Unmanned aerial vehicle flight path distribution method and device, medium and equipment

Abstract

The invention discloses an unmanned aerial vehicle flight path distribution method and device, a medium and equipment, and relates to the field of unmanned aerial vehicle control, and the method comprises the steps: building a multi-layer virtual pipeline model of a parcel no-fly airspace based on a distribution airspace of multiple unmanned aerial vehicles; multiple unmanned aerial vehicles in the multi-layer virtual pipeline model are regarded as ideal flowing particles in a virtual pipeline, a Laplace partial differential equation is adopted, a potential function and a flow function of the multiple unmanned aerial vehicles in an ideal fluid flow field of the virtual pipeline are generated, the boundary and the geometrical shape of the virtual pipeline are determined, and the virtual pipeline used for flight of the unmanned aerial vehicles is obtained; based on the state set, the action set, the state transition probability and the cost function of the unmanned aerial vehicles in the virtual pipeline, constructing a dynamic planning strategy of the multiple unmanned aerial vehicles with the minimum path distance and the minimum cost as targets; and based on a dynamic planning strategy of the multiple unmanned aerial vehicles, allocating an optimal flight path to a space access request of the multiple unmanned aerial vehicles in the virtual pipeline, and realizing rapid allocation of the optimal flight path of the unmanned aerial vehicles.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle control technology, and in particular to a method, device, medium and equipment for allocating a flight path of an unmanned aerial vehicle. Background Art

[0002] As an important carrier of digital low-altitude technology, drones are widely used in areas such as package delivery, aerial photography and air rescue. With the promotion of the low-altitude economy, the core competitiveness of the drone industry continues to increase, and drones will take on a large number of orders in terminal delivery services, which shows the broad prospects of drones in the field of logistics and distribution.

[0003] In the past few decades, scholars have proposed a variety of path planning methods to help drones avoid obstacles. Mathematical programming methods such as mixed integer linear programming, nonlinear programming, and dynamic programming are based on rigorous mathematical theories and are suitable for smaller-scale problems, but as the scale of the problem increases, the calculation time increases exponentially. Secondly, the artificial potential field method is known for its real-time performance, but when the target gravity is equal to the obstacle repulsion, it is easy to fall into the local optimum and may not even be able to obtain a feasible solution. Graph theory methods such as random path graph algorithm, Voronoi diagram algorithm, and A* algorithm can effectively find the optimal path, but it takes a long time in complex search space.

[0004] As the number of drones continues to grow, especially in urban environments, the safety of low-altitude airspace is becoming increasingly severe. Drones flying in cities face a high risk of path conflicts and mid-air collisions. To address this problem, a variety of methods have been developed in recent years, mainly divided into reactive collision avoidance methods and active collaborative path generation methods. In reactive collision avoidance methods, each drone only plans the optimal path to avoid environmental threats without considering collision avoidance and time coordination with other drones. When a collision risk is detected, local replanning is performed to avoid collision. A variety of algorithms based on reactive collision avoidance have been proposed, such as optimal reciprocal collision avoidance, distributed reactive collision avoidance, and intelligent algorithms. Reactive collision avoidance methods are widely used in practical applications due to their fast computational speed. However, such algorithms only plan local collision avoidance paths and may fall into local optimality, causing drones to fail to reach their targets. Active collaborative path generation methods are divided into coupling methods and decoupling methods. Coupling methods design a unified path planner for multiple drones, covering the start and end points of all drones, and plan collision-free collaborative paths. Reduction-based coupling methods obtain cooperative paths by simplifying the collaborative problem to a well-studied problem, such as linear programming. Search-based coupling methods, such as conflict-based search, transform the multi-UAV collaborative path planning problem into a global single-agent path search to generate collaborative paths for all UAVs. Although coupling methods are usually able to find the optimal path, their computation time may increase exponentially with the increase in the number of collisions between UAVs and the expansion of the dimension and scope of the planning environment, resulting in a slow speed in assigning flight paths to UAVs. Summary of the invention

[0005] The present invention provides a method, device, medium and equipment for allocating a UAV flight path, which are used to solve the above-mentioned problem existing in the prior art, that is, how to improve the speed of allocating a UAV flight path in the prior art. The present invention provides a method for allocating a UAV flight path, which includes:

[0006] Based on the delivery airspace of multiple drones, a multi-layer virtual pipeline model of parcel no-fly airspace is constructed;

[0007] The multiple UAVs in the multi-layer virtual pipeline model are regarded as particles of ideal flow in the virtual pipeline. The Laplace partial differential equation is used to generate the potential function and stream function of the ideal fluid flow field of multiple UAVs in the virtual pipeline, determine the boundary and geometric shape of the virtual pipeline, and obtain the virtual pipeline for UAV flight.

