Unmanned aerial vehicle autonomous landing path planning method and system based on time-energy

Abstract

The invention provides an unmanned aerial vehicle autonomous landing path planning method and system based on time-energy. The method comprises the following steps: constructing an autonomous landing model with optimal time-energy; designing an autonomous tracking mechanism based on an autonomous landing model: based on a target trajectory prediction sequence, generating a local target through a rolling space, executing local re-planning, taking the local target as a target position, and obtaining an unmanned aerial vehicle trajectory search sequence through a search algorithm; generating an unmanned aerial vehicle target trajectory prediction sequence based on the unmanned aerial vehicle trajectory search sequence; designing an autonomous landing mechanism: designing time and dynamic constraints, considering terminal constraints in autonomous landing, and optimizing an autonomous landing model; the autonomous landing model is solved through NLOPT; and finally, parameters are initialized, and landing is completed through an autonomous landing finite-state machine comprising the autonomous tracking mechanism and the autonomous landing mechanism. The method can adapt to different sites and environments, and can adapt to different conditions by adjusting planning parameters or algorithms.

Description

Technical Field

[0001] The present invention relates to an autonomous landing path planning technology for an unmanned aerial vehicle (UAV), and in particular to an autonomous landing path planning method and system for an unmanned aerial vehicle (UAV) based on time and energy. Background Art

[0002] The technical background of autonomous landing of drones on mobile platforms involves multiple fields such as high-precision positioning, dynamic environment perception, motion prediction and control algorithms. With the continuous expansion of drone application scenarios, especially the increasing demand for landing on dynamic platforms such as ships and mobile vehicles at sea, autonomous landing technology has become a research hotspot. However, there are the following problems with autonomous landing of mobile platforms: First, how to land smoothly and quickly on a dynamic platform. The drone is a typical underactuated system, and the state of the landing platform is unknown and time-varying, which introduces cross-dependent time and space conditions for the trajectory of the drone. In order to cope with this situation, the drone needs to flexibly adjust its flight time during the trajectory generation process. Second, how to deal with sudden changes in complex situations. A mobile platform with sudden changes in motion and severe external interference will cause the original planned trajectory to be abandoned. In order to allow drones to land quickly in complex environments, high-frequency replanning is essential. Third, how to ensure landing safety in high-frequency planning.

[0003] Limited by navigation technology and onboard computing power, the visual servo-based method is the easiest to implement and the most mature method for autonomous landing. This method directly plans the UAV at the control level, directly constructs the state deviation term based on target observation, and then uses PID or Lyapunov's second method to design a controller to complete the autonomous landing of the UAV. The landing method based on visual servo can capture the landing mark and execute the landing in real time, and has the characteristics of high accuracy and strong independence. However, this method often ignores environmental information and is short-sighted. It is difficult to introduce state constraints related to the landing platform. It has high requirements for the environment, requires sufficient lighting and clarity, and needs to be able to identify a suitable landing area. Summary of the invention

[0004] The purpose of the present invention is to provide a time-energy based UAV autonomous landing path planning method and system, which can meet the dynamic constraints of vertical take-off and landing UAVs, can adapt to different sites and environments, and adapt to different conditions by adjusting planning parameters or algorithms.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A time-energy-based UAV autonomous landing path planning method comprises the following steps:

[0007] Construct a time-energy optimal autonomous landing model;

[0008] An autonomous tracking mechanism is designed based on the autonomous landing model to generate an autonomous tracking path for the drone, including:

[0009] Generate local targets through rolling space and perform local replanning, use the local targets as target positions, and obtain the UAV trajectory search sequence through the search algorithm;

[0010] Based on the UAV trajectory search sequence, generate the UAV target trajectory prediction sequence;

[0011] Design an autonomous landing mechanism to land the drone, including:

[0012] Design time and dynamic constraints, and consider the terminal constraints in autonomous landing to optimize the autonomous landing model;

[0013] Solve the autonomous landing model through NLOPT;

[0014] Initialize motion planning parameters, and complete landing through an autonomous landing finite state machine including the autonomous tracking mechanism and the autonomous landing mechanism.

