibrium, the system makes decisions through traffic light policy adjustments to create a dedicated time window for the drone without ground traffic interference. In exchange, the drone must complete take-off and landing within the dedicated time window; otherwise, it will have to pay additional waiting energy costs. This achieves the trading and optimized allocation of airspace and ground right-of-way in the time dimension.

[0012] Optionally, step 4 of the execution and collaborative control also includes a roadside precision early warning step: based on the solved optimal take-off and landing path of the drone and the dynamic risk field, using the directional sound wave device and variable color temperature light effect device on the roadside smart light pole, an avoidance warning containing precise time and space information is issued to specific pedestrians or vehicles that are about to enter the spatiotemporal range of the drone's take-off and landing path, thereby improving the safety of control.

[0013] The UAV take-off and landing collaborative control system is used to realize the dynamic take-off and landing control method of UAVs. The system includes a sensing unit, edge computing nodes, execution unit and control platform. The sensing units are deployed on the roadside and include cameras, millimeter-wave radar, and V2X roadside units; Edge computing nodes are deployed on the roadside and communicate with sensing units to execute the method steps of claim 9; The execution unit is deployed on smart light poles or traffic signal poles, including drone take-off and landing base stations, directional sound wave early warning devices, and variable color temperature light effect devices. It is connected to edge computing nodes to execute control commands. The management and control platform is deployed in the central cloud and communicates with the sensing unit, edge computing nodes, and execution unit respectively, for global monitoring and model training.

[0014] In summary, the present invention has at least one of the following beneficial technical effects: This invention provides a dynamic take-off and landing control method and system for unmanned aerial vehicles (UAVs), enabling deep coordination between UAV take-off and landing and ground traffic. By constructing a dynamic risk field and a game-theoretic decision-making mechanism, it accurately matches the take-off and landing needs of UAVs with the operating status of ground traffic, effectively avoiding air-ground traffic conflicts and improving the safety of UAV take-off and landing operations.

[0015] Promote the efficient reuse of road ancillary facilities resources, and deploy control systems based on smart light poles and traffic signal poles. This eliminates the need for additional independent infrastructure construction and reduces the construction and operation costs of drone take-off and landing control.

[0016] It enables dynamic adaptive adjustment of electronic fences and drone flight paths, optimizes fence boundaries and take-off and landing paths based on real-time changes in ground traffic flow, improves the accuracy and flexibility of airspace control, and adapts to the complex and ever-changing traffic scenarios of urban roads.

[0017] Establish a multi-entity collaborative decision-making system that incorporates drones, traffic lights, and ground traffic flow into a unified decision-making framework. Maximize the utility of each entity through Nash equilibrium solution, taking into account both drone take-off and landing efficiency and ground traffic efficiency, thereby improving the overall efficiency of air and ground resource allocation.

[0018] This enables efficient linkage between precise roadside early warning and control execution. Relying on roadside sensing and early warning equipment, it releases precise avoidance information to traffic participants, further enhancing the safety protection capabilities of drone take-off and landing operations. At the same time, the layered deployment of edge computing nodes and the central control platform improves the system's real-time response capabilities and overall control capabilities. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the dynamic take-off and landing control method for unmanned aerial vehicles (UAVs) of the present invention. Figure 2 This is a schematic diagram of the architecture of the UAV dynamic take-off and landing control system of the present invention; Detailed Implementation

[0020] The present invention will be further described in detail below with reference to the accompanying drawings.

[0021] This invention discloses a method and system for dynamic take-off and landing control of unmanned aerial vehicles (UAVs).

