Professional Documents
Culture Documents
A Case-Based Online Trajectory Planning Method of Autonomous Unmanned Combat Aerial Vehicles With Weapon Release Constraints
A Case-Based Online Trajectory Planning Method of Autonomous Unmanned Combat Aerial Vehicles With Weapon Release Constraints
15040
© 2020, DESIDOC
ABSTRACT
As a challenging and highly complex problem, the trajectory planning for unmanned combat aerial vehicle
(UCAV) focuses on optimising flight trajectory under such constraints as kinematics and complicated battlefield
environment. An online case-based trajectory planning strategy is proposed in this study to achieve rapid control
variables solution of UCAV flight trajectory for the of delivery airborne guided bombs. Firstly, with an analysis of
the ballistic model of airborne guided bombs, the trajectory planning model of UCAVs is established with launch
acceptable region (LAR) as a terminal constraint. Secondly, a case-based planning strategy is presented, which involves
four cases depending on the situation of UCAVs at the current moment. Finally, the feasibility and efficiency of the
proposed planning strategy is validated by numerical simulations, and the results show that the presented strategy
is suitable for UCAV performing airborne guided delivery missions in dynamic environments.
Keywords: Unmanned combat air vehicle; UCAV; Trajectory planning; Receding horizon control; Threat
environment
374
Tang, et al.: A Case-based Online Trajectory Planning Method of Autonomous UCAVs with Weapon Release
375
Def. SCI. J., Vol. 70, No. 4, July 2020
x f − x ≤ εx
yf − y ≤ εy , (9)
h f − h ≤ ε h
where ε x , ε y and ε h denote the permitted distance errors
along the axis as determined by the range of the LAR, while
(x f , y f , h f ) represents the terminal point.
Moreover, the state constraints of the UCAV to launch
airborne bombs are
V (t f ) = V f
, (10)
ψ (t f ) = ξ f
amin ≤ a ≤ amax ,
ωmin ≤ u2 ≤ ωmax , (11)
γ ≤ u ≤ γ ,
min 3 max
V
r ≥ rmin = ,
Figure 1. Flow chart of acquiring LAR. ωmax
x = ∑ xi N p
i =1 (12)
, (8)
Np where p (h ) = (x, y ) refers to the 2D position of the UAV with
y = ∑ yi N p its flight height h . Fm ( p (h )) < 0 indicates that the UCAV is
i =1
capable to keep flying without considering the hard threat H m.
where (xi , yi ) indicates the i -th vertex of LAR envelope, and
N p denotes the number of vertices. Therefore, the terminal (b) Soft Threats : It refer to the safety threats related to air
constraint is transformed into defense weapons, such as IADS and air defense artillery.
376
Tang, et al.: A Case-based Online Trajectory Planning Method of Autonomous UCAVs with Weapon Release
Different from the complete avoidance of hard threats, UCAV the weights of the first and third objective, ϖ > β ≥ 0 and
ought to maintain a safe distance from soft threats. There ϖ + β = 1 ; κ > 0 means a scaling factor for equalisation of the
are M 2 soft threats, and the threat function of the j -th dimension of the two terms.
( j = 1, , M 2 ) soft threat is The complete constraints are listed as follows:
1
− 2 (p (h ) − B j (h ))
T
Vmin ≤ u1i ≤ Vmax
(13)
Th j ( p (h )) = λ (h )exp ,
amin ≤ a ≤ amas
Q −j 1 (h )(p (h ) − B j (h )) ω ≤ u i ≤ ω
min 2 max
where λ (h ) indicates the threat intensity related to the flight γ min ≤ u3i ≤ γ max
height h , while B j (h ) and Q j (h ) ≥ 0 denote the position and V
r ≥ rmin = ,
threat range respectively. When the UCAV flies at a constant ωmax (18)
altitude, or the threat intensity varies little as the flight height (
Fj pUAV i (t ) < 0 )
changes, λ (h ) = 1 .
