NL2031641B1 - Uav path planning method based on obstacle position prediction and improved artificial potential field method in unknown environment - Google Patents

Abstract

The present invention discloses a UAV path planning method based on obstacle position prediction and an improved artificial potential field method in an unknown environment. 5 In order to discretize position information of an obstacle, this method uses coordinates of the obstacle in a grid to represent the position information of the obstacle. By detecting and recording obstacle movement information, the UAV can obtain a position of the obstacle at next moment by means of Markov prediction according to probabilities that the obstacle moves toward various directions around. On this basis, the UAV avoids 10 the obstacle in advance by means of an artificial potential field, which improves the safety of UAV path planning. In addition, aiming at a local minimum of the traditional artificial potential field, the UAV can predict whether it will fall into the local minimum when predicting the position of the obstacle, and then take the minimum point as a virtual obstacle for avoidance. Moreover, after falling into a local oscillation, the UAV 15 gradually increases a gravitational coefficient to suppress sudden changes of gravitational and repulsive forces, and guides itself to escape from the local oscillation.

Description

UAV PATH PLANNING METHOD BASED ON OBSTACLE POSITION

PREDICTION AND IMPROVED ARTIFICIAL POTENTIAL FIELD METHOD

IN UNKNOWN ENVIRONMENT

TECHNICAL FIELD

[01] The present invention relates to the field of unmanned aerial vehicle (UAV) path planning, in particular to a UAV trajectory planning method for a UAV to avoid obstacles flying in a dynamic environment.

BACKGROUND ART

[02] In recent years, UAV technology has been widely used in the civil and military fields, and UAV path planning algorithms have attracted wide attention. However, there are potential threats of collisions during UAV flight.

[03] At present, UAV path planning technology has become a hot spot, and various algorithms emerge one after another. When classic algorithms deal with emergencies and dynamic environments, the path planning strategy has a low success rate and a large amount of calculation, making it difficult to complete real-time path updates. Heuristic algorithms that have low efficiency and fail to search in complex environments are not suitable for UAV path planning in unknown dynamic environments.

SUMMARY

[04] The technical solution of the present invention is: a UAV path planning method based on obstacle position prediction and an improved artificial potential field method in an unknown environment, characterized by including:

[05] establishing a UAV and obstacle model under the condition that a UAV operates at a fixed height, and obtaining an influence radius L of an obstacle;

[06] establishing a Markov obstacle path prediction model according to an obstacle grid map, to predict a trajectory of the obstacle; and

[07] improving the traditional artificial potential field, and combining with a predicted obstacle position, to obtain a dynamic UAV path planning.

[08] By adopting the above technical solution, the present invention can achieve the following technical effects: the UAV avoids an obstacle in advance by adopting an artificial potential field, which improves the safety of UAV path planning. In addition, aiming at the local minimum of the traditional artificial potential field, the UAV can predict whether it will fall into the local minimum when predicting the position of the obstacle, and then take the minimum point as a virtual obstacle for avoidance. Moreover, after falling into a local oscillation, the UAV gradually increases the gravitational coefficient to suppress sudden changes of gravitational and repulsive forces, and guides itself to escape from the local oscillation. The method of the present invention can improve the effect of avoiding dynamic obstacles in the artificial potential field method, and avoid the local minimum generated by the artificial potential field method, so that the UVA can escape from the local oscillation.

BRIEF DESCRIPTION OF THE DRAWINGS

[09] FIG. 1 is a flowchart of an implementation process of the present invention;

[10] FIG. 2 is an obstacle influence model diagram, a is an obstacle influence model diagram of the present invention, and b is an ideal obstacle influence model diagram;

[11] FIG. 3 is a flight attitude diagram of an obstacle of the present invention;

[12] FIG. 41s an obstacle grid map of the present invention;

[13] FIG. 5 is a schematic principle diagram of a local minimum;

[14] FIG. 6 is a schematic principle diagram of local oscillation at a target point;

[15] FIG. 7 is an instance parameter diagram of the present invention;

[16] FIG. 8 is an effect diagram of k-order Markov obstacle prediction according to the present invention;

[17] FIG. 9is a path principle diagram under the traditional artificial potential field method;

[18] FIG. 10 is a path principle diagram under the HOPA algorithm of the present invention;

[19] FIG. 11 is a UAV trajectory diagram under the traditional artificial potential field method;

[20] FIG. 12 is a UAV trajectory diagram under the HOPA algorithm of the present invention; [BIJ] FIG. 13 is an enlarged view of obstacle avoidance under the traditional artificial potential field method,