[0022] Reference Figure 1 and Figure 2 Example 1, a method for dynamic take-off and landing control of unmanned aerial vehicles, includes the following steps: Step 1: Real-time access to multi-source roadside sensing data, construct a real-time four-dimensional digital twin model containing dynamic traffic participants, and construct a dynamic risk field in the spatiotemporal dimension based on the predicted trajectories of each participant. Step 2: Receive take-off and landing requests from the drone, and define the drone, the intersection traffic light control system, and the ground traffic flow aggregated by the digital twin model as game participants, and initialize their respective strategy spaces; Step 3: In the digital twin environment, by iteratively calculating the comprehensive utility function under different strategy combinations, we seek the Nash equilibrium strategy combination that maximizes the utility of all participants. The Nash equilibrium strategy combination defines the take-off and landing timing, path, and timing coordination scheme of the UAV and traffic lights. Step 4: Based on the solved Nash equilibrium strategy combination, control commands are sent to the drone and traffic lights respectively to dynamically adjust the drone's flight path and the electronic fence boundary, thereby achieving coordinated optimization of air and ground resources.

[0023] By adopting the above technical solutions, and relying on roadside perception and digital twin technology to build a decision-making foundation for air-ground collaboration, drones, traffic signal control systems, and ground traffic flow are incorporated into a game-theoretic decision-making system. By solving the game equilibrium, the optimal strategy for multi-entity collaboration is determined, and then control commands are issued based on this strategy. The drone flight path and electronic fence boundary are dynamically adapted, ultimately achieving resource synergy optimization between drone take-off and landing and ground traffic. This allows drone take-off and landing control to conform to the real-time operation status of ground traffic, achieving linkage and adaptation between air and ground control.

[0024] Example 2, step 1 of constructing a dynamic risk field specifically includes: using a trajectory prediction model based on social generative adversarial networks or Transformer to predict the trajectories of pedestrians and non-motorized vehicles in the next 3-5 seconds and their probability distribution; based on the predicted trajectories, constructing a Gaussian risk field for each dynamic obstacle, and superimposing all individual risk fields to form a dynamic risk heat map of the intersection that changes continuously with time and space, which serves as the underlying mathematical model for the dynamic adjustment of the electronic fence.

[0025] By adopting the above technical solution, an advanced trajectory prediction model is used to accurately predict the short-term movement trajectories and probability distributions of pedestrians and non-motorized vehicles. Based on this, a Gaussian risk field is constructed for each dynamic traffic participant. By superimposing all individual risk fields, a dynamic heat map reflecting the spatiotemporal risk distribution at the intersection is formed. This heat map can quantify the risk level at different spatiotemporal locations of the intersection, providing a quantitative mathematical basis for the dynamic adjustment of the electronic fence, thus giving the electronic fence adjustment data support and a scientific model foundation.

[0026] Example 3, the strategy space of the UAV, specifically: Define the policy space of the UAV U as , This indicates that you should immediately descend along the nominal path. Indicates hovering and waiting After time, it descends along the dynamic programming path. This indicates that the aircraft has climbed up and is about to retake the aircraft to the waiting airspace. Define the policy space of traffic light T as , This indicates that the current signal timing scheme will be maintained. This indicates that the red light phase will be extended based on the current cycle. , This indicates that the red light phase of the next cycle will be activated ahead of schedule. ; Define the policy space of ground traffic flow G as , , These correspond to the high-density traffic state, medium-density traffic state, and low-density traffic state output by the trajectory prediction model, respectively.

[0027] By adopting the above technical solutions, differentiated strategy selection ranges are set for three core game participants: drone takeoff and landing, traffic light timing, and ground traffic flow. Three takeoff and landing strategies are planned for drones: immediate descent, hovering and waiting before descent, and pull-up and go-around waiting. Three timing strategies are set for traffic lights: maintaining timing, extending red lights, and starting red lights early. Three traffic flow state strategies are divided for ground traffic flow: high, medium, and low. By clarifying the strategy space of each participant, a decision-making framework for air-ground collaborative game is constructed, defining clear strategy boundaries for subsequent equilibrium solutions.