Th j ( p (h )) indicates the probability of UCAVs being
(
Th j pUAV i (h ) > ζ )
attacked or annihilated by the soft threat. Total soft threat (
C pUAV i1 , pUAV i2 ≤ 0 )
fitness is referred to as the summation of all soft threat values, i1 = 1, , N , i2 = 1, , N , i1 ≠ i2 ; j = 1, M 1
i.e.
M2 For the purpose of simplification, a receding horizon
Th ( p (h )) = ∑ Th j ( p (h )). (14) approach is adopted, i.e.
j =1
N N
( )
tk +δ
A lower soft threat fitness suggests a safer situation. min ∫ ϖ ∑ pUAV i (t ) − qTi (t ) + κβ∑ Th pUAV i (t ) dt ,
A threshold ζ for soft threats needs to be identified in {p
UAV i
(t ),uUAV i (t )}
tk
i =1 i =1
advance, the value of which represents the tolerance degree of (19)
soft threats.
where δ indicates the interval length of [tk , tk + δ ⋅ Dt ], and Dt
refers to the time step. It is noteworthy that an excessive value
2.3.2.3 Collision Avoidance Constraints
of Dt can result in a poor solution accuracy.
In coordinated missions where multiple UCAVs are
required, the collisions between the aerial vehicles must be
3. Case-based Solution model of
avoided. To this end, there is a safe distance threshold RUAV .
delivery trajectory planning
And
To solve the aforementioned trajectory planning problem,
( ) ( )−
2
C pUAV i , pUAV j = RUAV i + RUAV j pUAV i − pUAV j , (15) a novel and practical case-based method is put forward to
2
achieve online planning in dynamic battlefields.
is defined as the collision function between UAV i and UAV j ,
where ⋅ 2 represents 2-norm. Besides, no collision is
3.1 Key Parameters Identification
(
guaranteed by assuring C pUAV i , pUAV j ≤ 0 . ) The current state vector of UAV i at time k is
If UCAVs fly at the altitude-hold mode, which can be
easily achieved by a control law, i.e. X UAV i (k ) = [ pUAV
T
i (k ) v
T
UAV i
(k )]T
u3 = − K h (h − hdes ), (16) T
xUAV i (k ) yUAV i (k )hUAV i (k )vUAV
x
i
(20)
where K h is the proportional gain and hdes is a desired flight = .
(k )vUAV i (k )vUAV i (k )
y h
height), possible collisions can be eliminated by altering the
flight height.
In convenience, UCAVs fly in the altitude-hold mode,
2.4 Objective Function i.e. hUAV i (k ) ≡ hUAV i and vUAV
h
i (k ) ≡ 0 . Based on particle
The objectives of trajectory planning of UCAV are: kinematics of UCAV, the potential ranges of heading angle and
1) to approach the center of LAR as much as possible; velocity after one time step are
2) to avoid hard threats in the battlefield;
3) and to maintain a safe distance from those soft threats. VarψUAV = ψUAV i (k ) + ωmin ⋅ Dt , ψUAV i (k ) + ωmax ⋅ Dt ,
i
377
Def. SCI. J., Vol. 70, No. 4, July 2020
0 otherwise, VarψUAV
3.3 Procedures of Trajectory Planning Strategy
(24) Based on both threat models and mission demands, the
and situation of UCAVs falls into either of the following four
pTMIN UAV i (k ) = xUAV i (k ), yUAV i (k ), hUAV i
T
cases.
Case 1: With the current velocity and heading angle, it
π
(
sin ψUAV i (k ) + 2 − sin ψUAV i (k )
)
is highly likely for the UCAV to encounter a hard threat. The
vehicle should decelerate to the highest degree to turn as fast
vUAV i (k ) π as possible.
+
ωmin 2
(
− cos ψUAV i (k ) + + cos ψUAV i (k ) ). Case 2: With the current velocity and heading angle,
0 the UCAV is possible to encounter a highly risky area. The
vehicle should decelerate to the highest degree to turn as fast
as possible.