[0015] A time-energy based autonomous landing path planning system for unmanned aerial vehicles, comprising:

[0016] An autonomous landing model building unit, used to build a time-energy optimal autonomous landing model;

[0017] The autonomous tracking mechanism design unit designs an autonomous tracking mechanism based on the autonomous landing model to generate the autonomous tracking path of the UAV, including: generating local targets through rolling space and performing local replanning, taking the local targets as the target positions, obtaining the UAV trajectory search sequence through the search algorithm; based on the UAV trajectory search sequence, generating the UAV target trajectory prediction sequence

[0018] Autonomous landing mechanism design unit, used to design autonomous landing mechanism and realize UAV landing, including: designing time and dynamic constraints, considering terminal constraints in autonomous landing, optimizing autonomous landing model, and solving autonomous landing model through NLOPT;

[0019] The landing unit initializes motion planning parameters and completes the landing through an autonomous landing finite state machine including the autonomous tracking mechanism and the autonomous landing mechanism.

[0020] Compared with the prior art, the present invention has the following notable features:

[0021] (1) The motion planning algorithm proposed in this invention takes into account time adjustability and can dynamically adjust the motion trajectory according to environmental changes or task requirements, thereby improving the system's ability to cope with emergencies;

[0022] (2) The motion planning algorithm proposed in this invention is based on the trajectory optimization method of MINCO, which can perform spatiotemporal joint optimization of path points and motion time;

[0023] (3) The motion planning algorithm proposed in the present invention can be flexibly embedded in the state machine. When adding a new state or transition, only the corresponding code needs to be added without significantly modifying the existing logic, thereby improving the scalability of the system.

[0024] (4) The trajectory generation algorithm proposed in the present invention is designed for vertical take-off and landing UAVs and can meet the autonomous landing requirements of various vertical take-off and landing UAVs, including ordinary rotor UAVs and composite configuration vertical take-off and landing UAVs;

[0025] (5) The motion planning algorithm proposed in the present invention considers both the initial state constraints and the terminal state constraints of the UAV, adopts a hierarchical framework, and iteratively generates trajectories to improve the efficiency of autonomous landing of the UAV. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The figure is a flow chart of the time-energy based UAV landing path planning and optimization algorithm of the present invention.

[0027] Figure 2 This is a flow chart of the autonomous landing finite state machine of the mobile platform of the present invention.

[0028] Figure 3 It is a schematic diagram of the local planning strategy in the present invention.

[0029] Figure 4 Schematic diagrams of two scenarios of collision-induced re-planning and fixed-time re-planning in the present invention.

[0030] Figure 5 This is a comparison chart between the conventional A* algorithm and the dynamically extended A* algorithm. Figure 5 (a) is a schematic diagram of the conventional A* algorithm search. Figure 5 (b) is a schematic diagram of the dynamically expanded A* algorithm search.

[0031] Figure 6 Schematic diagrams of three situations of the UAV landing in the present invention.

[0032] Figure 7 Different from the present invention Degree of drop Schematic diagram of the curve. DETAILED DESCRIPTION

[0033] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] The present invention studies the trajectory generation of autonomous landing of UAV on mobile platform, and proposes an autonomous landing path planning and optimization method considering time adjustability. The MINCO trajectory class is used to plan the path of autonomous landing of UAV with fast speed on mobile platform. This method can meet the dynamic constraints of vertical take-off and landing UAV, combined with Figure 1 , specifically including the following steps:

[0035] Step 1: Build an autonomous landing finite state machine, receive the control signal to start the landing mission or set the waiting time to start the landing mission, and search the sequence according to the drone trajectory of the autonomous tracking mechanism and target trajectory prediction sequence Assist the drone to perform autonomous tracking tasks and complete autonomous landing according to the autonomous landing mechanism;

[0036] Step 2: Build a time-energy optimal autonomous landing model to smoothly transition the drone from a normal flight state to a motion state consistent with the target;

[0037] Step 3: Design an autonomous tracking mechanism based on the autonomous landing model to generate an autonomous tracking path for the drone, including:

[0038] Step 3-1: Generate local targets of the autonomous landing model based on the rolling space, perform local replanning, and obtain the UAV trajectory search sequence through the path search algorithm ;

[0039] Step 3-2: Use the time-energy optimal mobile target trajectory prediction algorithm to determine the target trajectory prediction sequence ;

[0040] Step 4: Design an autonomous landing mechanism to land the drone, including:

[0041] Step 4-1, optimizing the autonomous landing model by time and dynamic constraints;

[0042] Step 4-2, solve the autonomous landing model equations through NLOPT; use NLOPT to solve the unconstrained nonlinear optimization problem. Through the soft constraint method, the hard constraints are converted into penalty functions and added to the objective function to simplify the problem. Select the NLOPT_LD_LBFGS algorithm for solving, and the solved trajectory is converted into the desired state acceptable to the drone controller through the differential flatness tool to achieve precise control;

[0043] Step 5, initialize the motion planning parameters for autonomous landing, execute the autonomous landing finite state machine of step 1, and perform the autonomous landing mechanism and autonomous tracking mechanism during the execution process to achieve autonomous landing.

