Missile trajectory shaping using sampling-based path planning P. Pharpatara, B. Hérissé, R. Pepy, Y. Bestaoui 04/11/2013 - Tokyo, Japan

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Missile interception −→ Target detection by the radar station. I Mission preparation (À) : trajectories calculations. −→ Missile launch. I Midcourse guidance (Á) : I

I

ensure a proper collision course to the target with an acceptable velocity ; trajectory optimization with constraints.

−→ Target detection by the embedded radar. I Terminal guidance (Â) : I

engage the target at the Predicted Intercept Point (PIP).

PIP target Á À

2/18

IEEE IROS 2013 - Tokyo, Japan

 radar beam

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Objective To generate a missile-interceptor trajectory for the midcourse guidance (off-line or/and on-line) : I Optimal control problem : I

I

Terminal constraints at the PIP : I

I

I

maximize the final velocity (hit-to-kill). lower bound of terminal speed at the PIP (to ensure the target destruction) ; related orientation between the missile and the target at the PIP (due to the clearance of the embedded sensor).

Constraints related to the system/missile (here, single-stage missile) : I I I

structural limitation ; maximum angle of attack that can be balanced using the tail fins ; variation of air density.

=⇒ Complex, non-holonomic and non-linear problem 3/18

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Closed-loop midcourse guidance laws I

Singular Perturbation Theory [4] ;

I

Linear Quadratic Regulators [6] ;

I

Fuzzy Logic [9] ;

I

Neural Network [13] ; Kappa guidance [8] :

I

I

I

take the final orientation into account ; maximize the final velocity (locally optimal).

-with control limitation -no control limitation

Trajectories given by Kappa guidance

Problem of analytical guidance laws (ex : kappa guidance) I

They cannot anticipate the future changes of flight conditions : I

4/18

ex : loss of manoeuvrability at high altitude.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Trajectory generation using numerical methods [2]

I

State Dependant Riccati Equation (SDRE) technique [11] ;

I

Adaptive grid methods [1] ;

I

Pseudo-spectral method [12][10].

Problems of numerical methods I

Initialisation problem : I

I

5/18

a good first guess is recommended to ensure the acquisition of a solution.

Trade-off between accuracy and efficiency in terms of computational effort.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Proposed solution

I

Combination of trajectory planning methods and optimal control methods, here : I

Rapidly-exploring Random Tree (RRT) [7] : I I

I

Dubins’ paths (Dubins’ metric) [5][3] : I

6/18

it takes the future changes of flight conditions into account ; it can find a solution without any a priori guess. it is used to improve the optimality of the solution trajectory in 2D plane.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Missile modelling in 2 Dimensions ξ˙ = v,

(1)

mv˙ = fdrag (ρ, ||v||) + flift (ρ, ||v||) + fthrust − mg, θ˙ = q, I q˙ = τaero + τpert . flift fdrag

fthrust θ

α

Cg

v

7/18

IEEE IROS 2013 - Tokyo, Japan

(3) (4)

I

I

fthrust lasts 20s after launch (single-stage missile) ; flift , fdrag depend on ρ and v : I

mg

(2)

I

ρ small : loss of manoeuvrability at high altitude ; v large : structural limitation.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Mission and requirements (1) Mission I

xinit = [0 0 0 0]> with vertical launch for 1s ;

I

Xgoal = {x : kξ − ξ pip k < R, kvk > vmin , ∠(v, −vpip ) < φf } : Xgoal φf vpip φf

ξ pip vmin

R

I

vpip is in the horizontal direction ;

I

vmin = 500m/s ;

I

R = 500m ;

I

φf = π/8.

Exploration space

8/18

I

X = {x : ξ ∈ [−20km, 20km] × [0km, 30km]} ;

I

Xfree = {x : alt(ξ) > 0, ||v(t > tboost )|| > vmin }.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Mission and requirements (2) Admissible control input U(t, x) = {avc : αc 6 αmax (t, x)} with



stb struct αmax (t, x) = min αmax (t, x), αmax (t, x)



where stb (t, x) : the maximum achievable angle of attack using tail fins ; αmax struct (t, x) : the structural limit which is given by the maximum αmax lateral acceleration abmax in body frame that the missile can suffer before it breaks.

9/18

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Rapidly-exploring Random Trees - RRT [7]

What is RRT ? I

An incremental method designed to efficiently explore non-convex high-dimensional spaces.

Why choosing RRT ?

10/18

I

It uses the complete model ;

I

It takes the future changes of environment into account ;

I

It can find a solution without any a priori guess.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

G

11/18

IEEE IROS 2013 - Tokyo, Japan

Conclusions & perspectives

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

uniform distribution xrand

G

11/18

IEEE IROS 2013 - Tokyo, Japan

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

xrand xnear (t)

Dubins’ metric

G

11/18

IEEE IROS 2013 - Tokyo, Japan

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

xrand xnear (t) Kappa Guidance

G

11/18

IEEE IROS 2013 - Tokyo, Japan

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

xrand xnew (t + ∆t) xnear (t)

Integration of complete model G with constraints, ∆t = 2s

11/18

IEEE IROS 2013 - Tokyo, Japan

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

xrand xnew (t + ∆t) xnear (t)

Check if xnew ∈ Xfree

11/18

IEEE IROS 2013 - Tokyo, Japan

G

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Rapidly-exploring Random Tree - RRT [7] RRT Algorithm

G add xnew in G

11/18

IEEE IROS 2013 - Tokyo, Japan

Conclusions & perspectives

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

NearestNeighbor What is Dubins’ metric (based on Dubins’ curve [5][3]) ? I

Analytical method used to define the shortest path between two states while considering their orientation ; xfinal xfinal

xinit

xfinal

(i) CCC type I

xinit

xinit

(ii) CSC types

Metric function : the length of the shortest Dubins’ path.