[0008] Based on the position and state of the drone in the virtual pipeline, the state set is obtained. Based on the delivery action, the action set is obtained. Based on the probability of the drone transferring from the current state to the next target state in the action set, the state transition probability is obtained. Based on the path distance, time consumption and the cost of avoiding conflicts with other drones, the cost function is obtained. According to the state set, action set, state transition probability and cost function, a dynamic programming strategy for multiple drones is constructed with the goal of minimizing path distance and minimum cost.

[0009] Based on the dynamic planning strategy of multiple UAVs, the optimal flight paths are allocated to the spatial access requests of multiple UAVs in the virtual pipeline.

[0010] Optionally, the no-fly airspace is wrapped by using a fluid; wherein the no-fly airspace includes a building area and a restricted flight area.

[0011] Optionally, the action set includes moving forward in the virtual pipeline, staying at the current position next time, moving to an adjacent higher altitude, moving to an adjacent lower altitude, and unloading cargo at a target point.

[0012] Optionally, the state transition probability specifically includes:

[0013] The state transition probability is obtained using the following formula:

[0014]

[0015] Among them, P a (s, s') represents the transition from the current state s∈S to the state s'∈S under action a∈A. It represents the expected next state of each state s∈S under action a∈A, and P is the state transition probability.

[0016] Optionally, the cost function is obtained according to the distance between the next state and the target state, combined with the path conflict cost and the penalty term, specifically including:

[0017] The cost function is obtained using the following formula:

[0018]

[0019] Among them, d(s, s g ) is the distance between the next state and the target state, J 0 (s, a) is the path conflict cost, H is the penalty term, λ is the adjustment coefficient, and J(s, a) is the cost function.

[0020] Optionally, the dynamic programming strategy of multiple UAVs with the goal of minimizing path distance and minimum cost is constructed according to the state set, action set, state transition probability and cost function, specifically including:

[0021] Constructing a state machine, including a state for determining the state of the delivery airspace, including a first non-terminal state for determining whether a new drone submits an airspace use request, a second non-terminal state for determining whether a drone fails or a temporary no-fly airspace appears, a third non-terminal state for determining whether to redefine the virtual pipeline, and a fourth non-terminal state for determining whether to update the cost function and the state transition probability;

[0022] Determine in turn whether the non-terminal state of the state machine is triggered;

[0023] When the second non-terminal state or the third non-terminal state of the state machine is triggered, the state set, state transition probability and cost function are updated, and then the collision-free UAV planning strategy is updated by solving the dynamic programming DP equation;

[0024] When the first non-terminal state or the fourth non-terminal state of the state machine is triggered, the cost function and the state transition probability are updated, and the collision-free UAV planning strategy is updated by solving the dynamic programming DP equation;

[0025] When the non-terminal state of the state machine is not triggered, there is no need to update the current collision-free drone planning strategy.

[0026] Optionally, the updating of the collision-free UAV planning strategy by solving the dynamic programming DP equation specifically includes:

[0027] The updated collision-free UAV planning strategy is obtained using the following formula:

[0028]

[0029] Where S is the state set, A is the action set, V*(s′) is the optimal total cost from state s′, π*(s) is the updated collision-free UAV planning strategy, J(s,a) is the cost function, γ is the weight coefficient, and P a (s, s') represents the state transition probability from the current state s∈S to the state s'∈S under action a∈A.

[0030] The present invention provides a UAV flight path allocation device, comprising:

[0031] A multi-layer virtual pipeline model building module is used to build a multi-layer virtual pipeline model of the no-fly airspace for parcels based on the delivery airspace of multiple drones;

[0032] An acquisition module is used to regard multiple UAVs in the multi-layer virtual pipeline model as particles of ideal flow in the virtual pipeline, use Laplace partial differential equations to generate potential functions and stream functions of the ideal fluid flow field of multiple UAVs in the virtual pipeline, determine the boundary and geometric shape of the virtual pipeline, and acquire the virtual pipeline for UAV flight;

[0033] The allocation strategy building module is used to obtain a state set based on the position and state of the drone in the virtual pipeline, obtain an action set based on the distribution action, obtain the state transition probability based on the probability of the drone transferring from the current state to the next target state in the action set, obtain the cost function based on the path distance, time consumption and the cost of avoiding conflicts with other drones, and construct a dynamic programming strategy for multiple drones with the goal of minimizing path distance and minimum cost based on the state set, action set, state transition probability and cost function;

[0034] The allocation module is used to allocate optimal flight paths for spatial access requests of multiple UAVs in the virtual pipeline based on the dynamic planning strategy of multiple UAVs.

[0035] The present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned UAV flight path allocation method is implemented.