[0044] The finite state machine is a mathematical model that divides the control target into finite states in order to simplify the algorithm design. Generally, a state machine consists of three parts: a finite state set, an input set, and a state transition set. The autonomous landing mission of the drone involved in this patent is controlled by a state machine on the ROS system. The state machine is responsible for starting the planner and controlling the drone after receiving a signal from the system or reaching the start time.

[0045] The autonomous landing finite state machine starts autonomous landing path planning and controls the UAV after receiving a control signal to start the landing mission or the waiting time for starting the landing mission is reached; after starting the autonomous landing path planning, the UAV enters a waiting state, in which the UAV can receive human instructions; the end time of the waiting state is used as the landing start time, that is, the initial time; after the UAV enters the waiting state, it starts to receive target information, and if the plane distance between the UAV and the target is greater than the tracking distance , the drone is controlled to perform the autonomous tracking task. The drone autonomously tracks the target and shortens the distance between the drone and the target until the plane distance between the drone and the target is less than or equal to the landing distance. ; When the plane distance between the drone and the target is less than or equal to the landing distance , the drone is controlled to perform the autonomous landing mission and complete the autonomous landing mission; when the drone encounters an emergency during the autonomous tracking mission or autonomous landing mission (such as optimization failure, the drone is close to an obstacle), the drone is controlled to enter the hovering state and the drone remains hovering.

[0046] according to Figure 2 As shown in the figure, the autonomous landing finite state machine includes: after the program is started, the drone enters the waiting state, which is the fixed initial state of the drone. In the waiting state, the drone can receive human instructions. Once the mission start command is received, the state machine checks whether the target information is received. If the plane distance between the drone and the mobile platform is greater than the tracking distance , it will enter the autonomous tracking state. The autonomous tracking state is responsible for completing the tracking task without returning to the waiting state.

[0047] Furthermore, for the mobile platform landing mission of the vertical take-off and landing UAV, the planning goal of the autonomous landing model is to smoothly transition the UAV from the normal flight state to a motion consistent with the target motion. In this process, it is necessary to satisfy the dynamic differential constraints, boundary constraints and process constraints determined by the dynamic equations, and consider the control cost and landing time at the same time to achieve time-energy optimization; the autonomous landing model is modeled as a constrained optimization problem, and the optimization goal is to minimize the control cost and landing time. The differential flatness characteristics of the UAV are utilized, and the chain integral linear system is used to simplify the dynamic differential constraints. Finally, under the conditions of satisfying the boundary constraints and process constraints, the optimal control cost is solved to ensure that the UAV landing process is efficient and stable. Therefore, the landing motion planning of the UAV can be regarded as a constrained optimization problem. The objective function of the autonomous landing model is Described as:

[0048] (1)

[0049] in: To adjust the parameters, represents the control cost of the drone, Indicates the landing time. represents the weight matrix, and They represent equality constraints and inequality constraints respectively. Equality constraints include dynamic differential constraints and boundary constraints, and inequality constraints include process constraints.

[0050] The first equality constraint is the dynamic differential constraint of the system. By taking advantage of the differential flatness of the drone, a chain integral linear system is used to simplify the dynamic differential constraint:

[0051] (2)

[0052] in: express Drone status at any moment, express The first derivative of ; , , express OK The zero matrix of columns, express dimensional identity matrix; Indicates that the drone is The location at the moment, , and Respectively exist direction, Direction and Position in direction, , and Respectively The first, second, and third derivatives of ; Indicates the jump of the drone, the value is , used to simplify the control cost ; The system input in equation (2) uses the third-order derivative of displacement , the objective function of the autonomous landing model Simplified to:

[0053] (3)

[0054] The boundary constraints are expressed as:

[0055] (4)