[0028] Example 4, the formula for calculating the comprehensive utility function in step 3 is: ; in It is the integral risk value of the drone along its take-off and landing path. It is a quantification of the impact of drones on the average speed of ground traffic flow. The extra electrical energy consumed by drones while waiting or taking off again. It is a preset priority coefficient based on task type; The Nash equilibrium solution employs iterative elimination of disadvantageous strategies or optimization algorithms based on swarm intelligence to quickly search for the strategy combination that maximizes the individual utility of all game participants within a limited strategy space, serving as the final collaborative control scheme.

[0029] By employing the aforementioned technical solution, a comprehensive utility function is constructed to quantitatively evaluate the overall payoff of each participant in the game. This function integrates multiple dimensions of indicators, including the path risk of drone takeoff and landing, the degree of interference with ground traffic, its own energy consumption cost, and the urgency of the mission. Weighting is used to comprehensively consider each indicator. Simultaneously, an efficient Nash equilibrium solution algorithm is employed to iteratively eliminate inferior strategies within a limited strategy space or to optimize through swarm intelligence, quickly searching for the strategy combination that maximizes the utility of all participants. This determines the optimal collaborative management scheme, taking into account the interests of all participating entities.

[0030] In Example 5, the fast squared-moving algorithm is used to dynamically adjust the UAV flight path in step 4. The dynamic risk field generated in step 1 is used as the cost map to plan a real-time flight path with the lowest cumulative risk score from the current position of the UAV to the take-off and landing point. The real-time flight path is dynamically replanned as the risk field is updated.

[0031] By adopting the above technical solution, a dynamic risk field is used as the cost map for path planning, transforming risk values ​​into cost indicators for path planning. A fast squared-movement algorithm is employed for real-time planning of the UAV's flight path. The algorithm automatically searches for the path with the lowest cumulative risk score from the UAV's current position to the take-off and landing point. Simultaneously, with the real-time update of the dynamic risk field, the flight path is dynamically replanned to ensure that the UAV always flies along the path with the lowest risk, thus improving the safety of take-off and landing.

[0032] In Example 6, the specific steps of dynamically adjusting the UAV flight path in step 4 are as follows: using the fast squared forward algorithm, with the dynamic risk field generated in step 1 as the cost map, a real-time flight path with the lowest cumulative risk score from the current position of the UAV to the take-off and landing point is planned. This path is dynamically replanned as the risk field is updated.

[0033] By adopting the above technical solution, the dynamic risk field is used as the core cost basis for path planning. Utilizing the path search characteristics of the fast squared-movement algorithm, the algorithm plans the flight path of the UAV from its current position to the take-off and landing point with the objective of minimizing the cumulative risk integral. This algorithm can adapt to the spatiotemporal changes of the dynamic risk field. When the risk field is updated due to changes in ground traffic conditions, the UAV flight path is simultaneously recalculated and planned, ensuring the real-time optimality and safety of the flight path.

[0034] In Example 7, the boundary of the electronic fence in step 4 is a dynamic contour subset of the dynamic risk field, and its spatial morphology is linked with the risk field: in the high-value area of ​​the risk field, the fence boundary contracts inward to avoid high-density traffic flow; in the low-value area of ​​the risk field, the fence boundary expands outward to provide a greater safety margin, thereby realizing the spatiotemporal dynamic self-adaptation of the electronic fence morphology.

[0035] By adopting the above technical solution, the electronic fence boundary is deeply correlated with the dynamic risk field. The fence boundary is defined as a dynamic contour subset of the dynamic risk field, enabling the spatial morphology of the fence to adaptively adjust with changes in the risk field. In areas with dense ground traffic and high risk field values, the fence boundary is controlled to contract inward to avoid high-risk areas; in areas with sparse ground traffic and low risk field values, the fence boundary is controlled to expand outward to extend the safe space for drone take-off and landing, achieving dynamic adaptation of the electronic fence's spatiotemporal morphology.

[0036] Example 8 also includes a time-space exchange mechanism. By solving the Nash equilibrium, the system makes decisions to create a dedicated time window for the UAV without ground traffic interference through traffic light strategy adjustments. In exchange, the UAV must complete take-off and landing within the dedicated time window, otherwise it will have to pay additional waiting energy costs. This realizes the trading and optimized allocation of airspace and ground right-of-way in the time dimension.