[22] FIG. 14 is an enlarged view of obstacle avoidance under the algorithm of the present invention;

[23] FIG. 15 is a schematic principle diagram of a local minimum under the traditional artificial potential field method;

[24] FIG. 16 is a schematic principle diagram of a local minimum under the algorithm of the present invention; and

[25] FIG. 17 is a schematic principle diagram of solving a local oscillation by the algorithm of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[26] This example provides a UAV path planning method based on obstacle position prediction and an improved artificial potential field method in an unknown environment, including the following steps:

[27] Sl: in the presence of static and dynamic obstacles in an environment when a

UAV operates at a fixed height, a UAV and obstacle model is established, and an influence radius of the obstacles is obtained,

[28] Under normal circumstances, the flight speed of the UAV relative to the earth coordinate system may be described as: o =V cos cos wr =V cosOsing = =Vsm6G 29] dt (1)

[30] Where, (x, v, z) represents the position of the UVA in the earth coordinate system, J represents the speed of the UVA, # and ¢ represent the pitch angle and tilt angle of the UVA respectively. In the specific example of the present invention, the flight speed of the UVA is set to 5 ms. The flight height of the UVA ranges from 70 to 500 meters. In such a narrow flight space, the UAV should try to avoid collisions by changing the course or speed instead of the height. The present invention realizes collision avoidance by changing the course and speed in a horizontal plane, so the pitch angle is zero. The coordinates used in the obstacle avoidance issue are two-dimensional

Cartesian plane coordinates (x, y).

[31] In order to increase the safety of the UVA during flight, the UVA and the obstacle are regarded as a sphere respectively in the present invention. The geometric center (,,,., of the UVA is the origin of the UVA sphere, the longest length d,,., of the UVA body is the diameter, and then the radius of the UVA sphere is A1, =d,.,/ 2. Correspondingly, the center of the obstacle sphere is O, and the radius is KR, . Since the collision issue studied by the present invention occurs in a plane, the sphere representing the UAV becomes a circle with a center of (J, and a radius of R,;, in the two-dimensional

Cartesian plane coordinate system, and the obstacle sphere becomes a circle with a center of (J, and a radius of R, in the two-dimensional Cartesian plane coordinate system. In the example of the present invention, the radii of the UAV and the obstacle are both 0.5 m.

[32] When the UAV detects an obstacle, its course angle will be offset by Ap when the path planned by the artificial potential field method is used to avoid the obstacle. The

UAV 1s affected by its own physical performance, the offset speed of its course angle is w, and there is a maximum deflection angle Ag, Even if the artificial potential field method provides that the course angle Ao required to avoid the obstacle is greater than the Ag, the UAV can only be offset by Ag, to avoid the obstacle, but danger may occur at this time. The time required for deflecting Ag, is 7, = A9, /®. In this embodiment, referring to the parameters of the DJI UVA, the turning angle of the UVA is set to 60°, and the offset speed of the course angle is 150° /s, so 7, here is 0.8 s.

[33] The minimum safety distance of the UVA is Lm =R +R, . When the distance between the centers of the UVA and obstacle spheres is greater than ZL, it can be regarded as that the flight path of the UVA is safe.

[34] The UVA is assumed to fly towards a target point at a constant speed 1. The line connecting the position F, of the UVA and the position P, of the obstacle is

B. FP, - The projection of the UVA speed on the line PP, is F‚, . After the UAV is deflected by Ao, the projection speed of the UAV on the PP, is Vo tr — and

Vg ie sw <V. zw as shown in FIG. 2.

[35] The flight distance of the UAV in time ¢, is s=¥,, -¢, , and s is 4 m by calculation in the example of the present invention. Accordingly, L=s+ KR; +R, can be obtained, where L is 5 m, which is considered to be a safe distance between the UVA and the obstacle. When the UAV and the obstacle face each other, 5 =O Hogs),

The speed of the obstacle in this simulation example is also 57/5. Then, sis 8 m, L is 9 m, and a maximum obstacle collision distance can be obtained.

[36] When the distance between the UVA and the obstacle is less than L, there will be a potential collision risk. When the distance is between 2L and L, the UAV can observe the obstacle. In addition, when the distance is less than L, the UVA begins to avoid the obstacle. Because 2L > 27, , this ensures the flight safety of the UVA. In the present invention, the influence radius of the obstacle is set to L.