[0036] The present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned UAV flight path allocation method when executing the program.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a method for allocating flight paths of unmanned aerial vehicles, which generates an obstacle-free virtual pipeline geometry that safely surrounds buildings and no-fly zones with low computational cost by solving the Laplace partial differential equation; at the same time, by optimizing the pipeline generation method using an ideal fluid model, the spatial utilization rate of the pipeline is improved, and the maximum feasible path of the entire airspace is obtained; at the same time, the dynamic programming strategy of multiple unmanned aerial vehicles proposed in the present application manages the access of unmanned aerial vehicles to the pipeline through first-come, first-served priority sorting, and based on the dynamic programming strategy of multiple unmanned aerial vehicles, an action sequence with the optimal total cost is provided starting from the current state in the state set as the optimal strategy in the dynamic programming strategy of multiple unmanned aerial vehicles, and the airspace in the virtual pipeline requested by the unmanned aerial vehicle is allocated by adopting the optimized dynamic programming strategy of multiple unmanned aerial vehicles; by determining whether to optimize the dynamic programming strategy of multiple unmanned aerial vehicles by the trigger state of the non-terminal of the state machine, the flight path of the unmanned aerial vehicle can be quickly allocated, so as to carry out distribution work, which can effectively avoid path conflicts and ensure the safe and efficient operation of unmanned aerial vehicles in the virtual pipeline. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0039] Figure 1 A flowchart of a method for allocating a flight path for a drone provided by an embodiment of the present invention;

[0040] Figure 2 A multi-layer virtual pipeline model for drone traffic management provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of K-plane discretization provided in an embodiment of the present invention;

[0042] Figure 4 A virtual pipeline under a flow field provided by an embodiment of the present invention;

[0043] Figure 5 A state machine framework diagram for optimizing dynamic planning strategies for managing multiple drones provided by an embodiment of the present invention;

[0044] Figure 6 A schematic diagram of a four-layer virtual pipeline provided by an embodiment of the present invention;

[0045] Figure 7 50 drone delivery maps provided for embodiments of the present invention;

[0046] Figure 8 100 drone delivery maps provided by the embodiment of the present invention;

[0047] Fig. 9 150 drone delivery maps provided for embodiments of the present invention;

[0048] Fig.10 200 drone delivery maps provided for embodiments of the present invention;

[0049] Fig.11 300 drone delivery maps provided for embodiments of the present invention;

[0050] Fig.12 400 drone delivery maps provided for embodiments of the present invention;

[0051] Fig.13 A schematic diagram of a computer device for a method for allocating a flight path for a drone provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Figure 1 is a flow chart of a method for allocating a flight path of a UAV provided by an embodiment of the present invention. Figure 1 As shown, a method for allocating a flight path of a drone shown in this embodiment includes:

[0054] S1: Based on the delivery airspace of multiple drones, a multi-layer virtual pipeline model of parcel no-fly airspace is constructed.

[0055] S2: The multiple UAVs in the multi-layer virtual pipeline model are regarded as particles of ideal flow in the virtual pipeline. The Laplace partial differential equation is used to generate the potential function and stream function of the ideal fluid flow field of multiple UAVs in the virtual pipeline, determine the boundary and geometric shape of the virtual pipeline, and obtain the virtual pipeline for UAV flight.

[0056] Optionally, the no-fly airspace is wrapped by using a fluid; wherein the no-fly airspace includes a building area and a restricted flight area.

[0057] For example, drones can be considered as particles in an ideal flow, which move in a virtual tube that encloses the no-fly airspace area. The low-altitude airspace is represented by a finite number of parallel planes, each of which is an xy slice perpendicular to the z-axis. The no-fly zone represents buildings and restricted flight areas, such as Figure 2 As shown. The floor is represented by K∈R 2 It is represented and divided into a no-entry zone and a no-exit zone, where the no-entry zone is the projection of buildings and obstacles on the floor.

[0058] Assume that floor K contains 1 ,...,O n The n forbidden areas are defined by the disjoint closed sets of , and the complex variable z = x + iy is used to represent the position on the floor K. Obstacle O i to j The specific calculation of the distance of the center point is as follows:

[0059]

[0060] Among them, c oi =(x oi ,y oi ) indicates obstacle O i The center point of

[0061] When oioj <d threshold When i and O j Merge into new obstacle O ij The new obstacle set after merging is represented as O = {O 1 , O 2 , ..., O k}, where k≤n.

