Discrete mechanics, optimal control and formation flying spacecraft



Similar documents
Physics 9e/Cutnell. correlated to the. College Board AP Physics 1 Course Objectives

APPLIED MATHEMATICS ADVANCED LEVEL

11. Rotation Translational Motion: Rotational Motion:

Columbia University Department of Physics QUALIFYING EXAMINATION

Midterm Solutions. mvr = ω f (I wheel + I bullet ) = ω f 2 MR2 + mr 2 ) ω f = v R. 1 + M 2m

Lecture L30-3D Rigid Body Dynamics: Tops and Gyroscopes

Lecture 07: Work and Kinetic Energy. Physics 2210 Fall Semester 2014

Optimization of Supply Chain Networks

Center of Gravity. We touched on this briefly in chapter 7! x 2

3. Derive the partition function for the ideal monoatomic gas. Use Boltzmann statistics, and a quantum mechanical model for the gas.

3600 s 1 h. 24 h 1 day. 1 day

Lecture L29-3D Rigid Body Dynamics

KINEMATICS OF PARTICLES RELATIVE MOTION WITH RESPECT TO TRANSLATING AXES

PHY121 #8 Midterm I

Gravity Field and Dynamics of the Earth

Physics 1A Lecture 10C

Chapter 2. Mission Analysis. 2.1 Mission Geometry

State Newton's second law of motion for a particle, defining carefully each term used.

Numerical Methods for Differential Equations

Onboard electronics of UAVs

Torque Analyses of a Sliding Ladder

8.012 Physics I: Classical Mechanics Fall 2008

State Newton's second law of motion for a particle, defining carefully each term used.

Halliday, Resnick & Walker Chapter 13. Gravitation. Physics 1A PHYS1121 Professor Michael Burton

Dynamics. Basilio Bona. DAUIN-Politecnico di Torino. Basilio Bona (DAUIN-Politecnico di Torino) Dynamics / 30

Computer Graphics. Geometric Modeling. Page 1. Copyright Gotsman, Elber, Barequet, Karni, Sheffer Computer Science - Technion. An Example.

An Introduction to Applied Mathematics: An Iterative Process

Orbits of the Lennard-Jones Potential

Let s first see how precession works in quantitative detail. The system is illustrated below: ...

Penn State University Physics 211 ORBITAL MECHANICS 1

Physics 2A, Sec B00: Mechanics -- Winter 2011 Instructor: B. Grinstein Final Exam

Parameter Estimation for Bingham Models

The Technical Archer. Austin Wargo

Rigid body dynamics using Euler s equations, Runge-Kutta and quaternions.

Fluid Mechanics Prof. S. K. Som Department of Mechanical Engineering Indian Institute of Technology, Kharagpur

AP Physics: Rotational Dynamics 2

DYNAMICS OF A TETRAHEDRAL CONSTELLATION OF SATELLITES-GYROSTATS

OpenFOAM Optimization Tools

Halliday, Resnick & Walker Chapter 13. Gravitation. Physics 1A PHYS1121 Professor Michael Burton

Numerical methods for American options

N 1. (q k+1 q k ) 2 + α 3. k=0

AN INTRODUCTION TO NUMERICAL METHODS AND ANALYSIS

Kyu-Jung Kim Mechanical Engineering Department, California State Polytechnic University, Pomona, U.S.A.

Precise Modelling of a Gantry Crane System Including Friction, 3D Angular Swing and Hoisting Cable Flexibility

Dynamics and Control of an Elastic Dumbbell Spacecraft in a Central Gravitational Field

Presentation of problem T1 (9 points): The Maribo Meteorite

DIRECT ORBITAL DYNAMICS: USING INDEPENDENT ORBITAL TERMS TO TREAT BODIES AS ORBITING EACH OTHER DIRECTLY WHILE IN MOTION

Chapter 10 Rotational Motion. Copyright 2009 Pearson Education, Inc.

