A method for avoiding collision of manned aircraft and unmanned aircraft cooperative operation formation

Abstract

The application discloses a manned-unmanned aircraft cooperative operation formation collision avoidance control method, and belongs to the technical field of aircraft cooperative control; the method adopts a hierarchical architecture separating performance layer and safety layer: the performance layer generates nominal instructions for tracking operation slots by model predictive control; the safety layer filters threats in real time based on collision cones, and converts collision avoidance conditions into linear inequality constraints about unmanned aircraft acceleration by using control barrier functions. By setting hierarchical safety parameters for unmanned aircraft-manned aircraft interaction and unmanned aircraft-unmanned aircraft interaction, and imposing hard constraints on the former and soft constraints that can be relaxed on the latter in solving, the absolute safety priority of manned aircraft is ensured. The application realizes effective decoupling of high-performance formation tracking and real-time autonomous collision avoidance, improves the real-time performance and reliability of the system while ensuring safety, and is suitable for large-scale cooperative operation tasks.

Description

Technical Field

[0001] This invention relates to the field of aircraft cooperative control and autonomous safety collision avoidance technology, specifically to a method for manned and unmanned aircraft cooperative formation collision avoidance control. Background Technology

[0002] In manned / unmanned collaborative mission scenarios, a key mode involves a manned aircraft leading and multiple unmanned aircraft forming a specific formation to conduct operations or fly alongside it. In this scenario, the unmanned aircraft control system must simultaneously perform two core tasks: first, accurately tracking the dynamic slots determined by the manned aircraft's movement to maintain formation; and second, avoiding collisions with manned aircraft and other unmanned aircraft in real time. Performance and safety objectives are intertwined, and especially under complex maneuvers or dense formations, command conflicts are highly likely to arise, leading to system performance degradation and even safety incidents.

[0003] In existing technologies, collision avoidance constraints are typically embedded directly into artificial potential fields, velocity obstacle (VO) methods, or model predictive control (MPC) to solve such problems. However, these methods all have certain limitations: artificial potential field methods are prone to getting trapped in local optima, and the parameter tuning process is relatively complex; traditional velocity obstacle methods exhibit strong conservatism in dynamic target scenarios; and directly embedding complex nonlinear collision avoidance constraints into the MPC optimization problem significantly increases the computational load, affects the real-time performance of the system, and may even lead to an unsolvable optimization problem when constraints conflict. More importantly, existing methods often apply a uniform safety strategy to all members of the formation, failing to fully reflect the highest safety priority that manned machines, as critical platforms, should possess.

[0004] Therefore, there is an urgent need for a collaborative operation formation control method that can effectively decouple performance and safety objectives, prioritize manned and machine safety, and take into account real-time performance and reliability. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for collision avoidance control in collaborative operations between manned and unmanned aircraft.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This application provides a manned and unmanned aerial vehicle (UAV) collaborative formation collision avoidance control method, which adopts a layered control architecture that separates the performance layer and the safety layer. The method includes the following steps: S1. Define mission instructions for the UAV based on the motion state of the manned aircraft, establish a UAV dynamics model, and initialize graded safety parameters; S2. Design a nominal controller to solve for the nominal control commands used by each UAV to track the mission instructions; S3. Using the collision cone principle, assess the collision risk of each drone and identify targets with imminent threats; S4. Based on the control barrier function, the collision avoidance safety conditions for the imminent threat target are transformed into linear constraints on the UAV control commands, and the constraints are grouped according to the threat type; S5. Synthesize the nominal control instruction with the safety correction term constructed based on the linear constraint to form the desired control instruction; S6. Using the desired control command as a reference, solve online for the final control command that satisfies all the linear constraints and execution constraints, and output the execution command.

[0007] Optionally, S1 includes the following steps: S1.1. In the manned aircraft system, pre-allocate the work slot offset for each UAV, and convert it to the inertial coordinate system according to the manned aircraft status to generate the desired slot position and desired velocity for each UAV. S1.2. Establish a discrete dynamic model of the UAV outer loop based on a second-order integrator and apply input amplitude constraints; S1.3. Set graded safety parameters for UAV-managed machine interaction and UAV-UAV interaction, wherein the safety parameter value for UAV-managed machine interaction is greater than the corresponding parameter value for UAV-UAV interaction; subsequent risk assessment is only performed on objects within the trigger distance.

[0008] Optionally, S2 includes the following steps: S2.1. Construct a state vector using the UAV's position and velocity, and use the aforementioned discrete dynamics model as the prediction model; based on the assumption of constant speed or constant attitude extrapolation of the manned aircraft state, generate a slot reference state sequence in the prediction time domain; S2.2. Construct a quadratic cost function with the goal of minimizing slot tracking error, control energy, and control increment variation, and apply input amplitude constraints in the prediction time domain; S2.3. Solve the linear model predictive control optimization problem, and use the first term of the obtained optimal control sequence as the nominal outer loop acceleration command.

[0009] Optionally, S3 includes the following steps: S3.1. Calculate the relative position, relative speed, and relative distance between the local machine and the threat target, and match the corresponding security parameters according to the interaction type; S3.2. Calculate the collision cone half angle based on the relative distance and the minimum safe distance; calculate the angle between the relative velocity and the relative position; construct the collision cone safety margin function and calculate the predicted collision time; S3.3. Filter the set of triggering neighbors based on the triggering distance, and select the most urgent threat target subset from the set according to the safety margin function value, proximity and predicted collision time.