[0062] By defining the UAV coordinates as a compact set K∈R 2 The ideal fluid flow pattern on the surface can be obtained by defining the conformal mapping to obtain the potential function Φ(x, y) and stream function Ψ(x, y) of the ideal fluid flow field. The specific calculation formula is as follows:

[0063] f(z)=Φ(x,y)+iΨ(x,y)

[0064] Among them, Φ(x,y) and Ψ(x,y) satisfy the Laplace partial differential Laplace PDE equation and the Cauchy-Riemann condition:

[0065]

[0066]

[0067] Furthermore, the i th The x and y components of each UAV are constrained to follow the flow curve Ψ defined as i slide:

[0068]

[0069] Among them, i represents the flow curve, xi(t 0 ) and y i (t 0 ) is when the i th When a UAV passes through the boundary point and enters K, the i-th th The position of the UAV at the reference time t 0 The x and y components of .

[0070] Illustratively, analytical and numerical methods may be used to define Φ(x,y) and Ψ(x,y) on K, as described below.

[0071] Analytical solution: Define O 1 ,...,O n The defined stay-out or no-fly airspace regions can be safely wrapped by defining Φ and Ψ as the real and imaginary parts of complex functions:

[0072]

[0073] Among them, z i =x i +jy i and r i >0 respectively represent the nominal position and size of the i-th forbidden zone. i It must be large enough so that the i-th obstacle is safely enclosed by the closed domain generated by the analytical function f(z) defined by the above formula.

[0074] Numerical solution: When the environment contains any obstacles with non-circular cross-sections, the finite difference method is used to assign Ψ in the motion space. The K layer is evenly discretized into small areas using the graph G(v,ε), with increments in the x and y directions of Δx and Δy, respectively. The nodes and edges of G are defined by v = {1, ..., m} and ε∈v×v, respectively.

[0075] The set v = {1, ..., m} can be expressed as v = v B v I v o , where the disjoint sets v B ={1, ..., m B},v I ={m B +1, ..., m B +m I} and v O ={m B +m I +1, ..., m}, outer boundary control nodes, inner control nodes and obstacle nodes respectively. Figure 3 Demonstrated by represents the uniform discretization of the rectangular domain K with respect to its boundary.

[0076] For example, suppose the goal of the drone is to move safely from left to right, at v B and v o The following constraints are imposed on Ψ:

[0077]

[0078] Assume there are k obstacles, through the set k o = {1, ..., k} to identify obstacles, then v o It can be expressed as:

[0079]

[0080] Among them, u j is defined as all nodes contained in the jth obstacle. For obstacle j∈ko By definition:

[0081]

[0082] As the nominal position of the obstacle j∈k o , the flow value of the obstacle node is defined as follows:

[0083]

[0084] Assume that the K-layer airspace is uniformly discretized, and node i∈v I The flow value function at is updated to:

[0085]

[0086] Among them, ix,1 and ix,2 is the Ψ value at the adjacent node in the x direction, that is, (i x，1 ,i),(i x，2 , i)∈ε. Similarly, Ψ iy,1 and iy,2 is the Ψ value at the adjacent node in the y direction, that is, (i y，1 ,i),(i y，2 ,i)∈ε. p i is a node i∈v I The flux at, if i∈v I ∪v O , then p i =0.

[0087] By setting Ψ=[Ψ 1 ...Ψ m ] T Defined as a vector of aggregate node stream values, the updated stream value function can be organized as

[0088] -LΨ(t)+p(t)=0

[0089] Among them, i (t) = {Ψ(x i ,y i , t)|i∈v}, is the flux vector, is the Laplacian matrix of the coverage graph G with (i,j) entries, as follows:

[0090]

[0091] where deg(i) is the in-degree of node i, and the multiplicity of L with eigenvalue 0 is equal to the number of the largest reachable vertex set, i.e., the multiplicity of zero eigenvalue is the number of trees required to cover the graph G. Therefore, the matrix L has m equal to 0 B+m O eigenvalues, whose rank is mm B +m O .

[0092] Generation of virtual pipeline: Decompose the 3D environment into l layers, composed of K 1 ,...,K l , corresponding to different altitudes. K i ∈R 2 is a horizontal floor parallel to the xy plane, height z = h i .

[0093] set up Forbidden zone O j In K i Using numerical methods, we can obtain K i The stream function Ψ on i (x,y) to safely exclude And keep the discrete space:

[0094]

[0095] We get a finite number of pipes, bounded by Ψ i A horizontal curve is obtained when (x,y) is a constant.

[0096] Finishing section: For any point a∈T, the following formula is used to obtain the finishing section passing through a:

[0097]

[0098] Where C(a)=C(a u )=C(a d ) is called the finishing line or finishing section, such as Figure 4 As shown, where a is an arbitrary point on the finishing line. In addition, point a u (a), a d (a)∈R 2 is the distance between the finishing section C(a) and the tube boundary The intersection point, i.e. a u (a) = {x∈C(a):ψ(x i，d ,y i，d )},a d (a) = {x∈C(a):ψ(x j，d ,y i，d )}; the width of C(a) is represented by h(a), and the specific calculation formula of h(a) is as follows:

[0099]

[0100] The width of the virtual tube is further defined as The middle point m(a) of the finishing section C(a) is obtained using the following formula:

[0101]

[0102] Among them, m(a)∈C(a).