[0037] By adopting the above technical solution and based on the game theory solution of Nash equilibrium, a dedicated take-off and landing time window without ground traffic interference is allocated for drones by adjusting the traffic light timing strategy, thereby realizing the temporary allocation of right-of-way in the time dimension. Simultaneously, an energy consumption constraint mechanism is set up: if a drone fails to complete take-off or landing within the dedicated time window, additional waiting energy costs will be incurred. This establishes a time exchange rule for airspace and ground right-of-way, achieving optimized allocation of right-of-way in the time dimension, balancing the take-off and landing needs of drones with the efficiency of ground traffic.

[0038] Example 9, Step 4 of the execution and collaborative control also includes a roadside precision early warning step: Based on the solved optimal take-off and landing path of the UAV and the dynamic risk field, the directional sound wave device and variable color temperature light effect device on the roadside smart light pole are used to issue avoidance warnings containing precise time and space information to specific pedestrians or vehicles that are about to enter the spatiotemporal range of the UAV take-off and landing path, thereby improving the safety of control.

[0039] By employing the aforementioned technical solutions, based on the optimal take-off and landing path and dynamic risk field obtained from the solution, pedestrians and vehicles about to enter the spatiotemporal range of the drone's take-off and landing path can be accurately identified. Using directional sound waves and variable color temperature light effects devices mounted on smart roadside light poles, avoidance warnings containing precise spatiotemporal information are issued to target traffic participants, achieving directional and accurate transmission of warning information. This allows traffic participants to take avoidance actions in advance, reducing conflicts between drone take-off and landing and ground traffic, and improving the safety of the control process.

[0040] Example 10: Unmanned Aerial Vehicle (UAV) Take-off and Landing Cooperative Control System, used to implement a dynamic take-off and landing control method for UAVs. The system includes a sensing unit, an edge computing node, an execution unit, and a control platform. The sensing units are deployed on the roadside and include cameras, millimeter-wave radar, and V2X roadside units; Edge computing nodes are deployed on the roadside and communicate with sensing units to execute the method steps of claim 9; The execution unit is deployed on smart light poles or traffic signal poles, including drone take-off and landing base stations, directional sound wave early warning devices, and variable color temperature light effect devices. It is connected to edge computing nodes to execute control commands. The management and control platform is deployed in the central cloud and communicates with the sensing unit, edge computing nodes, and execution unit respectively, for global monitoring and model training.

[0041] By adopting the above technical solutions and a layered deployment system architecture, the roadside perception unit achieves comprehensive collection of multi-source traffic data, providing basic data support for control and management decisions; the roadside edge computing nodes complete core computing tasks such as data processing, game theory solving, and path planning nearby, improving the system's real-time response capability; the execution unit deployed on road ancillary facilities receives control commands and completes actual control actions such as drone take-off and landing and early warning issuance; the central cloud control platform achieves global monitoring of the entire system, while also conducting model training and optimization to provide better algorithm models for front-end control. All units work together to realize the full-process implementation of dynamic drone take-off and landing control.

[0042] The following specific embodiments illustrate the implementation principle of the present invention: This application focuses on the emergency take-off and landing control of drones at a key urban intersection. The intersection is equipped with roadside infrastructure such as traffic signal poles and smart light poles, along with sensing devices including cameras, millimeter-wave radar, and V2X roadside units, as well as execution equipment such as drone take-off and landing base stations, directional acoustic warning devices, and variable color temperature lighting devices. Edge computing nodes are set up on the roadside, and a central cloud management platform is deployed in the backend to achieve dynamic, end-to-end control of drone take-off and landing at the intersection. The specific implementation process is as follows: Constructing a four-dimensional digital twin model and dynamic risk field for the intersection: The roadside perception unit's cameras, millimeter-wave radar, and V2X roadside units collect multi-source perception data in real time, including the position, speed, and direction of movement of motor vehicles, pedestrians, and non-motorized vehicles within the intersection. This data is then synchronously transmitted to the roadside edge computing node. Based on this data, the edge computing node constructs a real-time four-dimensional digital twin model of the intersection, encompassing all dynamic traffic participants, thus reconstructing the spatiotemporal distribution and movement status of these participants.