[0010] Optionally, S4 includes the following steps: S4.1. Define the collision cone safety margin function as a barrier function, and match different control barrier function gain coefficients for different interaction types; S4.2. Apply a first-order control barrier function condition to the barrier function, requiring its time derivative and function value to satisfy a preset linear relationship; expand the time derivative in combination with the relative motion relationship and the assumption of "neighboring machine acceleration is set to zero"; S4.3. The standard form of the linear inequality constraint for the outer loop acceleration command of the machine is derived and grouped according to the interaction type into the UAV-maned machine constraint group and the UAV-UAV constraint group.

[0011] Optionally, S5 includes the following steps: S5.1. Define the nominal outer ring acceleration command as the initial direction quantity; S5.2. Based on the gradient direction and constraint constant term of the linear inequality constraint, construct a safety correction term to increase the safety margin; S5.3. Add the nominal outer loop acceleration command to the safety correction term to synthesize the desired outer loop acceleration command, and assign higher weights to the threats of UAVs and manned aircraft when constructing the correction term.

[0012] Optionally, S6 provides two online solution algorithms: a non-optimized amplitude adjustment method and a quadratic programming optimization method.

[0013] Optionally, the non-optimized amplitude adjustment method specifically includes: The desired outer loop acceleration command direction is fixed; Transform all linear inequality constraints into one-dimensional inequalities concerning a single control amplitude; By combining all one-dimensional inequalities and input limits, the total feasible range of the control amplitude is determined; Within this range, select the value closest to the desired amplitude and combine it with a fixed direction to obtain the final command; when the range conflict involves both UAVs and manned structures, activate the emergency safety mode.

[0014] Optionally, the quadratic programming optimization method specifically includes: Construct an optimization problem with the objective function of minimizing the deviation between the final instruction and the desired instruction; Hard constraints are imposed on the structure of both unmanned and manned aircraft; soft constraints are imposed on the structure of unmanned and manned aircraft by introducing slack variables. Solving this convex quadratic programming problem yields the final instruction; when the problem becomes infeasible due to hard constraints, an emergency safety mode is activated.

[0015] Optionally, the method further includes: When the actual distance is detected to be less than the corresponding minimum safe distance, the relevant drone is controlled to take emergency evasive action in a radial direction away from the threatening target until the safe state is restored and then the normal control process is returned.

[0016] Compared with the prior art, this application has the following beneficial effects: This invention proposes a manned-unmanned aerial vehicle (MAV) and unmanned aerial vehicle (UAV) collaborative formation collision avoidance control method. First, it establishes a relative kinematic model and formation geometry between the MAV and UAV, defines the operational slots within the MAV system, and transforms them to the inertial frame using an attitude matrix to generate UAV slot position and velocity references. Based on this, the core innovation lies in constructing a layered control architecture of "performance layer + safety layer": the performance layer employs linear model predictive control (Linear Model Predictive Control). The MPC (Multi-Purpose Control) method uses a rolling solution to obtain the nominal outer-loop acceleration command for slot tracking under input amplitude constraints. The safety layer selects the most pressing threat object based on collision cone risk assessment and uses a control barrier function (CBF) to transform collision avoidance safety conditions into linear inequality constraints on the outer-loop acceleration command, achieving safe filtering and correction of the nominal command. For UAV-managed (UM) and UAV-UAV (UU) interactions, graded safety distances and barrier parameters are set respectively. During the solution phase, hard constraints are applied to UM constraints, while soft relaxation is allowed for UU constraints, thus ensuring the priority safety of manned aircraft and the overall feasibility of the system. Simultaneously, this invention provides a non-optimized "direction + amplitude" low-computing-power solution embodiment and a quadratic programming (QP) high-performance solution embodiment, which can balance real-time performance, smoothness, and safety in large-scale operational formation scenarios. The safety layer is decoupled from the nominal controller structure, making the nominal controller replaceable and seamlessly integrated with safety constraints, providing an efficient and reliable autonomous control solution for manned / UAV collaborative operations, accompanying flights, and complex airspace missions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a collaborative formation collision avoidance control method for manned and unmanned aerial vehicles. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Furthermore, in this invention, an element referred to as fixed to or disposed on another element may be directly disposed on the other element, or there may be an intermediate element. When an element is considered to be connected to another element, it may be directly connected to the other element, or there may be an intermediate element present simultaneously. The terms vertical, horizontal, left, right, and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0020] First, the application scenarios of the technical solution of this application are described as follows: The technical solution of this application can be applied to collaborative operations of urban fire fighting and emergency rescue: In urban emergency rescue operations such as high-rise building fires, a manned rescue helicopter serving as a command and delivery platform works in coordination with multiple fire reconnaissance drones and material delivery drones. The manned helicopter hovers in a safe airspace to direct operations, while the drones need to approach the fire scene to conduct reconnaissance and drop fire extinguishing bombs or emergency supplies.

[0021] In complex low-altitude environments with numerous buildings, drones need to dynamically adjust their formation to avoid static obstacles such as buildings and chimneys, while also avoiding collisions with each other. The method of this invention enables drone swarms to maneuver safely and flexibly in narrow airspace, using manned aircraft as dynamic reference points. The safety layer not only handles collision avoidance (UU) between drones, but also strictly ensures that no drone will collide with the hovering manned command aircraft (UM) due to loss of control or misjudgment, providing reliable safety assurance for manned / unmanned collaborative operations in dense urban environments.

[0022] The technical solution presented in this application can be applied to the intensive scheduling and guidance of inbound and outbound flights at large logistics centers. In future large-scale air logistics hubs, manned monitoring / guidance aircraft and a large number of cargo drones entering and leaving the port may operate frequently in the same airspace; manned guidance aircraft will need to shuttle among the aircraft to perform tasks such as inspection and emergency guidance.