[0056] in: Indicates the landing start time, that is, the initial time; Indicates the state of the drone when it starts landing, that is, the initial state of the drone; Indicates the time when landing is completed. Indicates the status of the drone when it completes landing;

[0057] The process constraints on velocity and acceleration are expressed as:

[0058] (5)

[0059] in: Indicates that the drone is The speed of time, The maximum speed constraint of the UAV during the landing process; Indicates that the drone is The acceleration of time, Represents the maximum acceleration constraint of the drone during the landing process;

[0060] The process constraints on the time trajectory are expressed as:

[0061] (6)

[0062] in: Indicates time trajectory segment, the landing start time Until landing completion time According to the MINCO curve, it is divided into Time trajectory segments, the time trajectory sequence is recorded as , the duration of each track segment must be greater than 0.

[0063] In the UAV landing mission, since the UAV is close to the landing platform, the impact of environmental constraints on its motion planning is usually not considered; in addition, the objective function contains the high-order derivatives of the trajectory, which indirectly reflects the smoothness of the trajectory, so the smoothness constraint can be ignored in the optimization process. In summary, the objective function of the landing optimization problem is determined by equation (3), the equality constraint is determined by equation (4) and equation (2), and the inequality constraint is determined by equation (5) and equation (6).

[0064] Furthermore, the autonomous landing model is optimized through time and dynamic constraints, including:

[0065] The objective function The process constraint optimization about speed is , the objective function The process constraint optimization about acceleration is , the landing process only penalizes the maximum combined velocity and combined acceleration, so the objective function The process constraints on speed and acceleration are optimized as follows: :

[0066] (7)

[0067] (8)

[0068] in: and are the weights of speed penalty and acceleration penalty respectively;

[0069] Landing time Divided into fixed time and relaxation time Two parts, For relaxation time , introduce optimization duration , the relaxation time is optimized to ; For fixed time , introduce the maximum duration , optimizing the fixed time to ; then the landing time will be Optimized for , set the MINCO curve time trajectory sequence The initial value is ;

[0070] Define the unconstrained virtual time series as , for Time track segments Perform differential isomorphism mapping to obtain Unconstrained virtual time :

[0071] (9)

[0072] (10)

[0073] The objective function The process constraint optimization with respect to the time trajectory is ,use Represents the time trajectory sequence Convert to unconstrained dummy time series , then:

[0074] (11)

[0075] in: Represents the path point sequence of the MINCO curve, where This belongs to the prior art, and the derivation process will not be given in detail here.

[0076] A time and dynamics constrained optimization method is proposed, by defining an unconstrained virtual time , the time variable with constraints Transformed into an unconstrained optimization problem. Implemented using differential isomorphism mapping and After the optimization is completed, the actual time can be obtained by back-calculating To adapt to different scenarios, the total landing trajectory time is divided into a fixed part and a relaxed part, and the maximum fixed time is adjusted in real time during optimization. At the same time, the speed and acceleration constraints only penalize the maximum combined speed and combined acceleration to ensure the flexibility and feasibility of trajectory planning.

[0077] Trajectory planning takes the local target as the end point and the current position of the UAV as the starting point for searching. Considering the limited perception range of the UAV, planning outside the rolling space is inefficient, so the path search and trajectory optimization are restricted to the rolling space.

[0078] Furthermore, generating a local target of the autonomous landing model based on the rolling space specifically includes: before the path search begins, setting a global target for the mobile platform to land. When the global target is not in the rolling space ( Figure 3 When the global target is within the dotted circle in the middle, the global target is connected to the current position of the UAV, and the intersection of the connection line and the rolling space is set as the local target, such as Figure 3 As shown; the position and velocity of the local target can be calculated by solving the BVP and then using formula (12).

[0079] The rolling space is centered on the current position of the drone. The circular projection area with a radius of 200 is taken as the target’s current position as the global target, and it is determined whether the global target is in the rolling space. If it is in the rolling space, the global target is taken as the target position, and the next moment position of the drone is obtained through the search algorithm, thereby forming a drone trajectory search sequence. If it is not in the rolling space, first connect the current position of the drone with the global target, take the intersection of the line and the rolling space as the local target, take the local target as the target position, and obtain the next moment position of the drone through the search algorithm, thus forming the drone trajectory search sequence .