[0103] In this step, by solving the Laplace equation, the geometric shape of an obstacle-free virtual pipeline can be generated at a low computational cost, and a unidirectional, multi-level virtual pipeline can be constructed; by using an ideal fluid model to generate a virtual pipeline covering all feasible areas, the space utilization efficiency can be improved.

[0104] S3: Based on the position and state of the UAV in the virtual pipeline, the state set is obtained. Based on the delivery action, the action set is obtained. Based on the probability of the UAV transferring from the current state to the next target state in the action set, the state transition probability is obtained. Based on the path distance, time consumption and the cost of avoiding conflicts with other UAVs, the cost function is obtained. According to the state set, action set, state transition probability and cost function, a dynamic programming strategy for multiple UAVs is constructed with the goal of minimizing path distance and minimum cost.

[0105] Exemplarily, a dynamic programming DP model is defined to maximize the availability of low-altitude airspace by optimizing the allocation of drones in traffic pipes. DP includes a tuple of the following elements (S, A, P, J, Y), where S is a finite state set; A is a finite action set; P is a state transition probability; J is a cost function; and Y is a discount factor, Y∈[0,1].

[0106] Optionally, the action set includes moving forward in the virtual pipeline, staying at the current position next time, moving to an adjacent higher altitude, moving to an adjacent lower altitude, and unloading cargo at a target point.

[0107] Optionally, the state transition probability is obtained using the following formula:

[0108]

[0109] Among them, P a (s, s') represents the transition from the current state s∈S to the state s'∈S under action a∈A. It represents the expected next state of each state s∈S under action a∈A, and P is the state transition probability.

[0110] Optionally, based on the distance between the next state and the target state, combined with the path conflict cost and the penalty term, the cost function is obtained using the following formula:

[0111]

[0112] Among them, d(s, s g ) is the distance between the next state and the target state, J 0 (s, a) is the path conflict cost, H is the penalty term, and λ is the adjustment coefficient.

[0113] From S×A→R + Define the cost function J, where J(s,a) represents the numerical cost assigned to state s∈S under action a∈A.

[0114] like Figure 5 As shown, the state machine UTM specifically includes two terminal states (terminal state 1 and terminal state 2) and four non-terminal states (NT1-NT4) for dynamically managing the allocation of virtual pipelines. If no non-terminal state is triggered, the current collision-free drone planning strategy is considered acceptable. However, if the second non-terminal state NT2 or the third non-terminal state NT3 is triggered, the following operations are performed: the affected area (such as a failed drone or a temporary no-fly zone) needs to be defined as a temporary avoidance area, and the virtual pipeline is replanned, and the state set, transition probability, and cost function are updated to reflect the new airspace configuration; then the optimal strategy π is recalculated by solving the DP equation. * (s). If the first non-terminal state NT1 or the fourth non-terminal state NT4 is triggered, there is no need to modify the state set or the action set, but the cost function and the state transition function need to be adjusted, and the updated strategy is obtained by solving the DP equation; after obtaining the updated strategy, the state machine is updated to terminal state 1; when the non-terminal state of the state machine is not triggered, there is no need to update the current collision-free UAV planning strategy, and the state machine remains in terminal state 2. The updating method of the collision-free UAV planning strategy proposed in the present invention can dynamically adapt to changes in airspace conditions while maintaining the safety and operating efficiency of the UAV system.

[0115] For example, the DP value function + Definition corresponding to the sequence of states s = (s(1), ..., s(t), ...) and actions a = (a(1), ..., a(t), ...):

[0116]

[0117] Using the value iteration algorithm, iteratively calculate the value function V i (s):S→R + , where i represents the number of iterations, the value function V can be updated in the following way i (s):

[0118]

[0119] V i (s) converges monotonically to V in polynomial time * (s). The threshold δ specifies the numerical convergence requirement for the value in each state:

[0120]

[0121] Optimal strategy π * (s) is defined as providing the optimal total cost V starting from state s * (s) action sequence, and is obtained by the following formula:

[0122]

[0123] Under the premise of ensuring the safety of drone collaboration, it is required that every two points in the low-altitude airspace are "reachable". In order to ensure the safety of the state machine UTM, all air corridors on the same floor are required to have the same direction of movement. This requires at least four floors to authorize movement in the east, north, west and south traffic flow directions. Because the air corridor may have been assigned to some drones that are already using the airspace, without loss of generality, the low-altitude airspace is projected on four layers (n l =4) on, by C 1 ,...,C 4 Indicates that: the streamline is at C 1 and C 3 elongated along the x-axis; the streamline is at C 2 and C 4 Elongated along the y-axis.