[0023] The method of this invention can be deployed in the autonomous collision avoidance system of all UAVs and manned guide aircraft. When a UAV is flying along a planned route, its safety layer will continuously monitor its relative status with the manned guide aircraft and other UAVs. Once the manned guide aircraft needs to approach or cross the UAV flow due to mission requirements, the relevant UAVs will give priority and actively give way to the manned aircraft based on this method (UM constraint priority takes effect), and adjust their own trajectories under the coordination of the algorithm to avoid chain conflicts with other UAVs, thereby greatly improving the operational safety and scheduling flexibility of the entire airspace.

[0024] The technical solution of this application will be described in detail below based on the application scenarios: See Figure 1This application provides a manned and unmanned aerial vehicle (UAV) collaborative formation collision avoidance control method, which adopts a layered control architecture that separates the performance layer and the safety layer. The method includes the following steps: S1. Define mission instructions for the UAV based on the motion state of the manned aircraft, establish a UAV dynamics model, and initialize graded safety parameters.

[0025] S2. Design a nominal controller to solve for the nominal control commands used by each UAV to track the mission instructions.

[0026] S3. Using the collision cone principle, assess the collision risk of each drone and identify targets with imminent threats.

[0027] S4. Based on the control barrier function, the collision avoidance safety conditions for the imminent threat target are transformed into linear constraints on the UAV control commands, and the constraints are grouped according to the threat type.

[0028] S5. Synthesize the nominal control instruction with the safety correction term constructed based on the linear constraints to form the desired control instruction.

[0029] S6. Using the desired control command as a reference, solve online for the final control command that satisfies all the linear constraints and execution constraints, and output the execution command.

[0030] The following is a detailed explanation of steps S1-S6: S1 includes the following steps: S1.1. In the manned aircraft system, pre-allocate the work slot offset for each UAV, and convert to the inertial coordinate system according to the manned aircraft status to generate the desired slot position and desired velocity for each UAV.

[0031] Assume the system includes a manned aircraft and Unmanned aerial vehicles (UAVs); manned aircraft are designated with the following serial numbers: The drone's serial number is The set of all platform IDs is Define the inertial coordinate system as follows: Any platform The position and velocity are denoted as follows: , and all with Expression; Definition of manned-machine system as Its attitude transformation matrix to the inertial frame is , satisfying any The vector represented in the middle Its representation in an inertial frame is as follows In each control cycle Manned aircraft broadcast to drones As a reference input for the operational formation; preferably, when only the heading angle of the manned aircraft is available. When the complete attitude matrix is ​​unavailable, a simplified rotation matrix around the vertical axis can be used. Instead, it ensures that the formation rotates synchronously with the manned aircraft's heading in the horizontal plane.

[0032] To achieve collaborative formation in collaborative operations, in a manned-machine system The internal configuration pre-allocates work slot offsets for each drone. Its meaning is drone. The expected relative position of the manned aircraft within the aircraft system; after converting the slot offset to the inertial frame, the expected position of the UAV in the slot is obtained: .

[0033] The expected speed of the corresponding slot can be defined as: .

[0034] in The zero vector is preferred to represent that the slot is fixed within the manned-machine system; when a more precise description of the velocity term as the formation changes with attitude is required... It can be constructed from slot offset and attitude change rate; therefore, the cooperative operation control objective of the present invention is stated as: under the premise of satisfying safety constraints, to make the state of each UAV... track This allows the formation to evolve synchronously with the movement of manned aircraft and maintain the operational structure.

[0035] S1.2. Establish a discrete dynamic model of the UAV outer loop based on a second-order integrator and apply input amplitude constraints.

[0036] This invention employs a second-order integrator model for each UAV at the outer ring level; the continuous-time form is as follows: .

[0037] in Let the outer-loop acceleration command of the UAV in the inertial frame be ; and let the discrete control period be . Discrete time is Then the discrete-time form is: .

[0038] Further define the state as The linear discrete state-space form can be obtained. This is used for unified modeling of subsequent Linear MPC. It should be noted that the above outer loop model is valid only if the inner loop of the UAV can track the acceleration command of the outer loop within the physically feasible range. When there is thrust saturation, attitude limiting, or response lag, more stringent input limiting and rate of change constraints can be used in subsequent solutions to ensure overall executability.

[0039] To ensure the commands can be executed, an input limiting constraint is applied to each drone: .

[0040] Preferably, to facilitate the implementation of linear MPC and quadratic programming solutions, box-constrained approximation can be used. , , Alternative norm constraints; this substitution does not change the logic of the method of the present invention, but only changes the numerical implementation form and the way constraints are expressed.

[0041] S1.3. Set graded safety parameters for UAV-managed machine interaction and UAV-UAV interaction, wherein the safety parameter value for UAV-managed machine interaction is greater than the corresponding parameter value for UAV-UAV interaction; subsequent risk assessment is only performed on objects within the trigger distance.

[0042] To prioritize manned-machine safety, potential collision interactions are categorized into two types: UAV-Manned (UM) and UAV-UAV (UU), with safety parameters set for each category. For any UAV... The object corresponding to UM is The object corresponding to UU is The minimum safe distance is set to meet the following requirements: .

[0043] The set trigger distance satisfies: .

[0044] And set the gain of the control barrier function to satisfy: .

