Unmanned aerial vehicle autonomous flight method and system in dynamic unknown environment

Abstract

The invention discloses an unmanned aerial vehicle autonomous flight method and system in a dynamic unknown environment, and belongs to the technical field of rotor unmanned aerial vehicle planning control, and the method comprises the steps: carrying out the local path search, and generating a path trajectory; performing parameterization on the generated path trajectory; constructing an unmanned aerial vehicle safe flight channel according to the detection information; constructing a model prediction controller, tracking a path, solving the controller, and generating an optimal control action of the unmanned aerial vehicle; and the required angular velocity and thrust expectation are generated based on the differential flatness characteristic of the four rotors, and autonomous flight of the unmanned aerial vehicle in an unknown dynamic environment is realized. By adopting the method and the system, the autonomous flight requirement of the unmanned aerial vehicle in a complex and changeable environment is met.

Description

Technical Field

[0001] The present invention relates to the technical field of planning and controlling a rotary-wing unmanned aerial vehicle, and in particular to a method and system for autonomously flying a unmanned aerial vehicle in a dynamic unknown environment. Background Art

[0002] In recent years, quadcopter technology has developed rapidly and has been widely used in many fields such as environmental monitoring, survey operations, and multi-machine coordinated formation flight. However, these application scenarios often encounter unexpected obstacles and interference, such as small toys, birds, or sudden winds, which put severe tests on the stability and reliability of quadcopter mission execution. In unknown dynamic environments, the autonomous navigation capabilities of quadcopters still face major challenges.

[0003] Currently, autonomous flight of quadcopters generally adopts a structure in which the motion planner and controller are separated, such as Figure 1 As shown in the figure, it includes two frameworks of separation of planning and control. Under this architecture, the motion planner generates a flight trajectory that satisfies dynamic constraints and is located in a safe area based on the input target point position; then, the controller calculates the control instructions based on the deviation between the real-time state estimation feedback and the reference path to reduce the system deviation and ensure accurate tracking of the trajectory.

[0004] There are two problems with this architecture. First, the step-by-step processing flow increases the reaction time of the system. Although some methods can be used to relax the constraints and complete trajectory planning in a shorter time to reduce the delay time, this flexibility is likely to cause the trajectory optimization to fall into a local optimal solution when encountering unexpected obstacles, thereby increasing the risk of collision with obstacles. Secondly, in this separated architecture, the ability to handle interference is insufficient. Since dynamic interference factors are not considered in the planning process, the system's response is not agile enough when encountering sudden interference such as strong winds, which may pose a threat to the safety of the quadcopter. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for autonomous flight of a UAV in a dynamic unknown environment, so as to meet the autonomous flight requirements of the UAV in a complex and changeable environment.

[0006] To achieve the above object, the present invention provides a method for autonomous flight of a UAV in a dynamic unknown environment, the steps comprising:

[0007] S1, obtain UAV detection information in real time, use graph search-based algorithm to perform local path search, and generate path trajectory;

[0008] S2, parameterizing the generated path trajectory;

[0009] S3, constructing a safe flight channel for the UAV based on the detection information;

[0010] S4, construct a model predictive controller to track the parameterized path trajectory while keeping the UAV in the safe free space represented by the safe flight channel, and solve the controller to generate the optimal control action of the UAV;

[0011] S5. Based on the differential flatness characteristics of the quadrotor, the optimal control action is converted into angular velocity, and the thrust of the drone is calculated. The drone is tracked according to the angular velocity reference and thrust reference to achieve autonomous flight of the drone in an unknown dynamic environment.

[0012] Preferably, in step S1, the drone detection information is acquired in real time based on the flight control IMU and the laser radar carried by the drone.

[0013] Preferably, in step S1, a three-dimensional grid map is used to characterize the local environment, and an improved bidirectional A* algorithm is used to perform local path search to generate a path trajectory.

[0014] Preferably, the steps of improving the bidirectional A* algorithm include:

[0015] S11, initialization: put the drone start node into the open list OpenList1, put the drone end node into the open list OpenList2, initialize their evaluation function values ​​to 0 respectively, and create an empty closed list ClosedList;

[0016] S12, executing the algorithm until the termination condition is met, including:

[0017] a. If both open lists are empty, stop the loop; if not, continue the loop;

[0018] b. Select a node with the lowest cost from each of the two open lists as the current node, and add the two nodes to the ClosedList respectively;

[0019] c. If the adjacent node of the current node is the same as the search node in the opposite direction, stop the loop. If not, proceed to d.