Lab 7: Rotational Motion

Nonlinear Iterative Partial Least Squares Method

Orbital Mechanics. Angular Momentum

Influence of Crash Box on Automotive Crashworthiness

Gravitomagnetism and complex orbit dynamics of spinning compact objects around a massive black hole

Seminar 4: CHARGED PARTICLE IN ELECTROMAGNETIC FIELD. q j

(Most of the material presented in this chapter is taken from Thornton and Marion, Chap. 7)

Chapter 21 Rigid Body Dynamics: Rotation and Translation about a Fixed Axis

Lecture L22-2D Rigid Body Dynamics: Work and Energy

Free Fall: Observing and Analyzing the Free Fall Motion of a Bouncing Ping-Pong Ball and Calculating the Free Fall Acceleration (Teacher s Guide)

Lecture 2 Linear functions and examples

SOLID MECHANICS TUTORIAL MECHANISMS KINEMATICS - VELOCITY AND ACCELERATION DIAGRAMS

Lecture 13. Gravity in the Solar System

Worldwide, space agencies are increasingly exploiting multi-body dynamical structures for their most

Unit - 6 Vibrations of Two Degree of Freedom Systems

Sample Questions for the AP Physics 1 Exam

Lecture Notes to Accompany. Scientific Computing An Introductory Survey. by Michael T. Heath. Chapter 10

Analysis of Multi-Spacecraft Magnetic Field Data

G U I D E T O A P P L I E D O R B I T A L M E C H A N I C S F O R K E R B A L S P A C E P R O G R A M

Dynamics of Iain M. Banks Orbitals. Richard Kennaway. 12 October 2005

Displacement (x) Velocity (v) Acceleration (a) x = f(t) differentiate v = dx Acceleration Velocity (v) Displacement x

Mechanics lecture 7 Moment of a force, torque, equilibrium of a body

Numerical Methods For Image Restoration

Motion Control of 3 Degree-of-Freedom Direct-Drive Robot. Rutchanee Gullayanon

Equivalent Spring Stiffness

Module 1 : Conduction. Lecture 5 : 1D conduction example problems. 2D conduction

FUNDAMENTAL FINITE ELEMENT ANALYSIS AND APPLICATIONS

Macroeconomic Effects of Financial Shocks Online Appendix

Lecture 17. Last time we saw that the rotational analog of Newton s 2nd Law is

PHYSICS 111 HOMEWORK SOLUTION #9. April 5, 2013

Use the following information to deduce that the gravitational field strength at the surface of the Earth is approximately 10 N kg 1.

Finite Difference Approach to Option Pricing

EXPERIMENT: MOMENT OF INERTIA

MODULE VII LARGE BODY WAVE DIFFRACTION

CS Software Engineering for Scientific Computing. Lecture 16: Particle Methods; Homework #4

Solar System Fundamentals. What is a Planet? Planetary orbits Planetary temperatures Planetary Atmospheres Origin of the Solar System

State of Stress at Point

Lecture L6 - Intrinsic Coordinates

PS 320 Classical Mechanics Embry-Riddle University Spring 2010

A Bond Graph Approach for Modelling Systems of Rigid Bodies in Spatial Motion

AP Physics C. Oscillations/SHM Review Packet

Section 4: The Basics of Satellite Orbits

Simple Harmonic Motion

ACTUATOR DESIGN FOR ARC WELDING ROBOT

How To Understand The Dynamics Of A Multibody System

Derive 5: The Easiest... Just Got Better!

Isaac Newton s ( ) Laws of Motion

Transcription:

Discrete mechanics, optimal control and formation flying spacecraft Oliver Junge Center for Mathematics Munich University of Technology joint work with Jerrold E. Marsden and Sina Ober-Blöbaum partially supported by the CRC 376 Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.1

Outline mechanical optimal control problem direct discretization of the variational principle ( DMOC ) applications: low-thrust orbital transfer, hovercraft, spacecraft formation flying Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.2

Introduction Optimal control problem mechanical system, configuration space Q, to be moved from (q 0, q 0 ) to (q 1, q 1 ), force f, s.t. J(q, f ) = 1 0 C(q(t), q(t), f (t)) dt min Dynamics: Lagrange-d Alembert principle δ 1 L(q(t), q(t)) dt + 1 0 0 f (t) δq(t) dt = 0 for all δq with δq(0) = δq(1) = 0, Lagrangian L : TQ R. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.3