[0045] Preferably, to ensure initial feasibility, the initial state of the system satisfies as well as When a minor violation occurs, an emergency separation strategy is initiated to allow the system to return to a safe set before resuming formation control. This emergency separation strategy preferably follows a "manned aircraft priority" hierarchical rule: if any UAV and manned aircraft are detected to meet the following conditions... Then, the radial direction away from the manned aircraft is taken as the separation direction, and the outer ring acceleration command of the UAV is set to the maximum feasible amplitude and pointed in the direction of increasing distance, until the condition is met. And corresponding collision cone margin Then exit the emergency separation and proceed to step S2; if there is no UM violation but there is a UU violation, then the nearest collision object is used as the separation object and the same radial separation method is used until the condition is met. Then exit; it should be noted that, and This is a safety margin threshold used to prevent repeated triggering as soon as the loop returns to the boundary, thereby improving closed-loop stability.

[0046] Assume the system uses discrete-time rolling control, in each cycle drones Obtain self-state estimate Receiving broadcasts from manned aircraft And obtain the state estimates of neighboring platforms. To reduce computational load and avoid constructing unnecessary conservative constraints on distant objects, risk assessment and constraint construction are only performed on objects within the trigger distance: when Manned aircraft will be included in the constraints of this cycle; when Time will drone Included in the constraints of this cycle; it should be noted that the subsequent collision cone risk assessment and control barrier function linear inequality construction depend only on relative position and relative velocity, and can be derived from... and Calculations show that when the neighboring machine's acceleration is unavailable, the implementation assumption of "neighboring machine acceleration unknown / zeroed" can be used to transform the constraint into a command regarding the machine's outer loop acceleration. The linear inequalities ensure that the method is closed and implementable. When there is a delay or packet loss in communication, the state of the neighboring machine can be generated by the nearest hold or constant speed extrapolation for the current cycle calculation. This process does not change the form of the constraint, but only changes the conservatism of the constraint.

[0047] In one specific implementation, S2 includes the following steps: S2.1. Construct a state vector using the UAV's position and velocity, and use the aforementioned discrete dynamics model as the prediction model; based on the assumption of constant speed or constant attitude extrapolation of the manned aircraft state, generate a slot reference state sequence in the prediction time domain.

[0048] For any drone The discrete second-order integrator model from step S1 is used as the prediction model, and the state vector is defined as: .

[0049] The control quantity is the outer loop acceleration command. Then we have: .

[0050] in .

[0051] Alternatively, an equivalent discretization form can be used, such as position updates that do not include... The above modeling also applies; both are essentially second-order integrator discrete models; the above modeling ensures that the nominal outer-loop acceleration command output by the Linear MPC is consistent with the subsequent collision cone + control barrier function constraints. The linear inequality constraint forms are consistent, thus achieving structural compatibility between nominal control and security layer filtering.

[0052] Meanwhile, the reference state for the drone slot is defined as follows: .

[0053] And define the slot tracking error as: .

[0054] S2.2. Construct a quadratic cost function with the objectives of minimizing slot tracking error, control energy, and control increment variation, and apply input amplitude constraints in the prediction time domain.

[0055] Let the prediction time domain length be Let the predicted sequence be and Construct a Linear MPC objective function to enable the UAV to approximate the slot reference within the prediction domain while suppressing the amplitude and rate of change of the control variable. The preferred form is the following quadratic form: .

[0056] in , , This is the weight matrix. To control the increment and obtain a smoother nominal instruction, it should be noted that the present invention does not directly add collision avoidance constraints to the nominal layer, but maintains the structure of generating an attraction term towards the target first, and then correcting it by the safety layer. This makes it easier to uniformly filter and satisfy constraints on the nominal input in steps S5 and S6 by combining the collision cone with the control barrier function safety layer.

[0057] It should be noted that, in order to make the prediction problem of linear MPC solvable within each control cycle, a slot reference sequence within the prediction domain must be generated. The present invention preferably adopts a generation method consistent with the idea of ​​zero / constant velocity extrapolation when the acceleration of the external object is unknown: it assumes that the velocity and attitude of the manned aircraft remain constant or extrapolates according to constant velocity / constant attitude within the prediction domain, that is: .

[0058] This yields the slot reference prediction sequence: .

[0059] And construct accordingly When the system can obtain higher-order information about the manned aircraft (such as acceleration estimation or heading angular velocity), the above constant velocity / constant attitude assumption can also be replaced by corresponding kinematic extrapolation, but without changing the structure of the Linear MPC of this invention and the interface of the subsequent safety layer.

[0060] Regarding the constraint set, input limiting constraints are applied to the control quantity at each step within the prediction domain: .

[0061] Alternatively, a box-constrained approximation can be used; preferably, to further approximate the idea of ​​avoiding overly aggressive commands due to distant targets, a nominal reference pruning mechanism can be introduced: when When the value is too large, the slot reference is projected onto a surface centered at the current position with a radius of [missing information]. Within the spherical domain, a local slot reference is thus formed. and replace it It participates in forecasting and cost calculation to ensure that nominal control does not generate excessive acceleration requirements due to excessively distant references.

[0062] S2.3. Solve the linear model predictive control optimization problem, and use the first term of the obtained optimal control sequence as the nominal outer loop acceleration command.

[0063] In each control cycle drones In the current state With reference sequence of slots in the prediction domain Using the above linear MPC optimization problem as input, the optimal control sequence is obtained by solving the problem. The first term of the sequence is used as the nominal outer loop acceleration command for this period: .

[0064] It should be noted that, This invention is only used for slot / formation performance tracking and has not yet demonstrated consideration of collision risk and positive invariance of the safety set; therefore, in step S5, the present invention combines the safety correction term constructed based on the collision cone and control barrier function with... Synthesizing to obtain the desired input In step S6, the non-optimized direction and amplitude method or QP method is used to... Safety filtering and constraint satisfaction are performed to achieve an overall structure that combines attraction terms, repulsion terms, and constraint solutions. At the same time, Linear MPC is used to replace attraction terms to improve job slot tracking performance.