Why choosing Dubins’ metric ?

12/18

I

Its optimality in 2D is assured ;

I

It considers the orientations of the two states.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Results : Comparisons and Discussions (1) 1) No saturation problem 4

x 10

500

phase 1 (boost) phase 2

3

||av c (αmax )|| ||av kappa || ||av RRT||

450

400

2.5

300

m/s2

altitude

350

2 1.5

250

200

1

150

0.5 100

0

50

−2

−1

0 1 horizontal distance

2

0

4

0

0.1

x 10

Generated trajectories

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

t/tf

Lateral accelerations along the solution trajectories √ I < π/8 and < π/8 ; √ √ I ||vf || = 1586m/s > 500m/s and ||vf ; RRT kappa || = 1717m/s > 500m/s f ∠(vRRT , −vpip )

√

f ∠(vkappa , −vpip )

I RRT and kappa : respect the limit of control inputs. 13/18

IEEE IROS 2013 - Tokyo, Japan

1

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Ref.

Results : Comparisons and Discussions (2) 2) With saturation problem 4

500

x 10

phase 1 (boost) phase 2

3

||av c (αmax )|| ||av kappa || ||av RRT||

450

400

2.5

300

m/s2

altitude

350

2 1.5

250

200

1

150

0.5 100

0

50

−2

−1

0 1 horizontal distance

Generated trajectories

2 4

x 10

0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

t/tf

Lateral accelerations along the solution trajectories I ∠(vf , −vpip ) < π/8 and ∠(vf kappa , −vpip ) = 0.62 > π/8X ; RRT √ I ||vf || = 1309m/s > 500m/s ; RRT I RRT anticipates the loss of manoeuvrability at the end of the trajectory but kappa doesn’t. √

14/18

IEEE IROS 2013 - Tokyo, Japan

1

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Conclusions

Advantages I I

Satisfaction of constraints by integrating the complete model ; It anticipates future flight conditions : I

I

for example : for the 2nd scenario, a back-turn can be found.

A solution can be found without any a priori guess.

Drawbacks

15/18

I

Obtained solutions are not optimal ;

I

Number of iterations to find a solution can be large depending on how difficult the scenario is (ok for off-line trajectory generation).

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

Perspectives

I

Evolution of distance between RRT solution and the optimal one : I I

16/18

maximize the final velocity ; minimize the flight time.

I

Improve the computing time by reducing X ;

I

Improve the metric and the selection of control input ;

I

2D→3D.

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

References I [1]

D. A. Anisi, Adaptive node distribution for on-line trajectory planning, Congress of the International Council of the Aeronautical Sciences (ICAS) (2006).

[2]

J. T. Betts, Survey of munmerical methods for trajectory optimization, Journal of Guidance, Control, and Dynamics 21 (1998), no. 2, 193–207.

[3]

J. D. Boissonnat, A. Cérézo, and J. Leblond, Shortest paths of bounded curvature in the plane, Tech. report, Institut National de Recherche en Informatique et en Automatique, 1991.

[4]

V. H. L. Cheng and N. K. Gupta, Advanced midcourse guidance for air-to-air missiles, Journal of Guidance, Control, and Dynamics 9 (1986), no. 2, 135–142.

[5]

L. E. Dubins, On curves of minimal length with a constraint on average curvature and with presribed initial and terminal position and tangents, American Journal of Mathematics 79 (1957), 497–516.

[6]

F. Imado, T. Kuroda, and S. Miwa, Optimal midcourse guidance for medium-range air-to-air missiles, Journal of Guidance, Control, and Dynamics 13 (1990), no. 4, 603–608.

[7]

S. M. LaValle, Planning algorithms, Cambridge University Press, 2006.

[8]

C. F. Lin, Modern navigation guidance and control processing, Prentice-Hall, Inc., 1991.

[9]

C. L. Lin and Y. Y. Chen, Design of fuzzy logic guidance law against high-speed target, Journal of Guidance, Control, and Dynamics 23 (2000), no. 1, 17–25.

[10] J. A. Lukacs and O. A. Yakimenko, Trajectory-shape-varying missile guidance for interception of ballistic missiles during the boost phase, AIAA (2007). [11] P. K. Menon, T. Lam, L. S. Crawford, and V. H. Cheng, Real-time computational methods for sdre nonlinear control of missiles, American Control Conference (2002), 232–237. [12] J.-W. Park, M.-J. Tahk, and H.-G. Sung, Trajectory optimization for a supersonic air-breathing missile system using pseudo-spectral method, International Journal of Aeronautical and Space Sciences 10 (2009), 112–121. 17/18

IEEE IROS 2013 - Tokyo, Japan

Ref.

Introduction

Problem statement

Proposed solution : RRT+Dubins’ paths

Results

Conclusions & perspectives

References II

[13] E. J. Song and M. J. Tahk, Real-time neural-network midcourse guidance, Control Engineering Practice 9 (2001), no. 10, 1145 – 1154.

18/18

IEEE IROS 2013 - Tokyo, Japan

Ref.