[0043] Meanwhile, the edge computing nodes use the Transformer trajectory prediction model to accurately predict the movement trajectory and probability distribution of pedestrians and non-motorized vehicles in the next 4 seconds. Based on the predicted trajectory, a Gaussian risk field is constructed for each dynamic traffic participant. All individual Gaussian risk fields are superimposed to form a dynamic risk heat map of the intersection that changes continuously with time and space. This heat map serves as the underlying mathematical model for the dynamic adjustment of the electronic fence, quantifying the risk level at different temporal and spatial locations of the intersection.

[0044] Define the game participants and initialize the policy space: When an emergency inspection drone sends a landing request to the control system at the intersection, the edge computing node defines the drone, the intersection traffic light control system, and the ground traffic flow aggregated by the digital twin model as three types of game participants, and initializes the policy space of each participant: The strategic space of drones , To descend immediately along the nominal path, To hover and wait for 10 seconds before descending along the dynamic programming path, A waiting airspace was pre-defined to allow the aircraft to ascend and return to the intersection; Traffic signal strategy space , To maintain the current signal timing scheme, To extend the red light phase by 5 seconds based on the current cycle, To advance the start of the next red light phase by 5 seconds; Strategy space of ground traffic flow Based on the trajectory prediction model, the current surface traffic flow at this intersection is in a medium-density state, that is... .

[0045] Solve for the Nash equilibrium strategy combination: In a digital twin environment, edge computing nodes construct a comprehensive utility function and iteratively calculate the utility values ​​under different policy combinations. The formula for the comprehensive utility function is as follows: ;in , , , These are the weighting coefficients. The integral risk value along the drone's landing path. This is a quantification of the impact of drones on the average speed of ground traffic flow. The extra electrical energy consumed by drones while waiting or taking off again. The priority coefficient for this UAV emergency inspection mission is set to a higher value in this embodiment.

[0046] The Nash equilibrium solution employs an iterative strategy elimination algorithm, rapidly searching within a finite strategy space to ultimately obtain the Nash equilibrium strategy combination that maximizes the utility of all game participants, as { , , The strategy involves the drone hovering for 10 seconds before descending along a dynamically planned path, the traffic lights extending the red phase by 5 seconds based on the current cycle, and ground traffic maintaining a medium-density flow. This combination of strategies clarifies the drone's landing timing, flight path, and traffic light timing coordination scheme.

[0047] Dynamically adjust flight path and electronic fence boundary: The edge computing node sends control commands to the drone and traffic lights respectively based on the solution of the Nash equilibrium strategy combination. At the same time, it initiates a time-space exchange mechanism. The traffic lights execute the command to extend the red light phase by 5 seconds to create a dedicated landing time window for the drone without ground traffic interference. It is also set that the drone must complete the landing within this dedicated time window and the subsequent 5-second buffer period, otherwise additional waiting energy costs will be incurred.

[0048] During the drone hovering and waiting phase, the edge computing node uses the fast squared-step algorithm with a dynamic risk field as the cost map to plan the real-time flight path with the lowest cumulative risk score from the drone's current hovering position to the take-off and landing base station on the traffic signal pole. This path will be dynamically replanned as the dynamic risk field is updated in real time.

[0049] Meanwhile, the boundary of the electronic fence is defined as a subset of the dynamic contour lines of the dynamic risk field, so as to realize the spatiotemporal dynamic adaptation of the electronic fence form: in high-value areas of the risk field such as motor vehicle lanes and non-motor vehicle lanes at intersections, the boundary of the electronic fence shrinks inward to a range of 3 meters around the traffic signal pole; in low-value areas of the risk field such as the edge of the pedestrian crossing at intersections, the boundary of the electronic fence expands outward to a range of 8 meters, which avoids high-density traffic flow and provides sufficient safety margin for drone landing.