[0065] In one specific implementation, S3 includes the following steps: S3.1. Calculate the relative position, relative speed and relative distance between the local machine and the threat target, and match the corresponding security parameters according to the interaction type.

[0066] Based on the safety parameter initialization in step S1 and the nominal outer loop acceleration command generation in step S2, a collision cone discrimination mechanism is established: first risk assessment, then threat pair selection. This mechanism is used to determine the set of interaction objects that each UAV needs to focus on during the current control cycle. This set will serve as the input for the subsequent construction of linear constraints for the control barrier function and the calculation of safety correction terms, thereby significantly reducing the computational load and avoiding the imposition of unnecessary conservative constraints on distant objects without sacrificing theoretical consistency. During the control cycle... For any drone With any interactive object The relative position and relative velocity are defined as follows: .

[0067] Define relative distance as And assume in the calculation ;when When the value is extremely small, a positive lower bound can be used instead to avoid division by zero; to reflect human-machine priority safety, different safety distance and trigger distance parameters are matched for different types of interaction pairs: when When adopted , ;when When adopted , Therefore, subsequent collision cone determination and threat screening are carried out under a unified parameter system that is more conservative for UM and more conventional for UU.

[0068] S3.2. Calculate the collision cone half angle based on the relative distance and the minimum safe distance; calculate the angle between the relative velocity and the relative position; construct the collision cone safety margin function and calculate the predicted collision time.

[0069] Based on relative position and safe distance The collision cone half-angle (cone angle) is defined as: .

[0070] Its geometric meaning is: when the relative positions are fixed, if the direction of the relative velocity falls into the direction of... The axis is the half angle is the axis. Within the velocity cone, there exists a distance crossing along the current relative velocity direction. The risk; further define the angle between the relative velocity and the direction of the connecting line as: .

[0071] And construct the collision cone safety margin function: .

[0072] when When the relative velocity direction is located outside the collision cone, it indicates a safe state for the collision cone; when When the relative velocity direction falls into the collision cone, it indicates a collision cone risk state; to avoid misjudging a moving object as a threat, a proximity criterion is further introduced: when If the two are considered to be approaching, then they are considered to be separating; therefore, a valid threat can be determined when both conditions are met simultaneously. and The interaction pair.

[0073] To prioritize multiple threats, a collision time is defined. For the relative motion to satisfy the constant velocity assumption for the first time The smallest positive root (if no positive root exists, set to 0) From the quadratic equation: .

[0074] The discriminant can be obtained as follows: .

[0075] when When, it indicates that the relative trajectory does not intersect the safe sphere, let ;when and When the minimum positive root is: .

[0076] and only It is believed that there is a future collision time significance; it should be noted that if the current conditions have already been met... This constitutes an emergency situation where a safe distance has been reached, and orders can be issued directly. And assign it the highest priority for processing in subsequent security layers.

[0077] S3.3. Filter the set of triggering neighbors based on the triggering distance, and select the most urgent threat target subset from the set according to the safety margin function value, proximity and predicted collision time.

[0078] To reduce computational load while maintaining the trigger distance filtering principle, for each drone... During the control cycle Constructing the triggering neighbor set: .

[0079] in Select according to the category rules in step S3.1 (UM uses) UU uses ); in the set Inside, for each calculate and Based on this, construct the subset of the most pressing threats. Preferably, The construction of the object satisfies the following principle: First, all objects that meet the valid threat criteria are preferentially retained, i.e. and The interaction pairs; secondly, when the number of the above objects exceeds the preset limit. At that time, according to Sort by size from smallest to largest and select the first few. The most pressing threat subset is selected; secondly, when there is no valid threat but the trigger set is not empty, it can be selected as... or The minimum value is selected from one or more objects that are closest to the risk boundary as conservative monitoring objects to ensure that the safety layer has the ability to intervene in advance when the risk is critical.

[0080] In one specific embodiment, S4 includes the following steps: S4.1. Define the collision cone safety margin function as a barrier function, and match different control barrier function gain coefficients for different interaction types.

[0081] This step outputs the set of most pressing threats in step S3. Based on this, the angular safety condition of the collision cone is transformed into a control barrier function constraint, and the outer ring acceleration command of the UAV is further derived. The linear inequality form provides a strict constraint expression for subsequent calculations of safety correction terms and solutions to safety filters. To meet the needs of manned-machine collaborative operations, this step matches different safety distances and barrier parameters for UAV-manned (UM) and UAV-UAV (UU) operations, respectively, giving manned-machine protection constraints higher priority and stronger conservatism at the modeling level; for any UAV... and its most pressing threats The candidate barrier function is defined using the collision cone discriminant: .

[0082] in , and using the same relative amount as in step S3. , The geometric meaning of the barrier function is: when When the relative velocity direction is located outside the collision cone, the system is in a safe state around the collision cone; when At that time, the relative velocity direction enters the collision cone, posing a collision risk; to ensure If defined, the constraints are satisfied by default during construction. ,when When an emergency is declared and the mandatory safety policy is prioritized in subsequent solutions, to reflect the priority of human-machine safety, when (UM) time take , ;when (UU) time take , and satisfy , Thus, parameter hierarchies can be achieved under the same barrier function structure.

[0083] S4.2. Apply a first-order control barrier function condition to the barrier function, requiring its time derivative and function value to satisfy a preset linear relationship; expand the time derivative in combination with the relative motion relationship and the assumption of "neighboring machine acceleration is set to zero".