[37] S2: a Markov obstacle path prediction model is established based on an obstacle grid map to predict an obstacle trajectory.

[38] Specifically, the UAV detects the position of an obstacle through a sensor to obtain the position of the obstacle relative to the UAV. When the present invention predicts the position of the obstacle, a two-dimensional Cartesian rectangular grid is used to represent the environment. The selection of the grid size is also related to the performance of the sensor. If the sensor has high accuracy and fast speed, a smaller grid may be selected. The obstacle information measured by the sensor may be mapped to an environment coordinate system: {x =x, +dcos(T +1)

[39] zi =>, tdsin(l, +1;) 2)

[40] In formula (2), (x,,y.) is the coordinates of a measured point in the environment coordinate system, (x.,v. 7.) is the position of the UVA in the environment coordinate system, 7 is an angle between the UVA and the X axis of the environment coordinate system, 7, is an angle between the measured point and the X axis of the environment coordinate system, and d is the distance between the UAV and the measured point. The center of the UAV body is selected as the origin of the UAV coordinate system. Through formula (3), the coordinates (x,,y.) may be mapped to the corresponding cell(i, J) in the environment coordinate system. ; = x /ele+[e12] v.=| Vv /clc+/e/2

[41] © >: | 2] (3)

[42] In formula (3), (x,y) is the coordinates of the cell(i,]) in the environment coordinate system; c is the width of the cell(i, j). | | indicates that the number in the bracket 1s rounded up.

[43] In this grid coordinate system, the obstacle may fly from the current position to 9 flight course angles around, as shown in FIG. 3. The 9 flight course angles represent east, northeast, north, northwest, west, southwest, south, and southeast respectively. It is supposed that the obstacle can only move 1 cell around in each time interval t, so the width of w is a moving step length of the obstacle in the unit time t. This step length is a distance detected by the UVA that the obstacle moves within the unit detection interval. Therefore, the obstacle has 9 flight elements, which correspond to the 9 flight course angles described above. The nine course angles of the obstacle correspond to Ox, 1 1 3 5 3 7 . 7

A, =n, —A, A, —T, —n, —T and the original position from 1 to 9 respectively, as 4 2 4 4 2 4 shown in FIG. 4.

[44] The object studied in the present invention is an obstacle. Its state is gridded in space. The current position of the obstacle occupies a cell. The other nine directions that it may move to are nine states. The nine directions are respectively NW, N, NE, W, E,

SW, S, SE and the original position, which correspond to 9 states E.E, respectively. E € E,i=12,..9. The matrix P” is an m-order state transition matrix of the

Markov chain, and a recurrence relationship can be obtained by using the Chapman-

Kolmogorov equation:

[45] pr — pip — php! (4)

[46] Pp — (P' y (5)

Pa 7 == Pi

PP. - P,

[47] p=" 2 » (6) : bo 5 Fn == Fy

Me jj 12,9

Py = 5 7 JE 2 Se

[48] DN ; 7

[49] Where N, represents a number of times that the obstacle changes from state / to j. The number of changes is observed by the UVA when the distance between the UVA and the obstacle 1s between 2L and L.

[50] Based on the Markov chain, a predicted value of each position change is obtained by weighting:

[51] X()=aS(-1)P' +L +a, S(t-m)P" +L +a, S(t —k)P* (8)

[52] Inthe formula, 7 is the current detection interval, £—1 is the previous detection interval, and so on, until the m detection interval (£-m). X(t) is a probability of next position predicted by the weighting formula and is a 1x9 matrix, and each element is a probability of the corresponding flight element. S¢:—m).1<m<k represents a movement state of the UVA at the m detection interval. k is a maximum order in the prediction process.

[83] a,L ‚aL ‚a, are weights, which represents the degree of influence of the previous LL mL ‚& movement on the next movement, and can be determined experience. The present invention adopts an autocorrelation coefficient determination method to calculate the weights: km

A(t) A(t-m

[54] n= Zee A) Am) ( ) - (=m) (9) > À ()

[55] Inthe formula, r, represents an m-order autocorrelation coefficient, (7) is the current movement of the obstacle, and A(-m)e{E.E,.….E,} is the movement of the

UAV in the m time intervals at the current time. The autocorrelation coefficient needs to be normalized by formula (10), that is:

I]

[56] a, = k (10)

Dome m=1

[57] mm is the maximum order calculated according to the prediction requirement.

