Unmanned aerial vehicle flight system used under condition of dynamic environment change

Abstract

The invention discloses an unmanned aerial vehicle flight system used under the condition of dynamic environment change, and the system mainly comprises five parts: information release, improvement of an artificial potential field trajectory re-planning method, an unmanned aerial vehicle control module, an unmanned aerial vehicle flight module, and unmanned aerial vehicle / ground station feedback, and is mainly used for performing trajectory re-planning on an unmanned aerial vehicle autonomous flight task. The method has the advantages that unmanned aerial vehicle flight path oscillation and wandering caused by local minimum value searching of the repulsion field edge in path planning of the unmanned aerial vehicle are weakened, the problem of reducing flight energy efficiency is solved, the complexity of a process regression algorithm is reduced, and the online path planning efficiency of an artificial potential field method is improved and optimized.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle systems, and in particular to an unmanned aerial vehicle flight system used in a dynamically changing environment. Background Art

[0002] The UAV flight system mainly includes four parts: UAV flight trajectory planning method, UAV control module, UAV flight module, and UAV / ground station feedback. UAV flight trajectory planning is the core technology of UAV flight system. Trajectory planning in complex and dynamic maps has become a hot topic in the current UAV field. In order to solve the trajectory planning problem of UAV, domestic and foreign scholars have proposed a large number of trajectory planning algorithms, among which local planning algorithms have a wider range of applications. Traditional PRM and RRT algorithms often need to model obstacles in a certain space, which is complex to calculate and therefore not suitable for planning UAVs in complex environments; Dijkstra algorithm is simple to implement, but the search time in a large range of high-dimensional space is long and the memory requirement is high, so it is mostly used for two-dimensional static trajectory planning; artificial potential field method has the advantages of strong real-time performance and low computational complexity, and is suitable for trajectory planning tasks of UAVs under certain conditions. However, when planning trajectories, the artificial potential field method is prone to wandering around the minimum value when searching for the local minimum value, and the repulsive force direction changes too much, resulting in flight jitter, which leads to the unreachable flight target of the UAV and low control efficiency. These problems need to be solved urgently.

[0003] In summary, the flight efficiency and local planning issues of drones pose many challenges to the local planning method of drone trajectories. On the one hand, after completing trajectory planning, when the target point changes, the obstacle status information in the environment changes, and the execution trajectory deviates from the planned trajectory during the flight of the drone according to the pre-trajectory, it is necessary to quickly re-plan a feasible trajectory from the current state to the target state, which poses a real-time problem for the drone's own trajectory re-planning method. On the other hand, the trajectory planning method should reduce or solve the local minimum oscillation and wandering problems that occur in trajectory planning. Therefore, it is very necessary to study a real-time and efficient drone trajectory planning method for drone flight systems and drone autonomous flight missions. Summary of the invention

[0004] In order to solve the above-mentioned problems, the present invention discloses a UAV flight system for use in a dynamically changing environment.

[0005] A UAV flight system for use in a dynamically changing environment includes five parts: information release, improved artificial potential field trajectory replanning method, UAV control module, UAV flight module, and UAV / ground station feedback. The improved artificial potential field trajectory replanning method is used to obtain target information, UAV state information, and environmental state information, to construct an improved artificial potential field, predict the state trajectory, replan the improved artificial potential field, calculate the expectation and pass it to the UAV flight control module, and calculate the UAV flight control quantity in combination with the feedback, so as to achieve the UAV flight task under the dynamically changing environment.

[0006] The improved artificial potential field trajectory replanning method includes the following steps:

[0007] Step 1: Information acquisition, obtain the target point and status information. The target point information is the three-dimensional position of the target point in the world coordinate system. The status information includes the three-dimensional position coordinate data of the UAV in the world coordinate system, the speed data of the UAV in the world coordinate system, the attitude angle data in the body coordinate system, the angular velocity data of the UAV in the body coordinate system, and the environmental information status data of the UAV during flight, so as to prepare data for the subsequent steps;

[0008] The quadcopter drone model is an X-type drone, defined as x , y , z Indicates the three-dimensional position coordinates of the drone in the world coordinate system, X and Z are the coordinates of the drone and the target point respectively. v x , v y , v z Indicates the speed of the drone in the world coordinate system. V =[ v x v y v z ] T ; f , i , ψ They represent the attitude angles in the body coordinate system, f is the roll angle, i is the pitch angle, ψ is the yaw angle, p , q , r They are the angular velocity of the UAV in the body coordinate system;

[0009] Rotation matrix from body coordinate system to world coordinate system R for:

[0010] The environmental information status data during the UAV flight is defined as the obstacle status in the world coordinate system, as well as the historical information of the above data.