[0124] For example, L i Defined as Determined In addition, the present invention denotes a finite set of discrete values ​​of x and y coordinates by X and Y, respectively. The state set S is finite and is defined by the following formula:

[0125]

[0126] Where × represents the Cartesian product symbol, and the task state T = {t goal , t store}Indicates whether the drone is currently in a delivery mission or returning to the warehouse.

[0127] From the action set A = {a 1 , a 2 , a 3 , a 4 , a 5} defines five possible actions, with the following functions: a 1 Represents moving forward on the current path; a 2Indicates staying at the current position next time; a 3 Represents movement to an adjacent higher altitude; a 4 represents a move to an adjacent lower altitude; a 5 Represents unloading of cargo at the destination point.

[0128] Since dynamic programming (DP) is used for time planning, the transition between states is deterministic, which in turn means that the probability P a (s,s') is a binary variable. In order to obtain the transition probability Pa(s,s'), first define the expected next state as follows:

[0129] For example, using :S×A→S is used to define the expected next state of each state s∈S under action a∈A. The state transition is defined as:

[0130]

[0131] Exemplarily, by defining the subset So∈S as the set of inaccessible states that have been assigned to existing drones. Given a target state s g , as the target destination of the drone, the cost J(s,a) is defined as follows:

[0132]

[0133] Among them, s g is the final or target state of the new drone,

[0134]

[0135] Among them, d(s,s g ) is the next state s' and the target state s g The distance between them, where x(.), y(.) and z(.) represent the state s∈S or s g ∈S, and the path conflict cost J 0 (s,a) is defined as:

[0136]

[0137] By adding an additional penalty term H, the drone is encouraged to return to the warehouse O as soon as possible. When the drone does not return to the warehouse, an additional cost is added to prioritize completing the task and returning to the warehouse to avoid being stranded outside for a long time, which is expressed as:

[0138]

[0139] In the present invention, the present invention selects the discount factor Y=1 to optimally allocate the aerial pipelines to the drones. The optimal strategy π obtained by* (s) are assigned by value iteration.

[0140] S4: Based on the dynamic planning strategy of multiple UAVs, the optimal flight paths are allocated to the spatial access requests of multiple UAVs in the virtual pipeline.

[0141] In this step, the DP method is used to optimally allocate four layers of virtual pipelines to UAV systems. The computational cost of real-time updates is kept within a manageable range because most pipelines are already allocated to existing UAV systems and the state transitions of DP are updated only when new requests are submitted. Therefore, the DP method is able to dynamically allocate a pipeline to each UAV system at each request.

[0142] For example, in order to verify the airspace allocation and path planning in the logistics distribution scenario, the present invention sets up a small virtual environment containing four layers of pipes. The heights of each layer of pipes are z=11, z=12, z=13, and z=14 respectively. The environment contains multiple obstacles with different lengths, widths, and heights. Through the spatial planning method proposed by the present invention, fixed airspace corridors are defined for each layer, which can safely wrap around obstacles and avoid conflicts, such as Figure 6 shown.

[0143] In the simulation, the dynamic programming (DP) method is used to allocate the airspace requested by the drones, and the airspace usage of existing drones is given priority. Newly requested drones are assigned paths on a first-come, first-served basis. The state machine contains two terminal states and four non-terminal states to manage the update of dynamic programming and the security of airspace allocation. Figure 7 , Figure 8 and Fig. 9 The final trajectory distribution is shown when 50, 100, and 150 drones perform the mission.

[0144] Specifically, Figure 7 Take this as an example to illustrate: Figure 7 (a) shows the task execution effect of the algorithm proposed in the present invention to construct air traffic through a virtual pipeline; Figure 7 (b) shows the situation of using streamlines directly to complete the task (compared with Algorithm 1); Figure 7 (c) considers the performance of the UAV completing the task through the streamline (compared with Algorithm 2) under different starting point conditions. In the simulation, the UAVs are evenly divided into five groups and released in sequence, where the red points represent the target points and the blue points represent the blocked points; Figure 8 and Fig. 9 Can be followed with Figure 7The trajectory analysis is performed in the same way. By comparison, it can be found that the algorithm using the virtual pipeline significantly improves the traffic efficiency of drones and reduces the number of congestion points in the airspace. The results of further statistical analysis of 100 sets of experimental data are shown in Table 1.