[0020] d. Determine whether the length of the open list has changed. If not, proceed to step a. If it has changed, proceed to step e.

[0021] e. Select all child nodes, select the next node, and update the open list;

[0022] f. Modify the target node and enter a;

[0023] S13. Trace back in reverse order according to the parent node pointer to obtain the shortest path.

[0024] Preferably, in step S2, the path trajectory is generated by parameterization using a B-spline curve.

[0025] Preferably, step S3 of constructing a safe flight channel for the drone according to the detection information includes:

[0026] Construct a polyhedron, which is defined by a series of hyperplanes (C, d) to constrain the drone position p, as follows:

[0027] C·pd≤0;

[0028] Among them, C represents the normal vector matrix of the polyhedron hyperplane, which is of size K×3, d represents a K×1 vector of hyperplane constants, and K represents the number of hyperplanes of the polyhedron.

[0029] Preferably, step S4 constructs a model predictive controller to track the parameterized path trajectory while keeping the drone in a safe free space represented by a safe flight channel, including:

[0030] Sample N points to represent the prediction range and reference position point p of the model predictive controller ref,n (n=1,2,…,N) is located on the path trajectory. If p is outside the safe flight channel after calculation, it is adjusted to the farthest waypoint in the safe flight channel;

[0031] The model predictive controller expression is:

[0032]

[0033] Among them, the objective function By error reference Controls Control Variation and the kinetic feasible term R j F(c i ), R represents the weight coefficient, p n Represents the reference position point, R u Represents the weight coefficient of the control quantity, R c represents the weight coefficient of the smoothing term, Represents the dynamic constraint, which imposes dynamic constraints on the velocity V, acceleration A and jerk J in the three-axis direction, where a 1 ,b 1 ,c 1 ,a 2 ,b 2 ,c 2 Satisfies second-order continuity, c r represents the quadratic parameter, c j represents a piecewise function node, λ represents a one-time weight, c mrepresents a fixed coefficient, Δt represents the time interval, v represents the velocity, a represents the acceleration, j represents the jerk, i represents the reference point sequence, P(t) represents the flight trajectory, q represents the control point, u represents the control amount, t represents the time, and M represents the parameter matrix.

[0034] Preferably, in step S4, the model predictive controller is solved based on OSQP and Eigen3 to generate the optimal control action u.

[0035] Preferably, step S5 specifically includes:

[0036] S51. Based on the differential flatness characteristic of the quadrotor, the optimal control action u is converted into an angular velocity reference, which is expressed as:

[0037]

[0038] x C =[cosφ,sinφ,0] T ;

[0039]

[0040] p r =-h w ·y B ,q r =h w ·x B ;

[0041] r r =(φ r -φ)·z B (0,0,1) T ;

[0042] Among them, g represents the constant value of gravity acceleration, φ r and φ represent the expected yaw angle and the actually measured yaw angle of the quadrotor drone in the world coordinate system, respectively, and (p r ,q r ,r r ) is the desired pitch velocity, roll velocity and yaw velocity of the drone in its body coordinate system, acceleration j represents the optimal control action u in the world coordinate system, Z B Indicates the z-axis direction of the machine system, X C Indicates the intermediate system, y B Indicates the y-axis of the machine system, h w represents the intermediate system parameters;

[0043] S52. Calculate the thrust of the drone using the formula:

[0044] T r =C T ·‖t‖;

[0045] Among them, T r Indicates thrust, C T Indicates the throttle thrust coefficient;

[0046] S53, the UAV is tracked according to the angular velocity reference and the thrust reference, so as to realize the autonomous flight of the UAV in an unknown dynamic environment.

[0047] The present invention also provides an autonomous flight system for a UAV in a dynamic unknown environment, including a front-end module and a back-end module. The front-end module is responsible for executing local path search and parameterized processing of trajectories, and the back-end module performs control optimization in a fixed periodic cycle, including: constructing a safe flight channel for the UAV; constructing a model predictive controller, planning and controlling the model predictive controller to obtain an optimal control action; converting the obtained optimal control action into an expected input based on a differential flat characteristic, calculating the thrust of the UAV, and allowing the UAV to track according to an angular velocity reference and a thrust reference, thereby realizing autonomous flight of the UAV in an unknown dynamic environment.