[58] Only the maximum element in X(7) needs to be selected, the direction cell corresponding to the position of the maximum element is used as the predicted next moving direction, and then the moving coordinate value corresponding to the predicted next moving direction is added to the current position of the obstacle to obtain a predicted position of the obstacle.

[59] S3: the traditional artificial potential field is improved and combined with the predicted position of the obstacle to obtain a dynamic UAV path planning.

[60] When the UVA flies to the target point, there will be a certain point where the gravitational and repulsive forces are the same, so that the UVA is stuck in a stalemate and cannot move. This is called a local minimum. At this time, the resultant force of repulsive forces on the UAV is equal to the gravitational force in an opposite direction, as shown in FIG. 5.

[61] The present invention introduces obstacle position prediction, and the movement of the UAV itself can be grasped, so it can be predicted whether the UAV will encounter a local minimum in the future. A virtual obstacle can be placed on the possible local minimum point by means of the obstacle position prediction to avoid the local minimum. The new repulsive force of the virtual obstacle on the UVA is:

[62] FE B en ! yt Va Poes Pain) (11) rep virtual = ’ rE a 3 min on A (Pors By) d, qd’ (Peres Lo) ow

[63] The virtual obstacle is a mass point without radius, and its center of mass is P .

P_, represents the current position of the UAV, d(D.:Pwix) 1s the distance between the current UAV and the virtual obstacle, d, is an obstacle influence distance, and £ is a repulsion coefficient. In addition to the original repulsive force rr and gravitational force F, , the UVA will also receive a new repulsive force F,,, vrat of the virtual obstacle. Therefore, the potential field where the UAV is located is changed, so that the UAV has to re-plan the path.

[64] When a path is planned using the traditional artificial potential field method, unnecessary oscillations may occur. As shown in FIG. 6, when there are obstacles around the target point, the UVA will not be able to reach the target point.

[65] In face of this situation, the UAV can increase the gravitational coefficient. The temporary increment of the gravitational coefficient «« adopts a gradient increment method, and each gradient increment is a, . When the UAV falls into a local oscillation, the gravitational coefficient increases by à «,, making the gravitational force greater than the repulsive force. If the gravitational force is equal to the repulsive force just after a certain number of a, is increased, a a, continues to increase to determine à. The temporary increment Aa = Ag, .

[66] From the original gravitational force function and repulsive force function formulas of the artificial potential field method, after the sum of gravitational and repulsive forces is enhanced, the force on the UAV is shown in formula (12):

EF, (9) = (a + Aa) * d(g, Goat) 1 1 1

[67] FE (@)=8 TT) Vd(q,4,,) (12) ” d(q, Dons) d, d (3, as)

Pt (9) > Fo (9)

[68] Where F(q) is a gravitational force after the gravitational coefficient is adjusted, and d(q.q,,,) is the distance between the UVA and the target point. F, (9) is the repulsive force received by the UAV and remains unchanged when the gravitational force is increased, d(4,%,,,) is the distance between the UAV and the obstacle, and d, is the obstacle influence distance, that is, L above.

[69] FIG. 7 shows model parameter settings, where the gravitational coefficient, repulsion coefficient, UAV flight speed and angular velocity, maximum turning angle, and obstacle radius are set. An obstacle that moves irregularly in two-dimensional

Cartesian coordinates is set in the simulation. The order of the Markov chain ranges from 1 to 10, and the prediction accuracy of each order is averaged by 30 group of random moving obstacles. As shown in FIG. 8, the obstacle position prediction method proposed herein can effectively predict the position of a moving obstacle. Under normal circumstances, the probabilities that the obstacle moves in 9 directions are all 11.1%.

The prediction method proposed in the present invention has an accuracy of 40% in the first order, and when the order is changed from 1 to 2, the accuracy is significantly improved. After the order 5, the accuracy rate reaches relatively stable 73%.

[70] In the simulation diagram of the flight path of the UVA, the white hollow circle represents the starting point of the UVA, the hollow hexagram represents the target point, the black solid circles represent the obstacle, and the connection between two black solid circles and an arrow indicate that the obstacle moves along the arrow.

[71] In FIGS. 9 and 10, when the target point is (15,15), and the obstacle represented by the black circle flies along a straight line from (14,1) to (4,9), the UVA will collide with the obstacle. The UVA changes the way to avoid collision. FIG. 9 shows a UAV path using traditional APF. FIG. 10 shows a UVA path calculated by HOPA of the present invention. Compared with FIG. 9, the UAV in FIG. 10 turns in a safer different direction.