[0145] Table 1 Small range algorithm comparison table

[0146]

[0147] According to Table 1, it can be concluded that the algorithm of the present invention has the highest return rate and task completion rate in the execution of UAV tasks, and the lowest airspace congestion rate, which fully verifies the superiority and reliability of the algorithm in complex airspace environments.

[0148] The proposed UAV air traffic management algorithm based on virtual pipelines fully demonstrates its significant advantages in complex airspace environments in simulations, especially when considering the minimum safety distance between UAVs (set to 0.5 units), which further verifies the reliability and practicality of the algorithm. In simulation scenarios with 200, 300, and 400 UAVs, Fig.10 (a) Fig.11 (a) and Fig.12 (a) shows the task execution effect of building air traffic through virtual pipelines, which can effectively avoid obstacles and achieve efficient path planning while ensuring that drones maintain a safe distance. Fig.10 (b) Fig.11 (b) and Fig.12 Although the method of using streamlines to complete tasks directly in (b) of 10 is flexible, it is easy to cause path conflicts and airspace congestion in high-density drone scenarios, and it is difficult to meet safety requirements. Fig.11 (c) and Fig.12 (c) further analyzes the performance of the UAV in completing tasks through streamlines under different starting points. The results show that the randomness and instability of path planning are particularly prominent in high-density scenarios.

[0149] Combined with the statistical results of 100 sets of simulation data in Table 2, the algorithm based on virtual pipelines performs well in key indicators such as task completion rate, return rate and airspace congestion rate. Especially in high-density drone scenarios, the algorithm significantly reduces the airspace congestion rate and greatly improves the task completion efficiency. This is due to the algorithm's optimization of the construction and allocation of virtual pipelines through dynamic programming methods, which not only effectively ensures the safety of drones, but also significantly improves airspace utilization.

[0150] Table 2 Small range algorithm comparison table

[0151]

[0152] By optimizing the pipeline generation method and traffic rules, the present invention realizes efficient airspace management in multi-UAV logistics distribution tasks. Simulation experiments verify the superiority and applicability of this method at low computational cost, which can significantly improve the safety and space utilization rate of multi-UAV operations.

[0153] The above is the UAV flight path allocation method provided by one or more embodiments of this specification. Based on the same idea, this specification also provides a corresponding UAV flight path allocation device, including:

[0154] A multi-layer virtual pipeline model construction module, which is used to construct a multi-layer virtual pipeline model of the no-fly airspace for packages based on the distribution airspace of multiple UAVs;

[0155] An acquisition module, which is used to regard multiple UAVs in the multi-layer virtual pipeline model as particles flowing ideally in the virtual pipeline, use the Laplace partial differential equation to generate the potential function and stream function of multiple UAVs in the ideal fluid flow field of the virtual pipeline, determine the boundary and geometry of the virtual pipeline, and obtain the virtual pipeline for UAV flight;

[0156] A distribution strategy construction module, which is used to obtain a state set based on the position and state of the UAV in the virtual pipeline, obtain an action set based on the distribution actions, obtain the state transition probability based on the probability that the UAV transfers from the current state to the next target state among the actions in the action set, obtain a cost function based on the path distance, time consumption, and the cost of avoiding conflicts with other UAVs, and construct a dynamic programming strategy for multiple UAVs with the goal of minimizing the path distance and the lowest cost according to the state set, action set, state transition probability, and cost function;

[0157] An allocation module, which is used to allocate the optimal flight path for the spatial access requests of multiple UAVs in the virtual pipeline based on the dynamic programming strategy of multiple UAVs.

[0158] For the specific limitations of the UAV flight path allocation device, reference can be made to the limitations of the UAV flight path allocation method in the above text, which will not be elaborated here. Each module in the above UAV flight path allocation device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0159] The present invention also provides a computer-readable storage medium, which stores a computer program, and the computer program can be used to execute the above-provided UAV flight path allocation method.

[0160] The present invention also provides Fig.13 The structural diagram of the computer device shown in FIG. Fig.13 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the UAV flight path allocation method provided in the above embodiment.

[0161] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present invention.

Claims

1. A method for allocating flight paths of unmanned aerial vehicles, characterized in that: include: Based on the delivery airspace of multiple drones, a multi-layer virtual pipeline model of parcel no-fly airspace is constructed; The multiple UAVs in the multi-layer virtual pipeline model are regarded as particles of ideal flow in the virtual pipeline. The Laplace partial differential equation is used to generate the potential function and stream function of the ideal fluid flow field of multiple UAVs in the virtual pipeline, determine the boundary and geometric shape of the virtual pipeline, and obtain the virtual pipeline for UAV flight. Based on the position and state of the drone in the virtual pipeline, the state set is obtained. Based on the delivery action, the action set is obtained. Based on the probability of the drone transferring from the current state to the next target state in the action set, the state transition probability is obtained. Based on the path distance, time consumption and the cost of avoiding conflicts with other drones, the cost function is obtained. According to the state set, action set, state transition probability and cost function, a dynamic programming strategy for multiple drones is constructed with the goal of minimizing path distance and minimum cost. Based on the dynamic planning strategy of multiple UAVs, the optimal flight paths are allocated to the spatial access requests of multiple UAVs in the virtual pipeline.