[0084] To ensure safe assembly The positive invariance is determined by using a first-order control barrier function condition, and the class is... The function takes linear form This requires that each threat be addressed. satisfy: .

[0085] because By relative position With relative velocity The decision, when expanded using the chain rule, yields: .

[0086] Combining relative motion relationships and ,get: .

[0087] Substituting this into the control barrier function condition and rearranging, we get: .

[0088] It should be noted that in distributed job formation, the interaction objects outer ring acceleration command Typically, these constraints are unavailable or do not participate in the joint solution. To maintain constraint realizability, this invention employs the assumption that the acceleration of neighboring machines is unknown or zero during the constraint construction phase, i.e., it assumes... (Equivalent to assuming the neighboring machine's speed remains constant within a control cycle), thus unifying the constraints to only concern... The processing does not change the structure of the barrier function, only its conservatism; in manned aircraft scenarios, when the manned aircraft can broadcast higher-order motion information, it can... Using its estimated value to replace it reduces conservatism, but it is not a necessary condition for this step.

[0089] S4.3. The standard form of the linear inequality constraint for the outer loop acceleration command of the machine is derived and grouped according to the interaction type into the UAV-maned machine constraint group and the UAV-UAV constraint group.

[0090] In Adoption Then, the above equation can be written in the form of a standard linear inequality: .

[0091] in .

[0092] Considering , The definition above , This can be equivalently written in gradient form with respect to the local state variables, for example, using... Alternative and use Alternative Thus obtaining the same Linear constraint expression; any equivalent expression can be chosen in the implementation of this invention, as long as it guarantees... , This can be calculated from the currently measured / estimated relative position and relative velocity. To facilitate subsequent hierarchical solutions, all constraints are output in two groups according to their categories: when... Time constraints are grouped into the UM constraint group. ,when Time constraints are assigned to the UU constraint group. In each control cycle For drones Its threat set The internal constraints are stacked row by row to form a matrix: .

[0093] and input limiting constraints This serves as the basis for constructing the safety correction term in subsequent step S5 and as the input for the safety filtering solution in step S6, thereby realizing the nominal performance layer output. The overall structure is seamlessly integrated with the linear constraints of the collision cone and the control barrier function safety layer.

[0094] In one specific embodiment, S5 includes the following steps: S5.1. Define the nominal outer ring acceleration command as the initial direction quantity.

[0095] For any drone The nominal outer loop acceleration command obtained in step S2 is denoted as: .

[0096] And construct security modification items Thus, the direction of the composite acceleration is obtained: .

[0097] This invention adopts this interface format, and the desired outer loop acceleration command will be uniformly defined thereafter. This ensures seamless integration with both types of security filtering implementations.

[0098] S5.2. Based on the gradient direction and constraint constant term of the linear inequality constraint, construct a safety correction term to increase the safety margin.

[0099] For each threat Step S4 has obtained information about Linear inequality constraints: .

[0100] in , To ensure the system state moves towards a higher safety margin The present invention adopts the direction of evolution with The equivalent gradient direction is used as the correction direction, and a weighting function is applied to enhance the correction when the drone is close to the constraint boundary and weaken it when it is far from the boundary. Threat set structure: .

[0101] in For the number of threat targets, The gradient direction of the barrier function with respect to the local velocity variable (and) Consistent directional information), The risk level weighting function is used; preferably, the weighting function is: .

[0102] in , This is an adjustable parameter, and its function is: when the constraint approaches activation or approaches being broken (i.e., (smaller or even negative), Increase to enhance correction; when the system is far from the constraint boundary, Reduce to avoid unnecessary disturbances; it should be noted that when Time definition ,at this time In other words, nominal control is unaffected by security modifications.

[0103] S5.3. Add the nominal outer loop acceleration command to the safety correction term to synthesize the desired outer loop acceleration command, and assign higher weights to the threats of UAVs and manned aircraft when constructing the correction term.

[0104] After obtaining the direction of the composite acceleration Then, the desired outer-loop acceleration command for entering the safety filter solver is defined as follows: .

[0105] In the non-optimized implementation, the following... The direction is used as the control input direction, and the constraint is satisfied by selecting the amplitude; in the optimized implementation, subsequent steps are based on... The desired input for the quadratic programming problem is used as the input, and the closest value is found while satisfying the linear constraints of the control barrier function and the input limit. The feasible solution is thus achieved, thereby realizing a structure that first avoids risks in a directional manner and then strictly satisfies the constraints.

[0106] To reflect the priority of human-machine security and maintain consistency with the UM / UU classification parameters in step S4, it is preferable to assign a stronger weight or gain to UM threats in the security correction term construction, i.e., when Use larger weight parameters Or add a priority coefficient to the summation term. Thus At the directional level, priority is given to moving away from the manned-machine safety boundary; it should be noted that this increased priority does not change the rigid requirements of the linear inequality constraints themselves, but rather makes the desired input... Adjusting earlier and more smoothly toward satisfying UM constraints reduces conflict in step S6, minimizes control jitter, and improves the overall stability of the job formation.

[0107] In one specific implementation, S6 provides two online solution algorithms: a non-optimized amplitude adjustment method and a quadratic programming optimization method.

[0108] The non-optimized amplitude adjustment method specifically includes: fixing the direction of the desired outer loop acceleration command; transforming all linear inequality constraints into one-dimensional inequalities about a single control amplitude; combining all one-dimensional inequalities and the input limit to determine the total feasible interval of the control amplitude; selecting the value closest to the desired amplitude within this interval and combining it with the fixed direction to obtain the final command; and activating the emergency safety mode when the interval conflict involves both UAV and manned structures.