[0048] Therefore, the present invention adopts the above-mentioned method and system for autonomous flight of unmanned aerial vehicles in a dynamic unknown environment, which has the following beneficial effects:

[0049] (1) In this application, path planning and control are integrated into one framework, which saves the system's reaction time;

[0050] (2) The improved double A* algorithm is used to calculate the path from the current point to the minimum point in the openlist table on the other side, ensuring that the paths in the forward and reverse directions will eventually intersect, preventing the situation in some application scenarios where the forward and reverse searches cannot meet at the middle point due to the large number of obstacles and complex shapes;

[0051] (3) By constructing a series of overlapping polyhedral regions and placing the polyhedral regions as hard constraints in the optimization process, the UAV is prevented from colliding during flight and its ability to avoid dynamic obstacles is improved.

[0052] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a framework diagram of an existing UAV system in the background technology;

[0054] Figure 2 A framework diagram of a drone system according to an embodiment of the present invention;

[0055] Figure 3 This is a flow chart of an improved bidirectional A* algorithm according to an embodiment of the present invention;

[0056] Figure 4 Schematic diagram of a B-spline curve according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0058] Example

[0059] Reference Figure 2 The present invention provides an autonomous flight system for unmanned aerial vehicles in a dynamic unknown environment, including a front-end module and a back-end module. The front-end module is responsible for executing local path search and parameterization of trajectory. The system obtains the surrounding point cloud information and the real-time attitude information of the unmanned aerial vehicle in real time according to the flight control IMU and the laser radar carried by the unmanned aerial vehicle, and performs path planning and trajectory parameterization based on the detection information of the unmanned aerial vehicle. The back-end module uses the local map information and the conflict-free reference path after the front-end parameterization to execute at a fixed frequency of 100 times per second to track the reference path, including: constructing a safe flight channel (SFC) for unmanned aerial vehicles; constructing a model predictive controller (MPC), planning and controlling the MPC, and obtaining the optimal control action; converting the obtained optimal control action into the expected input based on the differential flat characteristic, calculating the thrust of the unmanned aerial vehicle, and making the unmanned aerial vehicle track according to the angular velocity reference and the thrust reference, so as to realize the autonomous flight of the unmanned aerial vehicle in an unknown dynamic environment. When constructing the SFC, a series of overlapping polyhedral regions are created based on the parameterized path trajectory. The system puts the SFC into the optimization process as a hard constraint to prevent the unmanned aerial vehicle from colliding during flight. At the same time, the MPC will generate the required reference state sequence in these SFCs. By calculating the objective function and combining the differential flatness property, the optimized MPC variables are converted into specific angular velocity instructions, so that the quadrotor drone can directly control the propeller speed through the underlying angular velocity controller.

[0060] The present invention also provides a method for autonomous flight of a UAV in a dynamic unknown environment, the steps comprising:

[0061] S1. Obtain drone detection information in real time, use a graph search-based algorithm to perform local path search, and generate path trajectories.

[0062] In this embodiment, the drone detection information is acquired in real time based on the flight control IMU and the laser radar carried by the drone. Since the visual range of the laser radar is limited, the quadcopter drone can only detect environmental data in a limited area, so its path planning process must be adjusted accordingly according to the real-time changes in the surrounding environment. The drone will update the surrounding environmental information at a fixed frequency, and must also perform a new path search after each update to move toward the predetermined local target. Using a three-dimensional grid map to represent the local environment, the motion space segmentation will be divided into uniform grid units, which can simplify calculations, reduce the pressure on the onboard computer, and improve the efficiency of grid-based operations and data processing. An improved two-way A* algorithm is used to perform local path search and generate a path trajectory. Figure 3 As shown, the steps of improving the bidirectional A* algorithm include:

[0063] S11, Initialization: put the starting node into an "open list" (OpenList1), put the end node into another "open list" (OpenList2), and initialize their evaluation function values ​​to 0. At the same time, create an empty "closed list" (ClosedList);

[0064] S12, executing the algorithm until the termination condition is met, including:

[0065] a. If both open lists are empty, stop the loop; if not, continue the loop;

[0066] b. Select a node with the lowest cost from each of the two open lists as the current node, and add the two nodes to the ClosedList respectively;

[0067] c. If the adjacent node of the current node is the same as the search node in the opposite direction, stop the loop. If not, proceed to d.

[0068] d. Determine whether the length of the open list has changed. If not, proceed to step a. If it has changed, proceed to step e.