(25) Case 3: Though there is neither hard nor soft threat posed
Namely, they are extreme points of UAV i that can be to the UCAV, the distance between the vehicle and the center
reached at its current speed with the maximum and minimum of LAR is not as close as required.
turning rate. Case 4: Not only there is neither hard nor soft threat posed
Based on these variables, the possible situations of UAV i to the UCAV, the distance between the vehicle and the center of
after a time step can be predicted. Once the UCAV is predicted LAR is also close enough. In this case, the vehicle should make
of possible encounter with threats, it needs to adjust its control self-adjustment to satisfy the position and posture conditions
variables immediately. for airborne bombs to be released.
To distinguish between the aforementioned cases, two
3.2 Threat Avoidance Strategy control variables are introduced, as shown in the following
Herein, the strategies applied to avoid hard and soft threats equations.
are introduced.
(k ) = VarψUAV
− H m (k ) ∩ Varψ (k ),
i i i
UAV UAV
VarCA (29)
3.2.1 Hard Threat Avoidance Strategy
(k ) = VarCAUAV (k ) ∩ VarψUAV
− S (k ).
i i i
UAV
VarRH (30)
It is supposed that a hard threat
H m = Am , Pm (m = 1, , M 1 ) satisfies Fm (pUAVTMIN
(k ))≥ 0 i
Besides the rules of the cases are summarised in Table 1.
Figure 2 provides the details of procedures of the proposed
or Fm (pUAV (k ))≥ 0 . This indicates a potential hard threat
TMAX
i
strategy. The computational time complexity of the proposed
encounter of UAV i . To avoid this, the acceptable variation algorithm depends on the loop of threat cases classification.
range of the heading angle ought to be Suppose that for a UCAV, there are M waypoints and the
Hm π Hm π total number of hard and soft threats is N . In the worst-
− H m (k ) = ψUAV i ( k ) −
i
VarψUAV , ψUAV i (k ) + , (26) case scenario, the complexity of the loop corresponds to the
2 2
Hm
multiplication of M by the sum of all threats, i.e. O (MN ) . The
where ψUAV i ( k ) is one of the perpendicular directions to the
other component of the algorithm is O (M ) . Consequently, the
tangential of the ellipse H m at the point where this ellipse computational time complexity is O (MN ) for the proposed
and the line through pUAV i (h ) and Am (h ) intersect, which is algorithm.
closer to the direction of vector Am (h ) pUAV i (h ) . The approach
Hm 4. Simulation results and analysis
of ψUAV i ( k ) is detailed .
15
The proposed planning strategy is validated using Matlab
Otherwise, if there is no indications that UAV i is possible 2016b. The main hardware parameters of the computer are 4x
− H (k ) = [0, 2π ) .
i
to encounter H m , VarψUAV Intel(R) Core(TM) i7 CPU @ 3.07GHz.
378
Tang, et al.: A Case-based Online Trajectory Planning Method of Autonomous UCAVs with Weapon Release
Control variables Serial number of cases Threats Position of center Radius Height h
number point (km) (km) (km)
UAV
VarCA (k ) = ∅
i
1
1 (13,17) 1 2
2
VarCA (k ) ≠ ∅, 2 (5,9) 1.5 1.5
i
UAV
UAV i
VarRH (k ) = ∅ 3 (13,4) 1 1.7
4 (10,10) 1 1
3
VarCA (k ) ≠ ∅,
i
UAV
5 (6,16) 1.5 2
UAV i
VarRH (k ) ≠ ∅,
UAV i
DisqTi (k ) > DisRH Table 3. Motion constraints of the UCAV
UAV i Velocity [Vmin , Vmax ]/(km/s) [0.082, 0.59]
VarRH (k ) ≠ ∅, Acceleration/ (km/s2) [−0.005, 0.005]
UAV i
DisqTi (k ) ≤ DisRH Turning rate/(°/s) [−20, 20]
hmax /km 12
379
Def. SCI. J., Vol. 70, No. 4, July 2020
(a) (a)
(b) (b)
380
Tang, et al.: A Case-based Online Trajectory Planning Method of Autonomous UCAVs with Weapon Release
5. Conclusions
This paper concentrates on obtaining on-board trajectories
for UCAVs executing ground attack missions. Allowing for
the complex battlefields with obstacles, no-fly zones and air
defense threats, the optimal control problem of trajectory
planning has been transformed into the situation classification
of four cases.