[0050] Roadside Precision Early Warning and Takeoff / Landing Execution: As the drone begins its descent along its planned path, edge computing nodes, based on the drone's optimal landing path and dynamic risk field, accurately identify pedestrians and non-motorized vehicles about to enter the drone's landing path's spatial and temporal range. These nodes then send warning commands to directional acoustic devices and variable color temperature lighting devices on the roadside smart light poles. The directional acoustic devices issue voice warnings containing precise spatiotemporal information to the target traffic participants, while the variable color temperature lighting devices simultaneously switch to warning light effects, reminding the target objects to avoid collisions with the drone.

[0051] Within a dedicated time window, the drone completes its landing along a dynamically planned low-risk path. Throughout the entire takeoff and landing process, the central cloud control platform receives data transmitted from the sensing unit, edge computing nodes, and execution unit in real time, enabling global monitoring of the drone's takeoff and landing process. At the same time, the relevant data from this takeoff and landing is incorporated into the model training sample library to provide data support for subsequent model optimization.

[0052] Overall system operation assurance: In this embodiment, the perception unit is responsible for collecting all roadside traffic data, providing basic data support for all control decisions; the edge computing node, as the core processing unit, completes computational tasks such as digital twin modeling, game theory solving, and path planning locally, ensuring the real-time issuance of control commands; the execution unit is deployed on road ancillary facilities such as traffic signal poles and smart light poles, efficiently executing actions such as drone take-off and landing, signal timing adjustment, and accurate early warning; the central cloud control platform realizes global monitoring and model iterative optimization, with the four-layer architecture working together to ensure the efficient, safe, and intelligent implementation of dynamic drone take-off and landing control.

[0053] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for dynamic take-off and landing control of unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Step 1: Real-time access to multi-source roadside sensing data, construct a real-time four-dimensional digital twin model containing dynamic traffic participants, and construct a dynamic risk field in the spatiotemporal dimension based on the predicted trajectories of each participant. Step 2: Receive take-off and landing requests from the drone, and define the drone, the intersection traffic light control system, and the ground traffic flow aggregated by the digital twin model as game participants, and initialize their respective strategy spaces; Step 3: In the digital twin environment, by iteratively calculating the comprehensive utility function under different strategy combinations, we seek the Nash equilibrium strategy combination that maximizes the utility of all participants. The Nash equilibrium strategy combination defines the take-off and landing timing, path, and timing coordination scheme of the UAV and traffic lights. Step 4: Based on the solved Nash equilibrium strategy combination, control commands are sent to the drone and traffic lights respectively to dynamically adjust the drone's flight path and the electronic fence boundary, thereby achieving coordinated optimization of air and ground resources.

2. The method for dynamic take-off and landing control of unmanned aerial vehicles according to claim 1, characterized in that, Step 1, which involves constructing a dynamic risk field, specifically includes: using a trajectory prediction model based on social generative adversarial networks or Transformer to predict the trajectories of pedestrians and non-motorized vehicles in the next 3-5 seconds and their probability distribution; based on the predicted trajectories, constructing a Gaussian risk field for each dynamic obstacle, and superimposing all individual risk fields to form a dynamic risk heat map of the intersection that changes continuously with time and space, serving as the underlying mathematical model for the dynamic adjustment of the electronic fence.

3. The method for dynamic take-off and landing control of unmanned aerial vehicles according to claim 2, characterized in that, The strategy space for drones is as follows: Define the policy space of the UAV U as , This indicates that you should immediately descend along the nominal path. Indicates hovering and waiting After time, it descends along the dynamic programming path. This indicates that the aircraft has climbed up and is about to retake the aircraft to the waiting airspace. Define the policy space of traffic light T as , This indicates that the current signal timing scheme will be maintained. This indicates that the red light phase will be extended based on the current cycle. , This indicates that the red light phase of the next cycle will be activated ahead of schedule. ; Define the policy space of ground traffic flow G as , , These correspond to the high-density traffic state, medium-density traffic state, and low-density traffic state output by the trajectory prediction model, respectively.