[0011] Step 2: Establish a potential field function, update the expected target point according to the data information in step 1, and construct an improved artificial potential field method to plan the virtual potential field function and virtual force function;

[0012] Defined in t At the moment, the position coordinates of the quadrotor drone are X t , the target point coordinates are Z t , Z t According to X t The Euclidean distance is less than a certain value, so that the drone is always affected by the gravity of the current target point before it reaches the target point. X t and Z t The gravitational force between them decreases as the distance between them decreases. X t and Z t If they coincide, the gravity is 0, and the gravitational field function is: U att ( t )= k att d 2 ( X t , Z t ) in, d ( X t , Z t )=|| Z t - X t || is the Euclidean distance between the drone and the target, k att is the known gravitational field constant, the attraction of the drone in the gravitational field F att ( t )and U att ( t )The relationship is: F att( t )= dU att ( t ) / dt in, F att ( t ) is the planned target point of the UAV Z t The traction force, F att ( t ) direction is X t and Z t Connecting lines between, direction pointing Z t .

[0013] Defining obstacle status O t and V ot , O t is t The position information of the obstacle in the earth coordinate system at all times, V ot yes t The speed status information of the obstacle in the earth coordinate system at all times;

[0014] The improved repulsive field function expression is: , in, d ( X t , O t )=|| O t - X t || is the Euclidean distance between the drone and the obstacle, k r1 is the constant of the position component of the repulsive field, k r2 is the velocity component constant of the repulsive field, d o is the radius of the obstacle repulsion field, V r for t The relative difference in speed between the UAV and the obstacle in the dynamic information at the moment is expressed as: V r =|| V t / ( Vt - V ot )|| V ot

[0015] The repulsive force on the drone in the repulsive field F r ( t )and U r ( t ) The relational expression is: F r ( t )= -dU r ( t ) / dt in, F r ( t ) is the planned target point of the UAV O t The repulsive force, F r ( t ) direction is X t and O t Connecting lines between, direction pointing X t ;

[0016] In the process of UAV flight trajectory planning, when the UAV enters the repulsive field, U r ( t ) produces mutation, repulsion F r ( t ) will cause the UAV flight trajectory to oscillate; in order to reduce flight oscillation and smooth the flight trajectory, the tangent function expression is defined as: , in, k q is the tangent function constant, d q is the radius of the virtual tangential force function range, and the center is O t . F q ( t )and U q ( t )The relationship is: F q ( t)= dU q ( t ) / dt in, F q ( t ) is the planned tangential force, the vertical line between the drone’s own position and the obstacle, and the direction is Z t .

[0017] Therefore, in the process of UAV flight trajectory planning, the gravitational field, repulsive field, and tangential function work together, and the synthetic potential field function and the resultant force expression are: U ( t )= U att ( t )+ U r ( t )+ U q ( t ) F ( t )= F att ( t )+ F r ( t )+ F q ( t )

[0018] Step 3: Gaussian process regression is used to predict the trajectory. A sparse Gaussian distribution function is constructed based on the local minimum value of the potential field function and the limited set of historical data obtained in step 1. The trajectory of the drone and obstacle status information are predicted through Gaussian process regression.

[0019] For the situation when the drone trajectory planning in step 2 converges to the local minimum point, the corresponding state of the drone motion trajectory state at any finite time set obeys the joint Gaussian distribution; the mean vector of the Gaussian distribution and the covariance matrix of the Gaussian distribution are set, and the trajectory state Gaussian distribution is formulated. The sparsity of the Gaussian distribution is used to expand it into the prior state distribution of the first trajectory state quantity, the prior state distribution of the last trajectory state quantity, and the prior state distribution of the adjacent trajectory state quantities. The posterior function is obtained according to the prior function and Bayesian reasoning and is expressed as: , in,{ s} is a flight event collection, p ( x ) is the prior probability; the likelihood function is designed as follows: , Among them, condition 1 is the satisfied condition for the motion event, and condition 2 is the unsatisfied condition for the motion event; according to the UAV environment perception system to obtain the flight status { x t}, using Gaussian process regression algorithm to predict the flight trajectory state of the drone at the next moment { x gp}, and according to the obstacle state information, the obstacle state information is predicted to be { o gp}.