[0080] During trajectory execution, the local occupancy grid map is updated using the projection method, and the update frequency is manually set according to the hardware conditions. There are two scenarios that will trigger replanning. First, when the planning module obtains the updated grid map data, it determines the safety of the current execution trajectory based on the latest map. If it is found that the current trajectory intersects or collides with an obstacle, replanning will be triggered to ensure the safety of the drone. Second, at fixed time intervals, the motion planning module will be called to recalculate the tracking points according to the time-energy optimal mobile target trajectory prediction algorithm to update the global and local targets, so that the motion planning can adapt to changes in mobile targets or mission requirements. Figure 4 Both replanning scenarios are shown.

[0081] Obtain the UAV trajectory search sequence when executing the path search algorithm When , the optimal control cost of the UAV is obtained by solving the boundary value method :

[0082] (12)

[0083] in: express The optimal state of the drone at all times; represents a 9-dimensional costate vector, Respectively represent dimensional covariate; , , , , , , , and Indicates that direction, Direction and The unknown constant to be solved in the direction, represents the Hamiltonian function, expressed as:

[0084] (13)

[0085] in: To adjust the parameters; , and Respectively represent displacement The third, second and first derivatives of the UAV, i.e. the jump, acceleration and speed of the UAV, are , and Respectively direction, Direction and direction.

[0086] Further, in obtaining the drone trajectory search sequence The search algorithm searches for the next position of the drone based on its current position, thus forming a drone trajectory search sequence. ; The search algorithm extends the third-order chain dynamics model and considers the cost function of the UAV based on the A-star algorithm;

[0087] Path search algorithm Using motion primitives as graph edges, we expand the third-order chain dynamics model and then generate motion primitives. Because of the differential flatness of the drone, we only need to consider the flat output. ,and You can also use Export, remember The drone status at the moment is , representing the position, velocity, and acceleration in the x, y, and z directions respectively, and the jump of the drone As a control cost , set the expansion step size to ,Will Discrete ,Will Discrete , then the dynamic differential constraint is discretized as:

[0088] (14)

[0089] in: , , represents the direct sum of the matrix on the diagonal; subscript is an integer representing the minimum discrete time times; uniformly discretize the control cost in three-dimensional space , when discrete, the control input is limited to a range, that is, , then there is a uniform discrete control input , represents the set value of the discrete coefficient, express The maximum value in three-dimensional space; the dynamic expansion represented by the state equation is as follows Figure 5 As shown in (b) of , motion constraints are incorporated into the search process by introducing a kinematic model. Figure 5 The conventional A* algorithm and RRT algorithm can only consider obstacle constraints, while the path generated by the method proposed in this patent not only considers obstacle constraints, but also satisfies the motion constraints of the drone.

[0090] Another key method in the path search algorithm , which calculates the cost function during the search process, and then The optimal node is selected according to the calculated cost function.

[0091] Following A*, the cost function of the current position of the drone is expressed as , Indicates the initial position of the drone Go to the current location of the drone The actual cost is actually expressed as the cumulative cost from the starting point to the current node. Indicates the current position of the drone The heuristic cost to the target location is the estimated cost from the current node to the target node; the actual cost of the drone Expressed as discrete control cost And landing time In order to achieve discrete control cost and landing time The trade-off is to set the search objective function Restrictions on landing time Define the minimization of energy-time cost:

[0092] (15)

[0093] in: To adjust the parameters;

[0094] In the path search algorithm, Calculated in each node expansion, based on the discrete control cost and expansion step Defining motion primitive costs for:

[0095] (16)

[0096] in: To adjust the parameters;

[0097] Motion primitive cost based Definition, for a certain The optimal search path composed of motion primitives, and the actual cost of the drone corresponding to the optimal search path It is expressed as:

[0098] (17)

[0099] Indicates The cost of motion primitives, and Respectively The corresponding discrete control cost and expansion step size, ; Another component of the cost function is the heuristic cost. When the heuristic cost is 0, the algorithm will degenerate into the Dijkstra algorithm.

[0100] This patent adopts a cost function that takes into account the actual flight conditions of the drone. Considering it as a two-point boundary value problem, in order to reduce the amount of calculation, the acceleration is relaxed, and then:

[0101]

[0102] (18)

[0103]

[0104] Then the heuristic cost of drone search is It is expressed as:

[0105] (19)

[0106] Get the optimal discrete control cost of the UAV for:

[0107] (20)

[0108] The cost function of the current position of the drone Optimized to:

[0109] (twenty one)

[0110] in: express The corresponding optimal discrete control cost.