[0109] For any drone The desired outer loop acceleration command obtained in step S5 is then applied. As the basis for direction generation, define the unit direction vector: .

[0110] when Time Selection Order Take any fixed unit vector or directly let Based on this, the final outer-loop acceleration command is restricted to this direction, and the constraint is satisfied only by amplitude adjustment, that is: .

[0111] Apply any linear inequality constraint obtained in step S4 Substitution , get about One-dimensional linear inequalities: .

[0112] make Then the inequality applies to Given feasible interval constraints: when From time to time ;when From time to time ;when Time constraints degenerate into If this condition is not met, it means that the constraint cannot be satisfied in that direction through amplitude adjustment; therefore, for all constraints that need to be considered... Calculate each pair The lower and upper bounds, and the input limiting range. By finding the intersection of all possible intervals, we can obtain the total feasible interval: .

[0113] in and Each of the above The sign rule is obtained by taking the upper bound if a constraint only gives a lower bound. If only the upper bound is given, then the corresponding lower bound is taken as... When satisfied When the feasible interval is not empty, the optimal choice is to make the amplitude as close as possible to the desired amplitude. That is, take: .

[0114] in Indicates will Saturation to range Final output .

[0115] It should be noted that when interval conflicts occur... Or exist and In the case of this, it indicates the current direction. It is impossible to satisfy all constraints simultaneously through amplitude adjustment. To reflect human-machine priority safety, this invention adopts a hierarchical processing strategy: when a conflict involves the UM constraint group (i.e., ... When the UM constraint cannot be satisfied in the current direction, it is preferable to enter the emergency safety mode to quickly restore the UM safety margin. The preferred control output rule is to reset the direction to the gradient direction of the most pressing UM constraint and take the maximum feasible magnitude, that is: .

[0116] when When necessary, it is preferable to use a direction radially away from the manned aircraft. Emergency safety mode will continue to run until the conditions are met. and Exit and restore the direction and amplitude rules output according to this step; if the conflict is mainly caused by the UU constraint group, it is preferable to relax only the low-priority UU constraints. Limitations (such as ignoring an upper or lower bound of a U that causes a conflict) are used to restore the feasible interval without breaking the UM constraint set, thus balancing real-time performance and overall feasibility.

[0117] The quadratic programming optimization method specifically includes: constructing an optimization problem with the objective function of minimizing the deviation between the final command and the expected command; applying hard constraints to the UAV and manned structure; applying soft constraints to the UAV and UAV structure by introducing slack variables; solving the convex quadratic programming problem to obtain the final command; and initiating an emergency safety mode when the problem becomes infeasible due to hard constraints.

[0118] In each control cycle For any drone The desired outer loop acceleration command obtained in step S5 For reference, a convex quadratic programming problem is constructed to find the closest approximation under the constraints. A feasible solution. Let the decision variable be the final outer-loop acceleration command. And only introduce slack variables for the UU constraint. Then, we establish an optimization problem: .

[0119] The following constraints must be satisfied: Hard constraints must be applied to the UM constraint group. Relaxable constraints are applied to the UU constraint group: .

[0120] And simultaneously satisfy the input limiting constraint: .

[0121] in The relaxation penalty coefficient is used to ensure that the relaxation of UU occurs only when necessary and is as small as possible. Since the objective function is a strictly convex quadratic form and the constraints are linear inequalities and convex amplitude constraints, the above problem is a convex optimization problem; therefore, if feasible, there exists a unique optimal solution. When the UM hard constraint set makes this QP problem infeasible, it is preferable to enter the emergency safety mode to prioritize restoring the UM safety margin. Its control output rule is preferably to maximize the UM constraint satisfaction within the input amplitude, i.e., to solve: .

[0122] Its closed-form solution is: .

[0123] when The preferred approach is to use a direction radially away from the manned aircraft; the emergency safety mode continues to be executed until the conditions are met. and Then exit and resume the QP security filtering solution for this step.

[0124] Solving for the optimal solution Then, it is used as the final outer loop acceleration command output for this cycle, that is, it is set to... If steps S6.1 and S6.2 are implemented simultaneously, it is preferable to use step S6.1 as the default implementation for low computing power, and switch to step S6.2 when a range conflict is detected or stronger hierarchical constraint processing is required, thereby forming a dual implementation method that is available in real time with low computing power and switchable with high performance.

[0125] In one specific implementation, the method further includes: when the actual distance is detected to be less than the corresponding minimum safe distance, controlling the relevant UAV to perform emergency evasion in a radial direction away from the threatening target until the safe state is restored and then returning to the normal control process.

[0126] The final outer loop acceleration command obtained in step S6.1 or step S6.2 The command is sent to the inner loop of the drone for execution, and will be executed in the next control cycle. Reacquire the state and neighbor information, and repeat steps S2 to S6 to form a rolling closed-loop control. It should be noted that the safety constraint construction and solution of the present invention are based on the relative position and relative velocity observed / estimated in the current cycle, which meets the characteristics of real-time online calculation. At the same time, through the structure of the expected input first being corrected by the directionality of the collision cone and the control barrier function, and then strictly satisfying the constraints through safety filtering, the operation formation can maintain the slot performance while having verifiable collision avoidance safety and manned-machine priority protection capabilities.