[0020] Step 4: Replan the potential field function, predict the secondary target point according to the trajectory predicted in step 3, interpolate the secondary target point between the expected target point and the UAV, and replan the virtual potential field function and virtual force function;

[0021] On the basis of the completed trajectory planning, when the drone is flying according to the pre-planned trajectory, the target point changes, the position of obstacles in the environment changes, and the execution trajectory deviates from the planned trajectory. At this time, it is necessary to quickly re-plan a feasible trajectory from the current state to the target state online.

[0022] Step 4.1: Obtain the secondary target point information, predict the next moment’s flight trajectory of the drone according to the Gaussian process regression algorithm in step 3, calculate interpolation, obtain the secondary target point information, and obtain obstacle prediction information at the same time to prepare data for step 4.2.

[0023] Gaussian process regression algorithm predicts the next moment flight trajectory of the drone x gp}, { x gp}The location information is X gp , { x gp The speed prediction information is V gp , in the prediction trajectory and Z t Select the secondary target point of trajectory planning between coordinate points Z * t ; The predicted obstacle state information is { o gp}, { o gp}The location information is O gp , { o gp}The speed information is V ogp .

[0024] Step 4.2: Replan the virtual force function. Plan the secondary target gravitational field function according to the secondary target point information obtained in step 4.1. Plan the improved repulsive field function and tangential function based on the obstacle prediction information. Combine the gravitational field function in step 2 with the secondary target gravitational field function, and calculate the resultant force function together with the improved repulsive field function and tangential function.

[0025] According to the target point information, the target gravitational field function is constructed as follows: , in, p ( s | x t ) is the secondary target point probability model, for The gravitational constant;

[0026] The secondary target gravity function is:

[0027] The combined force of target gravity and sub-target gravity is expressed as:

[0028] According to the obstacle status information, the improved repulsion field is replanned:

[0029] The velocity compensation function with the predicted repulsive field is expressed as:

[0030] The improved repulsive field of replanning is expressed as:

[0031] Replan the repulsive field:

[0032] Replan the tangent function:

[0033] Replan the tangential force function:

[0034] The final replanned resultant force function is expressed as:

[0035] Step 5: Calculate the expectation, based on the virtual force function re-planned in step 4 F * ( t ),according to F* ( t ) and calculate the expected velocity and expected acceleration of the state in the UAV control system.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention improves the artificial potential field trajectory planning method for UAVs, uses Gaussian process regression to predict the probability of the UAV's flight trajectory at the next moment, interpolates the planned trajectory in real time through probability distribution, re-plans the UAV trajectory, reduces the jitter problem caused by searching for local minima during the flight of the UAV, and optimizes the smoothness of the flight trajectory in the process of planning the UAV flight mission using the artificial potential field method. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the method of the present invention.

[0038] Figure 2 It is a schematic diagram of Gaussian process regression trajectory prediction according to an embodiment of the present invention.

[0039] Figure 3 It is a schematic diagram of replanning virtual force according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] 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 only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of 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.

[0041] See attached figure, Figure 1 This is a flow chart of the method of the present invention. When the UAV performs a flight mission, the target point is input into the method, the expected trajectory of the UAV control system is calculated, and input into the UAV control subsystem, the UAV power control input is obtained, and given to the UAV flight system, and then the perception system obtains the UAV / ground station feedback information and inputs the information into the method.

[0042] This method mainly includes 5 steps, as follows:

[0043] Step 1: Information acquisition, obtain the target point and status information. The target point information is the three-dimensional position of the target point in the world coordinate system. The status information includes the three-dimensional position coordinate data of the UAV in the world coordinate system, the speed data of the UAV in the world coordinate system, the attitude angle data in the body coordinate system, the angular velocity data of the UAV in the body coordinate system, and the environmental information status data of the UAV during flight, so as to prepare data for the subsequent steps;

[0044] The dynamic model of the X-type quadrotor drone is established as follows:

[0045] definition x , y , z Indicates the three-dimensional position coordinate value of the drone in the world coordinate system. X =[ x.y.z ] T , v x , v y , v z Indicates the speed of the drone in the world coordinate system. V =[ v x v y v z ] T ; f , i , ψ Respectively represent the attitude angles in the body coordinate system: f is the roll angle, i is the pitch angle, ψ is the yaw angle, p , q , r They are the angular velocity of the UAV in the body coordinate system; I x , I y , I z is the moment of inertia of the body coordinate system, m is the weight of the drone itself, which is a known quantity; u 1 is the total lift, u 2. u 3. u 4 represents the torque generated by the rotor; the rotation matrix from the body coordinate system to the world coordinate system R for:

[0046] The planned target point information in the UAV flight mission is the three-dimensional position of the target point in the world coordinate system. The environmental dynamic data and the UAV attitude and position perception data are obtained during the UAV flight. The state information of the environment collected by the environmental perception module is calibrated as the obstacle, that is, the position information and speed information of the obstacle in the earth coordinate system are used to form an obstacle state quantity based on the time series; the UAV flight state information collected by the UAV position and attitude perception module is used as the state sampling point of the UAV historical flight trajectory to form a UAV motion trajectory state quantity based on the time series, which is used as a set to describe the UAV motion trajectory state.