[0111] The pseudo code of the path search algorithm is shown in Table 1.

[0112] Table 1 Pseudo code of the path search algorithm

[0113]

[0114] Further, the time-energy optimal mobile target trajectory prediction algorithm is used to determine the target trajectory prediction sequence Specifically include: For the autonomous landing path point sequence of the UAV , which is divided into a prediction trajectory sequence And the search trajectory sequence from the drone to the current mobile platform , based on the UAV trajectory search sequence Determine the target trajectory prediction sequence Since environmental constraints are not considered, the next moment position of the drone is directly searched based on the current position of the drone and the current position of the target, thus forming a drone trajectory search sequence ;

[0115] For the mobile platform in the landing mission, this patent considers that its kinematic model is known, and the movement is fast and energy-saving. The current position of the target can be obtained by using onboard sensors or by using communication. With current speed The goal of the prediction algorithm is to estimate the future The position of the mobile platform after time , is a fixed known time. The time-energy optimal mobile target trajectory prediction algorithm is given. Before introducing the time-energy optimal mobile target trajectory prediction algorithm, this patent first gives the kinematic modeling of the mobile platform. In order to reduce the amount of calculation, this patent adopts a second-order chain integral model, that is, acceleration as the control input.

[0116] During the landing process of the drone, the target initial position and the target initial speed Calculate the predicted duration The estimated target position ; The second-order chain integral model is used to model the target dynamics, and the target acceleration is used as the system input, which is expressed as:

[0117] (twenty two)

[0118] in: express The state variable of the target at time instant, express The first derivative of ; , , express OK The zero matrix of columns, express dimensional identity matrix; Indicates the target The location at the moment, , and Respectively exist direction, Direction and Position in direction, and Respectively The first and second derivatives of ; Represents the target system input, and its value is ;

[0119] Time-energy optimal moving target trajectory prediction algorithm The function (row 2 in Table 2) calculates a rough end point position using the target motion model , the prediction duration is calculated by the target's dynamic model The estimated target position :

[0120] (twenty three)

[0121] It is just an estimate of the final position, and its purpose is to get the heuristic cost to reach the final position;

[0122] System input for the target Using discrete values, it is expressed as:

[0123] (twenty four)

[0124] in: represents the target maximum acceleration constraint, Indicates the set value of the discrete coefficient, which is usually an integer between 3 and 5;

[0125] From the sixth row in Table 2 Function Get The target state after ,based on The state variable of the target at the moment ,choose System input of the target at the moment :

[0126] (25)

[0127] in: represents the cost function of the target, represents the actual cost of the target, Represents the heuristic cost of the target:

[0128] (26)

[0129] in: Represents the adjustment parameter, which can be and The Euclidean distance between them is an indirect substitute; the optimal system input The corresponding state variable Add to target trajectory prediction sequence This step is done by the 7th row of Table 2 Function completed.

[0130] After the algorithm is cycled through, the desired , the pseudo code is shown in Table 2.

[0131] Table 2 Pseudo code of target trajectory prediction algorithm

[0132]

[0133] In trajectory prediction, an ideal situation is that the total trajectory time of the drone landing is equal to the predicted time ( ), so that the UAV lands exactly on the mobile platform. However, in the autonomous landing of the UAV, the fixed landing time may make it impossible to land, especially when the speed of the mobile platform is greater than the speed of the UAV. Fixed to , forcing the UAV to reach the target state is not feasible in some cases. For example, when the speed of the mobile platform is greater than the speed of the UAV, the UAV can never plan a trajectory that can achieve landing. Therefore, a flexible terminal constraint method is proposed to adapt to different situations by relaxing the total trajectory time, such as Figure 6 This method introduces the landing intention coefficient , dynamically adjust the landing strategy according to the plane distance between the UAV and the mobile platform: prioritize tracking when the distance is far, and prioritize landing when the distance is close. By adjusting the predicted target position, ensure that the UAV can land smoothly when it approaches the mobile platform.

[0134] Furthermore, the constraints also include terminal constraints.