The result shows that the obstacle position prediction of the k-order Markov chain in the

HOPA algorithm of the present invention can help the UAV find a shorter and safer path.

[72] FIGS. 11 and 12 show a trajectory under the traditional artificial potential field method and a trajectory based on the Markov and artificial potential field hybrid algorithm of the present invention when the UVA performs a flight mission. In the figures, the white hollow circle represents the starting point of the UVA, the hollow hexagram represents the target point, the black solid circles represent the obstacle, and the line connecting two black solid circles and an arrow indicate that the obstacle moves along the arrow. The UAV in FIG. 12 starts to avoid the obstacle earlier at a longer distance than the UAV in FIG. 11 when encountering the obstacle. The detailed comparisons of FIGS. 11 and 12 are shown in FIGS. 13 and 14, indicating that the algorithm of the present invention improves the safety of obstacle avoidance.

[73] FIG. 15 shows that under the traditional artificial potential field method, the

UAV travels to a specific point, the gravitational and repulsive forces received by the

UAV are the same, the UAV falls into a local minimum, then reaches an impasse, and cannot continue to move. The local minimum point in FIG. 15 is set as a virtual obstacle in FIG. 16 under the algorithm of the present invention, and the UAV receives a repulsive force of the virtual obstacle, so the local minimum point is avoided.

[74] As shown in FIG. 17, when the oscillation of the UAV is detected, the algorithm of the present invention increases the gravitational coefficient of the target point to increase the gravitational force of the target point on the UAV, so that the UAV can get rid of the oscillation between points 2 and 3, move in the order of points 3, 4, 5, and 6 by means of the improved algorithm of the present invention after a period of oscillation, and reach the target point smoothly.

Claims (3)