The main contributions of the papers are summarised as
(c) follows. Firstly, the trajectory planning problem of UCAVs
Figure 6. Variation of control variables and relative distance with airborne guided bombs is mathematically formulated as
in Scenario 2: (a) Variation of velocity, (b) Variation a traditional optimal control problem (OCP). In order to meet
of heading angle, and (c) Variation of relative the release conditions of guided bombs, the LAR is integrated
distance. into the planning constraints. Secondly, a case-based planning
strategy in combination with the receding horizon method is
which triggers deceleration and alter to turn its heading for the proposed to achieve real-time trajectories for UCAVs to deliver
avoidance of it. At the time of 99 s, the UCAV averts the pop- airborne bombs.
up threat with success and carries on flying towards the center To conduct a further research, the proposed online
of LAR. trajectory planning strategy is recommended to be combined
with some representative computational intelligence
4.4 Computational Burden algorithms, such as monarch butterfly optimisation (MBO)
In order to see whether the changes in the time horizon ( algorithm17, earthworm optimisation algorithm18 (EWA) and
Dt herein) for guidance have an impact on the final trajectory, elephant herding optimisation (EHO) algorithm19, which
the runtime, mission completion time, final relative distance is conductive to solving the global optimisation problems
of the algorithm are compared with different time steps. In effectively, while ensuring the missions with large-scale
this scenario, the UCAV initially locates at 17 km, 1 km, 2 km battlefields can be accommodated.
381
Def. SCI. J., Vol. 70, No. 4, July 2020
References 14. Wilson, S.A.; Vultetich, I.J.; Fletcher, D.; Jokic, M.; Brett,
1. Liu, J.-J.; Wang, W.-P.; Li, X.-B.; Wang, T. & Wang, T.-Q. M.; Boyd, C.; Williams, W. & Bryce, I. Guided weapon
A motif-based mission planning method for UAV swarms danger area & safety template generation-a new capability.
considering dynamic reconfiguration. Def. Sci. J., 2018, In AIAA Atmospheric Flight Mechanics Conference and
68(2), 159-166. Exhibit, 2008.
doi: 10.14429/dsj.68.11959 doi: 10.2514/6.2008-7123
2. Goerzen, C.; Kong, Z. & Mettler, B. A survey of motion 15. Jiang, H. & Liang, Y.-Q. Online path planning of
planning algorithms from the perspective of autonomous autonomous UAVs for bearing-only standoff multi-target
UAV guidance. J. Intell. Robot. Syst., 2010, 57, 65–100. following in threat environment. IEEE Access, 2018, 6,
doi: 10.1007/s10846-009-9383-1 22531-22544.
3. Wang, G.-G.; Guo, L.-H.; Duan, H.; Wang, H.-Q. Luo, L. doi: 10.1109/ACCESS.2018.2824849
& Shao M.-Z. A hybrid meta-heuristic DE/CS algorithm 16. Storm Shadow UCAV performance. Available from:
for UCAV three-dimension path planning. Sci. World J., http://www.aerospaceweb.org/design/ucav/main.html.
2012, 1–11. 17. Wang, G.-G.; Deb, S. & Cui, Z.-H. Monarch butterfly
doi: 10.1100/2012/583973 optimization. Neural Comput. Appl., 2019, 31(7), 1995-
4. Wang, G.-G. Moth search algorithm: a bio-inspired 2014.
metaheuristic algorithm for global optimisation problems. doi: 10.1007/s00521-015-1923-y
Memet. Comp. 2016. 18. Wang, G.-G.; Deb, S. & Coelho, L.S. Earthworm
doi: 10.1007/s12293-016-0212-3 optimization algorithm: a bio-inspired metaheuristic
5. Xu, C.-F.; Duan, H.-B. & Liu, F. Chaotic artificial bee algorithm for global optimization problems. Int. J. Bio-
colony approach to uninhabited combat air vehicle Inspired Comput., 2015, 12, 1-22.