4. The method for dynamic take-off and landing control of unmanned aerial vehicles according to claim 3, characterized in that, The formula for calculating the comprehensive utility function in step 3 is as follows: ; in It is the integral risk value of the drone along its take-off and landing path. It is a quantification of the impact of drones on the average speed of ground traffic flow. The extra electrical energy consumed by drones while waiting or taking off again. It is a preset priority coefficient based on task type; The Nash equilibrium solution employs iterative elimination of disadvantageous strategies or optimization algorithms based on swarm intelligence to quickly search for the strategy combination that maximizes the individual utility of all game participants within a limited strategy space, serving as the final collaborative control scheme.

5. The dynamic take-off and landing control method for unmanned aerial vehicles according to claim 4, characterized in that, In step 4, the UAV flight path is dynamically adjusted using the fast squared forward algorithm. Using the dynamic risk field generated in step 1 as the cost map, a real-time flight path with the lowest cumulative risk score is planned from the current position of the UAV to the take-off and landing point. The real-time flight path is dynamically replanned as the risk field is updated.

6. The method for dynamic take-off and landing control of unmanned aerial vehicles according to claim 5, characterized in that, The specific steps for dynamically adjusting the UAV flight path in step 4 are as follows: using the fast squared forward algorithm, with the dynamic risk field generated in step 1 as the cost map, a real-time flight path with the lowest cumulative risk score from the current position of the UAV to the take-off and landing point is planned. This path is dynamically replanned as the risk field is updated.

7. The method for dynamic take-off and landing control of unmanned aerial vehicles according to claim 6, characterized in that, In step 4, the boundary of the electronic fence is a dynamic contour subset of the dynamic risk field, and its spatial morphology is linked to the risk field: in the high-value area of ​​the risk field, the fence boundary contracts inward to avoid high-density traffic flow. In low-risk areas, the fence boundary expands outward to provide a greater safety margin, thereby achieving spatiotemporal dynamic adaptation of the electronic fence form.

8. The method for dynamic take-off and landing control of unmanned aerial vehicles according to claim 7, characterized in that, It also includes a time-space exchange mechanism. By solving the Nash equilibrium, the system makes decisions through traffic light policy adjustments to create a dedicated time window for drones free from ground traffic interference. In exchange, drones must complete take-off and landing within the dedicated time window; otherwise, they must pay additional waiting energy costs. This achieves the trading and optimized allocation of airspace and ground right-of-way in the time dimension.

9. The dynamic takeoff and landing control method for unmanned aerial vehicles according to claim 8, characterized in that, Step 4 of the execution and collaborative control also includes a roadside precision early warning step: based on the solved optimal take-off and landing path of the drone and the dynamic risk field, using the directional sound wave device and variable color temperature light effect device on the roadside smart light pole, an avoidance warning containing precise time and space information is issued to specific pedestrians or vehicles that are about to enter the spatial and temporal range of the drone's take-off and landing path, thereby improving the safety of control.

10. A collaborative control system for takeoff and landing of unmanned aerial vehicles (UAVs), characterized in that, To implement the UAV dynamic take-off and landing control method of claim 9, the system includes a sensing unit, an edge computing node, an execution unit, and a control platform; The sensing units are deployed on the roadside and include cameras, millimeter-wave radar, and V2X roadside units; Edge computing nodes are deployed on the roadside and communicate with sensing units to execute the method steps of claim 9; The execution unit is deployed on smart light poles or traffic signal poles, including drone take-off and landing base stations, directional sound wave early warning devices, and variable color temperature light effect devices. It is connected to edge computing nodes to execute control commands. The management and control platform is deployed in the central cloud and communicates with the sensing unit, edge computing nodes, and execution unit respectively, for global monitoring and model training.

Images