2. The method for allocating a flight path of a UAV according to claim 1, characterized in that: The no-fly airspace is wrapped by using fluid; wherein the no-fly airspace includes a building area and a restricted flight area.

3. The method for allocating a UAV flight path according to claim 1, characterized in that: The action set includes moving forward in the virtual pipe, staying at the current position next time, moving to the adjacent higher altitude, moving to the adjacent lower altitude, and unloading the cargo at the target point.

4. The method for allocating a flight path of a UAV as claimed in claim 1, characterized in that: The state transition probability specifically includes: The state transition probability is obtained using the following formula: Among them, P a (s, s') represents the transition from the current state s∈S to the state s'∈S under action a∈A. It represents the expected next state of each state s∈S under action a∈A, and P is the state transition probability.

5. The method for allocating a UAV flight path as claimed in claim 1, characterized in that: According to the distance between the next state and the target state, the cost function is obtained by combining the path conflict cost and the penalty term, which specifically includes: The cost function is obtained using the following formula: Among them, d(s, s g ) is the distance between the next state and the target state, J0(s, a) is the path conflict cost, H is the penalty term, λ is the adjustment coefficient, and J(s, a) is the cost function.

6. The method for allocating a UAV flight path according to claim 1, characterized in that: The dynamic programming strategy of multiple UAVs is constructed based on the state set, action set, state transition probability and cost function with the goal of minimizing path distance and minimum cost, specifically including: Constructing a state machine, including a state for determining the state of the delivery airspace, including a first non-terminal state for determining whether a new drone submits an airspace use request, a second non-terminal state for determining whether a drone fails or a temporary no-fly airspace appears, a third non-terminal state for determining whether to redefine the virtual pipeline, and a fourth non-terminal state for determining whether to update the cost function and the state transition probability; Determine in turn whether the non-terminal state of the state machine is triggered; When the second non-terminal state or the third non-terminal state of the state machine is triggered, the state set, state transition probability and cost function are updated, and then the collision-free UAV planning strategy is updated by solving the dynamic programming DP equation; When the first non-terminal state or the fourth non-terminal state of the state machine is triggered, the cost function and the state transition probability are updated, and the collision-free UAV planning strategy is updated by solving the dynamic programming DP equation; When the non-terminal state of the state machine is not triggered, there is no need to update the current collision-free drone planning strategy.

7. The method for allocating a flight path of a UAV as claimed in claim 6, characterized in that: The collision-free UAV planning strategy is updated by solving the dynamic programming DP equation, specifically including: The updated collision-free UAV planning strategy is obtained using the following formula: Where S is the state set, A is the action set, V*(s′) is the optimal total cost from state s′, π * (s) is the updated collision-free UAV planning strategy, J(s,a) is the cost function, γ is the weight coefficient, P a (s, s') represents the state transition probability from the current state s∈S to the state s'∈S under action a∈A.

8. A drone flight path allocation device, characterized in that: include: A multi-layer virtual pipeline model building module is used to build a multi-layer virtual pipeline model of the no-fly airspace for parcels based on the delivery airspace of multiple drones; An acquisition module is used to regard multiple UAVs in the multi-layer virtual pipeline model as particles of ideal flow in the virtual pipeline, use Laplace partial differential equations to generate potential functions and stream functions of the ideal fluid flow field of multiple UAVs in the virtual pipeline, determine the boundary and geometric shape of the virtual pipeline, and acquire the virtual pipeline for UAV flight; The allocation strategy building module is used to obtain a state set based on the position and state of the drone in the virtual pipeline, obtain an action set based on the distribution action, obtain the state transition probability based on the probability of the drone transferring from the current state to the next target state in the action set, obtain the cost function based on the path distance, time consumption and the cost of avoiding conflicts with other drones, and construct a dynamic programming strategy for multiple drones with the goal of minimizing path distance and minimum cost based on the state set, action set, state transition probability and cost function; The allocation module is used to allocate optimal flight paths for spatial access requests of multiple UAVs in the virtual pipeline based on the dynamic planning strategy of multiple UAVs.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the method for allocating the flight path of a UAV as described in any one of claims 1 to 7 is implemented.

10. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for allocating the flight path of a UAV as described in any one of claims 1 to 7 is implemented.

Images