(UCAV) path planning. Aerosp Sci. Technol., 2010, 14, doi: 10.1504/IJBIC. 2015.10004283
535–541. doi:10.1016/j.ast.2010.04.008 19. Wang, G.-G.; Deb, S.; Gao, X.-Z. & Coelho, L.S. A
6. Wang, G.-G.; Guo, L.-H.; Duan, H.; Liu, L.; Wang, H.-Q. new metaheuristic optimization algorithm motivated by
& Shao M.-Z. Path planning for uninhabited combat aerial elephant herding behavior. Int. J. Bio-Inspired Comput.,
vehicle using hybrid meta-heuristic DE/BBO algorithm. 2016, 8(6), 394.
Adv. Sci., Eng. Med., 2012, 4(6), 550–564. doi: 10.1504/IJBIC.2016.10002274
doi: 10.1166/asem.2012.1223
7. Duan, H.-B.; Liu, S.-Q. & Wu, J. Novel intelligent water Contributors
drops optimization approach to single UCAV smooth
trajectory planning. Aerosp Sci. Technol., 2009, 13(8), Ms Jiayu Tang is currently pursuing her PhD in Naval
442–449. Aviation University, Yantai, China. Her research interests
include: cooperative task assignment and path planning of
doi: 10.1016/j.ast.2009.07.002
multi-UAV.
8. Wang, G.-G.; Chu, H.-C. & Mirjalili, S. Three-dimensional In the current study, she has done simulation, post processing
path planning for UCAV using an improved bat algorithm. of the results and preparation of the manuscript.
Aerosp Sci. Technol., 2016, 49, 231-238.
doi: 10.1016/j.ast.2015.11.040 Prof. Xiangmin Li obtained his PhD from Northwestern
9. Betts, J.T. Survey of numerical methods for trajectory Polytechnical University. He is currently a professor and a
optimization. J. Guid Control Dyn., 1998, 21(2), 193-207. doctoral supervisor in Naval Aviation University, China. His
doi: 10.2514/2.4231 research interests include: Fire command and control.
10. Zhang, Y.; Chen, J. & Shen, L.-C. Real-time trajectory In the current study, he has provided guidance in examining
the results and reviewed the final manuscript .
planning for UCAV air-to-surface attack using inverse
dynamics optimization method and receding horizon Dr Jinjin Dai received his PhD (weapon systems and application
control. Chi. J. Aeronaut., 2013, 26(4), 1038–1056. engineering) in 2012 from Naval Aviation University. He is
doi: 10.1016/j.cja.2013.04.040 currently a lecturer in Naval Aviation University, China. His
11. Zhu, W.-R. & Duan, H.-B. Chaotic predator-prey research interests include: Fire command and control.
biogeography-based optimization approach for UCAV In the current study, he has provided guidance in executing
path planning. Aerosp Sci. Technol., 2014, 32, 153-161. the battlefield modelling and the simulation.
doi: 10.1016/j.ast.2013.11.003
12. Duan, H.-B. & Huang, L.-Z. Imperialist competitive Dr Ning Bo received his PhD (weapon systems and application
engineering) in 2018 from Naval Aviation University. He is
algorithm optimized artificial neural networks for UCAV
currently an engineer in Unit 91213 of PLA, Yantai, China.
global path planning. Neurocomputing, 2014, 125, 166-171. His research interests include: Weapon system simulation.
doi: 10.1016/j.neucom.2012.09.039 In the current study, he has provided guidance in the solution
13. Curiac, D.I. & Volosencu, C. Path planning algorithm of LAR.
based on Arnold cat map for surveillance UAVs. Def. Sci.
J., 2015, 65(6), 483-488.
doi: 10.14429/dsj.65.8483
382