[0047] Step 2: Establish a potential field function, update the expected target point according to the data information in step 1, and construct an improved artificial potential field method to plan the virtual potential field function and virtual force function;

[0048] In this embodiment, when the altitude of the UAV flight mission trajectory is relatively fixed, only the UAV in the earth coordinate system is considered. x and y Direction problem; the position coordinates of the quadcopter are X t , the target point coordinates are Z t , Z t along with X t The update threshold is d m , is a known quantity; the drone did not reach Z t Before, it was always affected by gravity, and the gravitational field function was: U att ( t )= k att d 2 ( X t , Z t ) / 2 in, d ( X t , Z t )=|| Z t - X t || is the Euclidean distance between the drone and the target, k att is the known gravitational field constant, the attraction of the drone in the gravitational field Fatt ( t )and U att ( t )The relationship is: F att ( t )= dU att ( t ) / dt=k att d ( X t , Z t ) in, F att ( t ) is the planned target point of the UAV Z t The traction force, F att ( t ) direction is X t and Z t Connecting lines between, direction pointing Z t .

[0049] Obstacle Status O t yes t The position information of the obstacle in the earth coordinate system at all times, O t =[ x o y o z o ] T , V ot yes t The velocity state information of the obstacle in the earth coordinate system at the moment, the improved repulsive field function is: , in, d ( X t , O t )=|| O t - X t || is the Euclidean distance between the drone and the obstacle, k r1 is the constant of the position component of the repulsive field,k r2 is the velocity component constant of the repulsive field, d o is the radius of the obstacle repulsion field, V r for t The relative difference in speed between the UAV and the obstacle in the dynamic information at the moment is expressed as: V r =|| V t / ( V t - V ot )|| V ot

[0050] The repulsive force on the drone in the repulsive field F r ( t )and U r ( t )The relationship is: F r ( t )= -dU r ( t ) / dt in, F r ( t ) is the planned target point of the UAV O t The repulsive force, F r ( t ) direction is X t and O t Connecting lines between, direction pointing X t .

[0051] In the process of UAV flight trajectory planning, when the UAV enters the repulsive field, U r ( t ) produces mutation, repulsion F r ( t ) will cause the UAV flight trajectory to oscillate; in order to reduce flight oscillation and smooth the flight trajectory, the tangent function expression is defined as: , in, kq is the tangent function constant, n is a finite constant, d q is the radius of the virtual tangential force function range, and the center is O t , in this embodiment, it is defined d q > d o ; F q ( t )and U q ( t )The relationship is: F q ( t )= dU q ( t ) / dt in, F q ( t ) is the planned tangential force, the vertical line between the drone’s own position and the obstacle, and the direction is Z t .

[0052] In the process of UAV flight trajectory planning, the gravitational field, repulsive field, and tangential function work together to synthesize the potential field function and the resultant force expression is: U ( t )= U att ( t )+ U r ( t )+ U q ( t ) F ( t )= F att ( t )+ F r ( t )+ F q ( t )

[0053] Step 3: Gaussian process regression predicts trajectory. According to the local minimum value of the potential field function and the limited set of historical data obtained in step 1, a sparse Gaussian distribution function is constructed, and the trajectory and obstacle status information of the drone are predicted by Gaussian process regression. This embodiment uses the sparsity of Gaussian distribution to establish a probabilistic trajectory model and introduces a trajectory based on maximizing the posterior probability, which is expressed as: x gp ( t )=argmax x(t) p ( x | s ) in, s For flight events, p ( x | s ) is the posterior function in the Bayesian formula, and the posterior function expression is: , in, p ( x ) is the prior probability, p ( s ) is a constant, p ( s | x ) is the likelihood function; then the trajectory model is transformed into the design of the prior function and the likelihood function.