[0069] e. Select all child nodes, select the next node, and update the open list;

[0070] f. Modify the target node and enter a;

[0071] S13. Trace back in reverse order according to the parent node pointer to obtain the shortest path.

[0072] The calculation method of the improved bidirectional A* algorithm is not to calculate the path from the current point to the end point, but the path from the current point to the minimum point in the openlist table on the other side, so as to ensure that the paths in the forward and reverse directions will eventually intersect.

[0073] S2. Parameterize the generated path trajectory.

[0074] In this embodiment, the path trajectory is generated by parameterization of B-spline curve. The path point set generated by the improved double A* algorithm is discrete. When the UAV flies along this path, it will experience a sudden change in speed and acceleration at each turning point of the path, which will undoubtedly reduce the overall efficiency of the flight. By parameterizing the trajectory, the detailed state of the object during the movement can be described in detail, including key parameters such as position, speed and acceleration.

[0075] A B-spline curve is a curve that is approximated by a set of control points, but it does not pass through these control points exactly. Figure 4 shown.

[0076] A B-spline curve contains n+1 control points {q 0 ,q 1 ,q 2 ,…,q n} and m+1 nodes {t 0 ,t 1 ,...,t m} and the degree p of the curve, there is a clear equation relationship between the three m = n + p + 1. If the nodes of the B-spline curve are equidistant, that is, m + 1 nodes satisfy This type of curve is called a uniform B-spline curve. If the nodes of the B-spline curve are not all equidistant, then such a curve is generally called a non-uniform B-spline curve.

[0077] For a spline curve, its domain is [t p ,t m-p ], any interval in the domain In the example, each corresponds to a local trajectory, so there are J = n + 1-p = m-2p segments. For uniform B-spline curves, all basis matrices are the same constant matrix, as shown below:

[0078]

[0079] Where u represents the control quantity, t represents the time parameter, M represents the parameter matrix, C(t) represents the flight trajectory, and i represents the reference point sequence.

[0080] B-spline curves have derivative properties. The derivative of a p-order B-spline curve is a p-1-order B-spline curve defined by the original node vector and the new n control points. Using this recursive property, the derivatives of the curve can be calculated, especially the calculation formulas for the first and second order derivatives are as follows:

[0081]

[0082] For uniform B-spline curves, due to The control points of the B-spline curve derivative can be simplified to:

[0083]

[0084] S3. Construct a safe flight channel for the UAV based on the detection information.

[0085] In this embodiment, constructing a safe flight channel for the drone based on the detection information includes:

[0086] Construct a polyhedron, which is defined by a series of hyperplanes (C, d) to constrain the drone position p, as follows:

[0087] C·pd≤0;

[0088] Among them, C represents the normal vector matrix of the polyhedron hyperplane, which is of size K×3, d represents a K×1 vector of hyperplane constants, and K represents the number of hyperplanes of the polyhedron.

[0089] S4. Construct a model predictive controller to track the parameterized path trajectory while keeping the UAV in the safe free space represented by the safe flight channel, and solve the controller to generate the optimal control action of the UAV.

[0090] In this embodiment, constructing a model predictive controller to track the parameterized B-like curve while keeping the drone in a safe free space represented by a safe flight channel includes:

[0091] Sample N points to represent the prediction range and reference position point p of the model predictive controller ref,n (n=1,2,…,N) is located on the path trajectory. If p is outside the SFC after calculation, it is adjusted to the farthest waypoint within the SFC;

[0092] The model predictive controller expression is:

[0093]

[0094] Among them, the objective function By error reference Controls Control Variation and the kinetic feasible term Rj F(c i ), R represents the weight coefficient, p n Represents the reference position point, R u Represents the weight coefficient of the control quantity, R c represents the weight coefficient of the smoothing term, Represents the dynamic constraint, which imposes dynamic constraints on the velocity V, acceleration A and jerk J in the three-axis direction, where a 1 ,b 1 ,c 1 ,a 2 ,b 2 ,c 2 Satisfies second-order continuity, c r represents the quadratic term parameter, c j represents a piecewise function node, λ represents a one-time weight, c m represents a fixed coefficient, Δt represents the time interval, v represents the velocity, a represents the acceleration, j represents the jerk, i represents the reference point sequence, P(t) represents the flight trajectory, q represents the control point, u represents the control amount, t represents the time, and M represents the parameter matrix.