[0127] On the other hand, based on the same inventive concept, embodiments of this application also provide a manned and unmanned aerial vehicle (UAV) collaborative formation collision avoidance control system, the system comprising: The initialization module is used to define mission instructions for the UAV based on the motion state of the manned aircraft, establish the UAV dynamics model, and initialize the hierarchical safety parameters. The design module is used to design the nominal controller and solve for the nominal control commands for each UAV to track the mission instructions. The screening module is used to assess the collision risk of each drone using the collision cone principle and screen out targets with imminent threats. The conversion module is used to convert collision avoidance safety conditions into linear constraints on UAV control commands for the imminent threat target based on the control barrier function, and to group the constraints according to the threat type; A synthesis module is used to synthesize the nominal control instruction and a safety correction term constructed based on the linear constraints to form the desired control instruction; The execution module is used to solve online for the final control command that satisfies all the linear constraints and execution constraints, with reference to the desired control command, and output the execution command.

[0128] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0129] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for collision avoidance control in a coordinated formation involving manned and unmanned aircraft, characterized in that, The method employs a layered control architecture that separates the performance layer and the security layer, and includes the following steps: S1. Define mission instructions for the UAV based on the motion state of the manned aircraft, establish a UAV dynamics model, and initialize graded safety parameters; S2. Design a nominal controller to solve for the nominal control commands used by each UAV to track the mission instructions; S3. Using the collision cone principle, assess the collision risk of each drone and identify targets with imminent threats; S4. Based on the control barrier function, the collision avoidance safety conditions for the imminent threat target are transformed into linear constraints on the UAV control commands, and the constraints are grouped according to the threat type; S5. Synthesize the nominal control command with the safety correction term constructed based on linear constraints to form the desired control command; S6. Using the desired control command as a reference, solve online for the final control command that satisfies all linear constraints and execution constraints, and output the execution command.

2. The method according to claim 1, characterized in that, S1 includes the following steps: In the manned aircraft system, each UAV is pre-assigned a work slot offset, and the position and speed of each UAV are generated by converting to the inertial coordinate system according to the manned aircraft status. A discrete dynamic model of the outer loop of the UAV based on a second-order integrator is established, and input amplitude constraints are applied. Leveled safety parameters are set for drone-to-manned and drone-to-drone interactions, with the safety parameter value for drone-to-manned interaction being greater than the corresponding parameter value for drone-to-drone interaction; subsequent risk assessment is only performed on objects within the trigger distance.

3. The method according to claim 1, characterized in that, S2 includes the following steps: A state vector is constructed using the UAV's position and velocity, and a discrete dynamics model is used as the prediction model. Based on the assumption of constant speed or constant attitude extrapolation of the manned UAV's state, a slot reference state sequence is generated in the prediction time domain. Construct a quadratic cost function with the objectives of minimizing slot tracking error, control energy, and control increment variation, and apply input magnitude constraints in the prediction time domain; Solve the linear model predictive control optimization problem, and use the first term of the obtained optimal control sequence as the nominal outer loop acceleration command.

4. The method according to claim 1, characterized in that, S3 includes the following steps: Calculate the relative position, relative speed, and relative distance between the local machine and the threat target, and match the corresponding security parameters according to the interaction type; Calculate the collision cone half angle based on the relative distance and minimum safe distance; calculate the angle between the relative velocity and the relative position; construct the collision cone safety margin function and calculate the predicted collision time; The trigger neighbor set is filtered based on the trigger distance, and the most urgent threat target subset is selected from the set based on the safety margin function value, proximity and predicted collision time.

5. The method according to claim 1, characterized in that, S4 includes the following steps: The collision cone safety margin function is defined as a barrier function, and different control barrier function gain coefficients are matched for different interaction types. Apply a first-order control barrier function condition to the barrier function, requiring its time derivative and function value to satisfy a preset linear relationship; expand the time derivative by combining the relative motion relationship and the assumption that the acceleration of the adjacent machine is zero; The standard form of the linear inequality constraint for the machine's outer loop acceleration command is derived, and these constraints are grouped according to interaction type into UAV-maned aircraft constraint group and UAV-UAV constraint group.

6. The method according to claim 1, characterized in that, S5 includes the following steps: The nominal outer ring acceleration command is defined as the initial direction quantity; Based on the gradient direction and constraint constant term of the linear inequality constraint, a safety correction term is constructed to increase the safety margin. The nominal outer-loop acceleration command is added to the safety correction term to synthesize the desired outer-loop acceleration command, and higher weights are given to the threats of UAVs and manned aircraft when constructing the correction term.

7. The method according to claim 1, characterized in that, The S6 provides two online solution algorithms: the non-optimized amplitude adjustment method and the quadratic programming optimization method.

8. The method according to claim 7, characterized in that, The non-optimized amplitude adjustment method specifically includes: The desired outer loop acceleration command direction is fixed; Transform all linear inequality constraints into one-dimensional inequalities concerning a single control amplitude; By combining all one-dimensional inequalities and input limits, the total feasible range of the control amplitude is determined; Within this range, select the value closest to the desired amplitude and combine it with a fixed direction to obtain the final command; when a conflict in the range involves interaction between the UAV and the manned aircraft, activate the emergency safety mode.

9. The method according to claim 7, characterized in that, The quadratic programming optimization method specifically includes: Construct an optimization problem with the objective function of minimizing the deviation between the final instruction and the desired instruction; Apply hard constraints to the interaction between UAVs and manned machines; apply soft constraints to the interaction between UAVs and other UAVs by introducing slack variables. Solving this convex quadratic programming problem yields the final instruction; when the problem becomes infeasible due to hard constraints, an emergency safety mode is activated.

10. The method according to claim 1, characterized in that, The method further includes: When the actual distance is detected to be less than the corresponding minimum safe distance, the relevant drone is controlled to take emergency evasive action in a radial direction away from the threatening target until the safe state is restored and then the normal control process is returned.

Images