Conclusies L UAV-padplanningssysteem op basis van obstakelpositievoorspelling en een verbeterd kunstmatigepotentiaalveldwerkwijze in een onbekende omgeving, gekenmerkt door het volgende te omvatten: een straalverwervingsmodule die geconfigureerd is om een UAV-en- obstakelmodel tot stand te brengen in een NVIDIA CarmelARM® v8.2 van Jetson Xavier NX op voorwaarde dat een UAV werkzaam is op een vaste hoogte, om een invloedsstraal L van een obstakel te verkrijgen, en de invloedsstraal op te slaan in een Samsung 128 GB geheugenkaart van het systeem; een trajectorievoorspellingsmodule die geconfigureerd is om een Markovobstakelpadvoorspellingsmodel tot stand te brengen in de NVIDIA CarmelARM® v8.2 van Jetson Xavier NX volgens een obstakel rasterkaart, het model op te slaan in de Samsung 128 GB geheugenkaart van het systeem, en een trajectorie van het obstakel te voorspellen; en een padplanningsmodule die geconfigureerd is om het traditionele kunstmatigepotentiaalveld te verbeteren, en het verbeterde kunstmatigepotentiaalveld te combineren met een voorspelde obstakelpositie in de NVIDIA CarmelARM* v8.2 van Jetson Xavier NX om een dynamische UA V-padplanning te verkrijgen.Conclusions L UAV path planning system based on obstacle position prediction and an improved artificial potential field method in an unknown environment, characterized by comprising: a beam acquisition module configured to establish a UAV and obstacle model in an NVIDIA CarmelARM® v8.2 from Jetson Xavier NX under the condition that a UAV operates at a fixed altitude, to obtain an influence radius L of an obstacle, and store the influence radius in a Samsung 128 GB memory card of the system; a trajectory prediction module configured to create a Markov obstacle path prediction model in the NVIDIA CarmelARM® v8.2 of Jetson Xavier NX according to an obstacle raster map, store the model in the system's Samsung 128 GB memory card, and create a trajectory of the obstacle predictable; and a path planning module configured to enhance the traditional artificial potential field, and combine the enhanced artificial potential field with a predicted obstacle position in the NVIDIA CarmelARM* v8.2 of Jetson Xavier NX to obtain dynamic UA V path planning. 2. UAV-padplanningssysteem op basis van obstakelpositievoorspelling en een verbeterd kunstmatigepotentiaalveldwerkwijze in een onbekende omgeving volgens conclusie 1, met het kenmerk dat tijdens het tot stand brengen van een UAV-en- obstakelmodel, de straalverwervingsmodule specifiek geconfigureerd is om: de UVA en het obstakel respectievelijk als een bol te beschouwen, waarbij het geometrische centrum Oyy 4 van de UVA de oorsprong van de UVA-bol is, de langste lengte dyy4 van het UVA-lichaam de diameter is, dan de straal van de UVA-bol Ry, = dyva/2 1s, en evenredig, het centrum van de obstakelbol Ops is en de straal Rops is; waarbij, aangezien de botsing in een vlak plaatsvindt, de UA V-bol een cirkel wordt met een centrum van Oyy4 en een straal van Ryya4 in een tweedimensionaal Cartesiaansvlakcoördinatensysteem, en de obstakelbol een cirkel wordt met een centrum van Oops en een straal van Rops in het tweedimensionale Cartesiaansvlakcoördinatensysteem.2. UAV path planning system based on obstacle position prediction and an improved artificial potential field method in an unknown environment according to claim 1, characterized in that during the creation of a UAV and obstacle model, the beam acquisition module is specifically configured to: the UVA and the obstacle to be regarded respectively as a sphere, where the geometric center Oyy 4 of the UVA is the origin of the UVA sphere, the longest length dyy4 of the UVA body is the diameter, then the radius of the UVA sphere Ry, = dyva /2 1s, and proportional, the center of the obstacle sphere is Ops and the radius is Rops; where, since the collision occurs in a plane, the UA V sphere becomes a circle with a center of Oyy4 and a radius of Ryya4 in a two-dimensional Cartesian plane coordinate system, and the obstacle sphere becomes a circle with a center of Oops and a radius of Rops in the two-dimensional Cartesian plane coordinate system. S13 -S13 - 3. UAV-padplanningssysteem op basis van obstakelpositievoorspelling en een verbeterd kunstmatigepotentiaalveldwerkwijze in een onbekende omgeving volgens conclusie 1, met het kenmerk dat: indien het obstakel gedetecteerd wordt, de koershoek van de UAV verschoven wordt met A, waarbij de verschuivingssnelheid van de koershoek w is, en de tijd die genomen wordt om de hoek AQ, af te buigen tg = A@/w is; aangenomen wordt dat de UVA richting een doelpunt vliegt met een constante snelheid Vgay, en de lijn die de positie Py4y van de UVA en de positie Pops van het obstakel verbindt Py4y Poes is; waarbij de projectie van de UVA-snelheid op de lijn Pyav Pops het volgende is Vref yay; waarbij nadat de UAV met Ag afgebogen is, de projectiesnelheid van de UAV op de Pyay Pors het volgende is Vier vav nieuw, en Vier vav_nreuw < Vref VAV; de vliegafstand van de UAV in een tijd tg het volgende is s = Vysy tg, L = s + Ryay + Rops verkregen wordt, en deze afstand als een veilige afstand tussen de UVA en het obstakel gezien wordt, waarbij indien de UAV en het obstakel naar elkaar gericht zijn, 5s = Vyay +Vops) tp , en een maximale obstakelbotsingsafstand verkregen wordt; indien de afstand tussen de UVA en het obstakel minder dan L is, er een mogelijk botsingsrisico zal zijn; waarbij indien de afstand tussen 2L en L is, de UAV het obstakel kan observeren; waarbij aanvullend, indien de afstand minder dan L is, de UVA begint met het ontwijken van het obstakel, waarbij de invloedsstraal van het obstakel op L ingesteld wordt.3. UAV path planning system based on obstacle position prediction and an improved artificial potential field method in an unknown environment according to claim 1, characterized in that: if the obstacle is detected, the heading angle of the UAV is shifted by A, where the heading angle shifting speed is w , and the time taken to deflect the angle AQ, is tg = A@/w; it is assumed that the UVA is flying towards a target with a constant velocity Vgay, and the line joining the position Py4y of the UVA and the position Pops of the obstacle is Py4y Poes; where the projection of the UVA velocity on the Pyav Pops line is Vref yay; where after the UAV is deflected with Ag, the projection speed of the UAV onto the Pyay Pors is Four vav new, and Four vav_nreuw < Vref VAV; the flight distance of the UAV in a time tg is obtained as follows: s = Vysy tg, L = s + Ryay + Rops, and this distance is considered a safe distance between the UVA and the obstacle, where if the UAV and the obstacle are directed towards each other, 5s = Vyay +Vops) tp , and a maximum obstacle collision distance is obtained; if the distance between the UVA and the obstacle is less than L, there will be a possible risk of collision; where if the distance is between 2L and L, the UAV can observe the obstacle; additionally, if the distance is less than L, the UVA starts avoiding the obstacle, setting the obstacle's radius of influence to L.