[0135] Set the landing destination of the drone as the target trajectory prediction sequence The last position of , when the end point is constrained, only to process;

[0136] Use the landing degree Evaluate whether the drone performs landing action: If , then maintain the state; if , then execute the landing action; if Figure 7 , the degree of decline for:

[0137] (27)

[0138] in: Characterizes the maximum amplitude of the landing relative to the tracking, , Represents the initial time , A constant that characterizes the range of falling speeds. The faster the falling speed, The bigger;

[0139] Build landing level and The connection between them is:

[0140] (28)

[0141] The endpoint constraints for constructing the drone landing are:

[0142]

[0143] (29)

[0144]

[0145] The initialized motion planning parameters include:

[0146] The state machine initializes related parameters: tracking distance d1 and landing distance d2;

[0147] Path search algorithm and trajectory prediction algorithm initialization parameters: target initial position , Target initial speed , UAV initial position , the predicted duration of the target's future trajectory , Target initial state , UAV initial state , UAV maximum speed constraints , UAV maximum acceleration constraint .

[0148] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

[0149] This embodiment also provides a time-energy based autonomous landing path planning system for a UAV, including:

[0150] An autonomous landing model building unit, used to build a time-energy optimal autonomous landing model;

[0151] The autonomous tracking mechanism design unit designs an autonomous tracking mechanism based on the autonomous landing model to generate the autonomous tracking path of the UAV, including: generating local targets through rolling space and performing local replanning, taking the local targets as the target positions, obtaining the UAV trajectory search sequence through the search algorithm; based on the UAV trajectory search sequence, generating the UAV target trajectory prediction sequence

[0152] Autonomous landing mechanism design unit, used to design autonomous landing mechanism and realize UAV landing, including: designing time and dynamic constraints, considering terminal constraints in autonomous landing, optimizing autonomous landing model, and solving autonomous landing model through NLOPT;

[0153] The landing unit initializes motion planning parameters and completes the landing through an autonomous landing finite state machine including the autonomous tracking mechanism and the autonomous landing mechanism.

[0154] The present embodiment provides a method for autonomous landing path planning of a UAV based on time and energy, which is an autonomous landing path planning technology specially designed for vertical take-off and landing UAVs and meets the requirements of dynamic constraints and real-time performance.

[0155] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A time-energy based UAV autonomous landing path planning method, characterized by: Includes steps: Construct a time-energy optimal autonomous landing model; An autonomous tracking mechanism is designed based on the autonomous landing model to generate an autonomous tracking path for the drone, including: Generate local targets through rolling space and perform local replanning, use the local targets as target positions, and obtain the UAV trajectory search sequence through the search algorithm; Based on the UAV trajectory search sequence, generate the UAV target trajectory prediction sequence; Design an autonomous landing mechanism to land the drone, including: Design time and dynamic constraints, and consider the terminal constraints in autonomous landing to optimize the autonomous landing model; Solve the autonomous landing model through NLOPT; Initialize motion planning parameters, and complete landing through an autonomous landing finite state machine including the autonomous tracking mechanism and the autonomous landing mechanism.

2. The method for autonomous landing path planning of a UAV based on time-energy according to claim 1, characterized in that: The objective function of the autonomous landing model is designed as: ； in, Indicates the landing time. To adjust the parameters, Respectively The third-order derivative of Indicates that the drone is The position at the moment.

3. The method for autonomous landing path planning of a UAV based on time-energy according to claim 2, characterized in that: The constraints of the objective function include: The dynamic differential constraint is: ； ； in, express Drone status at any moment, express The first derivative of ; parameter matrix , express OK The zero matrix of columns, express dimensional identity matrix; , and Respectively exist direction, Direction and Position in direction, , and Respectively The first, second, and third derivatives of ; Indicates the jump of the drone, the value is ; The boundary constraints are: ； in, Indicates the landing start time, that is, the initial time; Indicates the state of the drone when it starts landing, that is, the initial state of the drone; Indicates the time when landing is completed. Indicates the status of the drone when it completes landing; The process constraints on velocity and acceleration are: ； ； in, Indicates that the drone is The speed of time, The maximum speed constraint of the UAV during the landing process; Indicates that the drone is The acceleration of time, Represents the maximum acceleration constraint of the drone during the landing process; Process constraints on time trajectory: the landing start time Until landing completion time According to the MINCO curve, it is divided into time trajectory segments, denoted as , the duration of each time trajectory segment must be greater than 0: .