[0054] The collected UAV flight status information is used as the state sampling point of the UAV's past flight trajectory. Then the Gaussian process can be described as a set description consisting of a finite number of supporting states that obey the joint Gaussian distribution. For any finite time set t ={ t 0,…, t N}, its corresponding state x Obey the following joint Gaussian distribution: x =[ x 0… x N ] T ~ N ( m , k ) in, , D is the dimension of the state variable; m is the mean vector of the Gaussian distribution, k is the covariance matrix of the Gaussian distribution, m and k The expression is: m=[ m 0… m N ], k =[ k ( t i , t j )]| ij,0≤i,j≤N

[0055] This trajectory is a Gaussian process, which can be expressed as: x ( t )=GP( m ( t ), k ( t , t ′))

[0056] The corresponding Gaussian distribution can be expressed as: , Among them, the exponential part uses the Mahalanobis distance, which is: ;

[0057] p ( x ) is a priori function, p ( x ) to decompose: , in, Represents the prior state distribution on the first trajectory state quantity, expressed as:

[0058] Represents the prior state distribution on the state quantity of the final trajectory, expressed as:

[0059] Represents the Gaussian process distribution between two adjacent states, expressed as:

[0060] In this way, we can get x gp ( t ); see the attached figure, Figure 2 Schematic diagram of Gaussian process regression trajectory prediction according to an embodiment of the present invention. X gp express{ x gp}UAV position prediction information in;

[0061] Similarly, to predict the obstacle status information, the obstacle likelihood function is expressed as: , in, d p Contains the distance between the obstacle and the drone and the state of the obstacle itself. When this function is activated, it will predict the state of the obstacle and prompt the drone's flight trajectory to avoid the obstacle and its repulsive field. Finally, p ( s obs | o t ) to predict obstacle status information { o gp}.

[0062] Step 4: Replan the potential field function, predict the secondary target point according to the trajectory predicted in step 3, interpolate the secondary target point between the expected target point and the drone, and replan the virtual potential field function and virtual force function:

[0063] After the trajectory planning has been completed, when the drone is flying according to the pre-planned trajectory, if the target point changes, the position of obstacles in the environment changes, or the execution trajectory deviates from the planned trajectory, it is necessary to quickly re-plan a feasible trajectory from the current state to the target state online.

[0064] Step 4.1: Obtain secondary target point information, predict the next moment flight trajectory of the drone according to the Gaussian process regression algorithm in step 3, calculate interpolation, obtain secondary target point information, and obtain obstacle prediction information at the same time to prepare data for step 4.2; Gaussian process regression algorithm predicts the next moment flight trajectory of the drone { x gp}, { x gp}The location information is X gp , { x gp The speed prediction information is V gp , in the prediction trajectory and Z t Select the secondary target point of trajectory planning between coordinate points Z * t , this embodiment selects Z * t = X gp ; The predicted obstacle state information is { o gp}, { o gp}The location information is O gp , { o gp}The speed information is V ogp .

[0065] Step 4.2: Replan the virtual force function. Plan the secondary target gravitational field function according to the secondary target point information obtained in step 4.1. Plan the improved repulsive field function and tangential function based on the obstacle prediction information. Combine the gravitational field function in step 2 with the secondary target gravitational field function, and calculate the resultant force function together with the improved repulsive field function and tangential function.

[0066] According to the target point information, the target gravitational field function is constructed as follows: , in, p ( s | x t ) is the secondary target point probability model, for The gravitational constant;

[0067] The secondary target gravity function is:

[0068] The combined force of target gravity and sub-target gravity is expressed as:

[0069] According to the obstacle status information, the improved repulsion field is replanned:

[0070] The velocity compensation function with the predicted repulsive field is expressed as:

[0071] The improved repulsive field of replanning is expressed as:

[0072] Replan the repulsive field:

[0073] Replan the tangent function:

[0074] Replan the tangential force function:

[0075] See attached figure, Figure 3 Schematic diagram of re-planning virtual force according to an embodiment of the present invention, wherein Fatt ( t ), F * att ( t ), F * r ( t )and F * q ( t ) is the virtual force control after replanning the trajectory of the UAV, and the final resultant force function is expressed as:

[0076] Step 5: Calculate the expectation, based on the virtual force function re-planned in step 4 F * ( t ),right F * ( t ) can be used to calculate the expected acceleration and expected velocity of the drone's world coordinate system. The calculation expression is: , in, a * ( t ) is the expected acceleration of the UAV after trajectory replanning, v * ( t ) is the expected velocity of the UAV after trajectory replanning.