[0095] When solving, it involves solving the variable u=[u 0 ,u 1 ,...,u N-1 ] T , based on OSQP and Eigen3 C++ library. When the objective function generates the optimal solution in the kth step, a series of optimal control actions u=[u 0 ,u 1 ,...,u N-1 ] T is recorded, at which point the model predictive controller will 0 As the control input of the current step, when the optimization problem cannot be solved in the kth step, the time information between the last successful step and the failed step k is obtained, and the kth control action u of the last successful solution is output. Based on this control strategy, MPC is more robust in the face of failures and reduces the occurrence of accidents.

[0096] S5. Based on the differential flatness characteristics of the quadrotor, the optimal control action is converted into angular velocity, and the thrust of the drone is calculated. The drone is tracked according to the angular velocity reference and thrust reference to achieve autonomous flight of the drone in an unknown dynamic environment.

[0097] After solving the MPC problem, the optimal control action u, i.e., jerk j, defined in the world coordinate system cannot be directly applied to the quadrotor in the real world because j is not a command that the motor can execute. Therefore, the system needs to convert the jerk j and acceleration into an angular velocity reference based on the differential flatness characteristics of the quadrotor. Finally, the angular velocity reference and thrust reference are tracked by the low-level controller implemented on the autopilot to generate commands that the motor can execute.

[0098] In this embodiment, the optimal control action is converted into angular velocity based on the differential flatness characteristics of the quadrotor, the thrust of the drone is calculated, and the drone is tracked according to the angular velocity reference and the thrust reference to achieve autonomous flight of the drone in an unknown dynamic environment, specifically including:

[0099] S51. Based on the differential flatness characteristic of the quadrotor, the optimal control action u is converted into an angular velocity reference, which is expressed as:

[0100]

[0101] x C =[cosφ,sinφ,0] T ;

[0102]

[0103] p r =-h w ·y B ,q r =h w ·x B ;

[0104] r r =(φ r -φ)·z B (0,0,1) T ;

[0105] Among them, g represents the constant value of gravity acceleration, φ r and φ represent the expected yaw angle and the actually measured yaw angle of the quadrotor drone in the world coordinate system, respectively, and (p r ,q r ,r r ) is the desired pitch velocity, roll velocity and yaw velocity of the drone in its body coordinate system, acceleration j represents the optimal control action u in the world coordinate system, Z B Indicates the z-axis direction of the machine system, X C Indicates the intermediate system, y B Indicates the y-axis of the machine system, h w represents the intermediate system parameters;

[0106] S52. Calculate the thrust of the drone to control the movement of the quadrotor drone along the Z axis. The formula is:

[0107] T r =C T ·‖t‖;

[0108] Among them, T r Indicates thrust, C T Indicates the throttle thrust coefficient;

[0109] S53, the UAV is tracked according to the angular velocity reference and the thrust reference, so as to realize the autonomous flight of the UAV in an unknown dynamic environment.

[0110] Therefore, the present invention adopts the above-mentioned method and system for autonomous flight of a drone in a dynamic unknown environment, which saves the system's response time by integrating path planning and control into one framework, and meets the autonomous flight requirements of the drone in a complex and changeable environment.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for autonomous flight of a drone in a dynamic unknown environment, characterized in that the steps include: S1, obtain UAV detection information in real time, use graph search-based algorithm to perform local path search, and generate path trajectory; S2, parameterizing the generated path trajectory; S3, constructing a safe flight channel for the UAV based on the detection information; S4, construct a model predictive controller to track the parameterized path trajectory while keeping the UAV in the safe free space represented by the safe flight channel, and solve the controller to generate the optimal control action of the UAV; S5. Based on the differential flatness characteristics of the quadrotor, the optimal control action is converted into angular velocity, and the thrust of the drone is calculated. The drone is tracked according to the angular velocity reference and thrust reference to achieve autonomous flight of the drone in an unknown dynamic environment.

2. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 1, characterized in that: In step S1, the drone detection information is obtained in real time based on the flight control IMU and the laser radar carried by the drone.

3. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 1, characterized in that: In step S1, a three-dimensional grid map is used to represent the local environment, and an improved bidirectional A* algorithm is used to perform local path search to generate a path trajectory.

4. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 3, characterized in that: The steps to improve the bidirectional A* algorithm include: S11, initialization: put the drone start node into the open list OpenList1, put the drone end node into the open list OpenList2, initialize their evaluation function values ​​to 0 respectively, and create an empty closed list ClosedList; S12, executing the algorithm until the termination condition is met, including: a. If both open lists are empty, stop the loop; if not, continue the loop; b. Select a node with the lowest cost from each of the two open lists as the current node, and add the two nodes to the ClosedList respectively; c. If the adjacent node of the current node is the same as the search node in the opposite direction, stop the loop. If not, proceed to d. d. Determine whether the length of the open list has changed. If not, proceed to step a. If it has changed, proceed to step e. e. Select all child nodes, select the next node, and update the open list; f. Modify the target node and enter a; S13. Trace back in reverse order according to the parent node pointer to obtain the shortest path.

5. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 4, characterized in that: In step S2, the path trajectory is generated by parameterization using a B-spline curve.

6. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 5, characterized in that: Step S3 constructs a safe flight channel for the UAV based on the detection information, including: Construct a polyhedron, which is defined by a series of hyperplanes (C, d) to constrain the drone position p, as follows: C·pd≤0; Among them, C represents the normal vector matrix of the polyhedron hyperplane, which is of size K×3, d represents a K×1 vector of hyperplane constants, and K represents the number of hyperplanes of the polyhedron.

7. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 6, characterized in that: Step S4 constructs a model predictive controller to track the parameterized path trajectory while keeping the UAV in a safe free space represented by a safe flight channel, including: Sample N points to represent the prediction range and reference position point p of the model predictive controller ref,n (n=1,2,…,N) is located on the path trajectory. If p is outside the safe flight channel after calculation, it is adjusted to the farthest waypoint in the safe flight channel; The model predictive controller expression is: s.t.u n =j n ,n=1,2,...,N; C·pd≤0; Among them, the objective function By error reference Controls Control Variation and the kinetic feasible term R j F(c i ), R represents the weight coefficient, p n Represents the reference position point, R u Represents the weight coefficient of the control quantity, R c represents the weight coefficient of the smoothing term, Represents the dynamic constraint, which imposes dynamic constraints on the velocity V, acceleration A and jerk J in the three-axis direction, where a1, b1, c1, a2, b2, c2 satisfy the second-order continuity, c r represents the quadratic term parameter, c j represents a piecewise function node, λ represents a one-time weight, c m represents a fixed coefficient, Δt represents the time interval, v represents the velocity, a represents the acceleration, j represents the jerk, i represents the reference point sequence, P(t) represents the flight trajectory, q represents the control point, u represents the control amount, t represents the time, and M represents the parameter matrix.

8. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 7, characterized in that: In step S4, the model predictive controller is solved based on OSQP and Eigen3 to generate the optimal control action u.

9. The method for autonomous flight of a drone in a dynamic unknown environment according to claim 8, characterized in that: Step S5 specifically includes: S51. Based on the differential flatness characteristic of the quadrotor, the optimal control action u is converted into an angular velocity reference, which is expressed as: x C =[cosφ,sinφ,0] T ; p r =-h w ·y B ,q r =h w ·x B ; r r =(φ r -φ)·z B ·(0,0,1) T ; Among them, g represents the constant value of gravity acceleration, φ r and φ represent the expected yaw angle and the actually measured yaw angle of the quadrotor drone in the world coordinate system, respectively, and (p r ,q r ,r r ) is the desired pitch velocity, roll velocity and yaw velocity of the drone in its body coordinate system, acceleration j represents the optimal control action u in the world coordinate system, Z B Indicates the z-axis direction of the machine system, X C Indicates the intermediate system, y B Indicates the y-axis of the machine system, h w represents the intermediate system parameters; S52. Calculate the thrust of the drone using the formula: T r =C T ·‖t‖; Among them, T r Indicates thrust, C T Indicates the throttle thrust coefficient; S53, the UAV is tracked according to the angular velocity reference and the thrust reference, so as to realize the autonomous flight of the UAV in an unknown dynamic environment.

10. A system for autonomous flight of a drone in a dynamic unknown environment according to any one of claims 1 to 9, characterized in that: It includes a front-end module and a back-end module. The front-end module is responsible for executing local path search and trajectory parameterization processing, and the back-end module operates in a fixed periodic cycle to perform control optimization, including: building a safe flight channel for drones; building a model predictive controller, planning and controlling the model predictive controller to obtain the optimal control action; converting the obtained optimal control action into the expected input based on the differential flatness characteristic, calculating the drone thrust, and allowing the drone to track according to the angular velocity reference and thrust reference, thereby realizing autonomous flight of the drone in an unknown dynamic environment.

Images