4. The method for autonomous landing path planning of a UAV based on time-energy according to claim 3, characterized in that: Designing time and dynamic constraints to optimize the autonomous landing model includes: Regarding the process constraints of speed and acceleration, the objective function Optimized to: ； ； ； in, and are the weights of speed penalty and acceleration penalty respectively, For the optimization objective of the process constraint on acceleration, Optimizing objectives for process constraints regarding speed; Regarding the process constraints of the time trajectory, the objective function Optimized to: ； ； ； in, , is an unconstrained virtual time series, A sequence of path points representing a MINCO curve.

5. The method for autonomous landing path planning of a UAV based on time-energy according to claim 1, characterized in that: Generating local targets through rolling space specifically includes: The current position of the target is taken as the global target, and it is determined whether the global target is in the rolling space. If it is in the rolling space, the global target is taken as the target position, and the next moment position of the drone is obtained through the search algorithm, thereby forming a drone trajectory search sequence. If it is not in the rolling space, first connect the current position of the drone with the global target, take the intersection of the line and the rolling space as the local target, take the local target as the target position, and obtain the next moment position of the drone through the search algorithm, thus forming the drone trajectory search sequence ; When executing the search algorithm, the optimal control cost of the drone is obtained by solving the boundary value method .

6. The method for autonomous landing path planning of a UAV based on time and energy according to claim 5, characterized in that: Optimal control cost for: ； in, express The optimal state of the drone at all times; represents a 9-dimensional costate vector, Respectively represent dimensional covariate; , , , , , , , and Indicates that direction, Direction and The unknown constant to be solved in the direction, Represents the Hamiltonian function, which is: ； in, To adjust the parameters; , and Respectively represent displacement The third, second and first order derivatives of the UAV, i.e. the jump, acceleration and speed of the UAV, subscript , and Respectively direction, Direction and direction.

7. The method for autonomous landing path planning of a UAV based on time and energy according to claim 5, characterized in that: The search algorithm adopts the A-star algorithm.

8. The method for autonomous landing path planning of a UAV based on time and energy according to claim 7, characterized in that: The cost function of the A-star algorithm is: ； ； in, , To adjust the parameters, and Respectively The corresponding discrete control cost and expansion step size, express The corresponding discrete control cost is for The corresponding optimal discrete control cost is, Search objective function for landing time constraints.

9. The method for autonomous landing path planning of a UAV based on time-energy according to claim 1, characterized in that: Based on the UAV trajectory search sequence, a UAV target trajectory prediction sequence is generated, which includes: The second-order chain integral model is used to model the target dynamics, and the target acceleration is used as the system input, which is: ； ； ； in, express The state variable of the target at time instant, express The first derivative of ; Indicates the target The location at the moment, , and Respectively exist direction, Direction and Position in direction, and Respectively The first and second derivatives of ; Represents the target system input, and its value is ; Calculate the predicted duration through the target's dynamic model The estimated target position : ； The initial position of the target , target initial speed ; Enter the target system Using discrete values, it is expressed as: ； in, represents the target maximum acceleration constraint, Indicates the set value of the discrete coefficient; based on The state variable of the target at the moment ,choose System input of the target at the moment : ； in, Indicates the actual cost, Represents the heuristic cost: ； ； in, represents the adjustment parameter; Input the optimal system The corresponding state variable Add to target trajectory prediction sequence .

10. An autonomous landing path planning system for a UAV implementing the method described in any one of claims 1 to 9, characterized in that: include: An autonomous landing model building unit, used to build a time-energy optimal autonomous landing model; The autonomous tracking mechanism design unit designs an autonomous tracking mechanism based on the autonomous landing model to generate the autonomous tracking path of the UAV, including: generating local targets through rolling space and performing local replanning, taking the local targets as the target positions, obtaining the UAV trajectory search sequence through the search algorithm; based on the UAV trajectory search sequence, generating the UAV target trajectory prediction sequence Autonomous landing mechanism design unit, used to design autonomous landing mechanism and realize UAV landing, including: designing time and dynamic constraints, considering terminal constraints in autonomous landing, optimizing autonomous landing model, and solving autonomous landing model through NLOPT; The landing unit initializes motion planning parameters and completes the landing through an autonomous landing finite state machine including the autonomous tracking mechanism and the autonomous landing mechanism.

Images