[0077] The improved artificial potential field trajectory planning method for UAV of the present invention uses Gaussian process regression to predict the probability of the UAV flight trajectory at the next moment, interpolates the planned trajectory in real time through probability distribution, re-plans the UAV trajectory, reduces the jitter problem caused by searching for local minimum values ​​during the flight of the UAV, and optimizes the smoothness of the flight trajectory in the process of planning the UAV flight mission using the artificial potential field method.

[0078] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A UAV flight system for use in a dynamically changing environment, comprising five parts: information release, improved artificial potential field trajectory replanning method, UAV control module, UAV flight module, and UAV / ground station feedback. The improved artificial potential field trajectory replanning method is used to obtain target information, UAV state information, and environmental state information, construct an improved artificial potential field, predict the state trajectory, replan the improved artificial potential field, calculate the expectation and pass it to the UAV flight control module, and calculate the UAV flight control quantity in combination with the feedback, so as to realize the UAV flight task under the dynamically changing environment; its characteristics are: The improved artificial potential field trajectory replanning method includes 5 steps: Step 1: Information acquisition, obtain the target point and status information. The target point information is the three-dimensional position of the target point in the world coordinate system. The status information includes the three-dimensional position coordinate data of the UAV in the world coordinate system, the speed data of the UAV in the world coordinate system, the attitude angle data in the body coordinate system, the angular velocity data of the UAV in the body coordinate system, and the environmental information status data of the UAV during flight, so as to prepare data for the subsequent steps; Step 2: Establish a potential field function, update the expected target point according to the data information in step 1, and construct an improved artificial potential field method to plan the virtual potential field function and virtual force function; Step 3: Gaussian process regression is used to predict the trajectory. A sparse Gaussian distribution function is constructed based on the local minimum value of the potential field function and the limited set of historical data obtained in step 1. The trajectory of the drone and obstacle status information are predicted through Gaussian process regression. Step 4: Replan the potential field function, predict the secondary target point according to the trajectory predicted in step 3, interpolate the secondary target point between the expected target point and the UAV, and replan the virtual potential field function and virtual force function; Step 5: Calculate the expectation. Integrate the virtual force function replanned in step 4 to calculate the expected acceleration and expected velocity of the drone's world coordinate system.

2. The UAV flight system for use in a dynamically changing environment according to claim 1, characterized in that: The improved artificial potential field method plans the virtual potential field function and the virtual force function, including six modules: gravitational field function, gravitational function, improved repulsive field function, improved repulsive function, tangential function, and tangential force function. The gravitational field function is used to calculate the virtual gravitational field in the UAV trajectory planning, and the gravitational function calculates the virtual gravity through the gravitational field. The improved repulsion field function is used to calculate the virtual repulsion field in the trajectory planning of the UAV. The improved repulsion function calculates the virtual repulsion by improving the repulsion field function. The improved repulsion field function and the improved repulsion function are described as: ， in, U r ( t ) is to improve the repulsive field function, F r ( t ) is the planned target point of the UAV O t The repulsive force, V r for t The relative difference in speed between the UAV and the obstacle in the dynamic information at the moment is expressed as: V r =|| V t / ( V t - V ot )|| V ot The tangential function and tangential force function are used to calculate the virtual tangential force in the trajectory planning of the UAV. The function description is: ， in, U q ( t ) is the tangent function, F q ( t ) is the planned tangential force.

3. The UAV flight system for use in a dynamically changing environment according to claim 1, characterized in that: The Gaussian process regression prediction trajectory determines that the input is state information. If it is, a sparse Gaussian distribution function is constructed using a limited set of historical data to construct a Gaussian process regression model of the drone and obstacles, and the flight trajectory state and obstacle state information of the drone at the next moment are predicted; if not, information acquisition is returned.

4. The UAV flight system for use in a dynamically changing environment according to claim 1, characterized in that: The replanning of the potential field function includes two steps: obtaining a secondary target point and replanning a virtual force function; Step 4.1: Obtain the secondary target point information, predict the next moment flight trajectory of the drone according to the Gaussian process regression algorithm in step 3, calculate interpolation, obtain the secondary target point information, and obtain obstacle prediction information at the same time to prepare data for step 4.2; Step 4.2: Replan the virtual force function. Plan the secondary target gravitational field function according to the secondary target point information obtained in step 4.

1. Plan the improved repulsive field function and tangential function based on the obstacle prediction information. Combine the gravitational field function in step 2 with the secondary target gravitational field function, and calculate the resultant force function together with the improved repulsive field function and tangential function.

Images