Introduction Equality constrained optimization problem Minimize subject to (q, f ) J(q, f ) L(q, f ) = 0. Standard approach: derive differential equations, discretize (multiple shooting, collocation), solve the resulting (nonlinear) optimization problem. Here: discretize the Lagrange-d Alembert principle directly. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.4

Discretization techniques: comparison cost function + EL cost function + Ld Ap variation discretization discrete cost function + discretized ode Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.5

Discretization techniques: comparison cost function + Ld Ap cost function + EL variation discretization discrete cost function + discretized ode discretization discrete cost function + discrete Ld Ap variation discrete cost function + DEL Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.6

Discretization techniques: comparison cost function + Ld Ap cost function + EL variation discretization discrete cost function + discretized ode finite differences, multiple shooting, collocation,... discretization discrete cost function + discrete Ld Ap variation discrete cost function + DEL Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.7

Discretization techniques: comparison cost function + Ld Ap cost function + EL variation discretization discrete cost function + discretized ode finite differences, multiple shooting, collocation,... discretization discrete cost function + discrete Ld Ap variation discrete cost function + DEL DMOC Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.8

Discrete paths Figure: J.E. Marsden, Lectures on Mechanics Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.9

Discrete paths Figure: J.E. Marsden, Lectures on Mechanics Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.10

Discretization of the variational principle Replace the state space TQ by Q Q, a path q : [0, 1] Q by a discrete path q d : {0, h,..., Nh = 1} Q, the force f : [0, 1] T Q by a discrete force f d : {0, h, 2h,..., Nh = 1} T Q. Discrete Lagrangian L d : Q Q R, virtual work L d (q k, q k+1 ) (k+1)h kh f k δq k + f + k δq k+1 L(q(t), q(t)) dt, (k+1)h f k, f + k T Q: left and right discrete forces. kh f (t) δq(t) dt, Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.11

Discretizing the variational principle Discrete Lagrange-d Alembert principle Find discrete path {q 0, q 1,..., q N } s.t. for all variations {δq 0,..., δq N } with δq 0 = δq N = 0 N 1 L d (q k, q k+1 ) + N 1 δ k=0 k=0 f k δq k + f + k δq k+1 = 0. Forced discrete Euler-Lagrange equations Discrete principle quivalent to D 2 L d (q k 1, q k ) + D 1 L d (q k, q k+1 ) + f + k 1 + f k = 0. k = 1,..., N 1. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.12

Boundary Conditions We need to incorporate the boundary conditions q(0) = q 0, q(0) = q 0 q(1) = q 1, q(1) = q 1 into the discrete description. Legendre transform FL : TQ T Q FL : (q, q) (q, p) = (q, D 2 L(q, q)), Discrete Legendre transform for forced systems F f + L d : (q k 1, q k ) (q k, p k ), p k = D 2 L d (q k 1, q k ) + f + k 1. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.13

The Discrete Constrained Optimization Problem Minimize N 1 J d (q d, f d ) = C d (q k, q k+1, f k, f k+1 ), k=0 subject to q 0 = q 0, q N = q 1 and D 2 L(q 0, q 0 ) + D 1 L d (q 0, q 1 ) + f 0 = 0, D 2 L d (q k 1, q k ) + D 1 L d (q k, q k+1 ) + f + k 1 + f k = 0, D 2 L(q N, q N ) + D 2 L d (q N 1, q N ) + f + N 1 = 0, k = 1,..., N 1. Solution by, e.g., sequential quadratic programming. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.14

Implementation: quadrature q(t) q k+1 q(t) q( t k+t k+1 ) = q k+q k+1 2 2 q k Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.15

Implementation: quadrature Discrete Lagrangian tk+1 tk+1 t k t k tk+1 ( ( tk + t k+1 L q t k 2 ( qk + q k+1 = hl 2 L(q, q) dt L( q(t), q(t)) dt =: L d (q k, q k+1 ) ), q, q k+1 q k 2 ( )) tk + t k+1 ) 2 dt Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.16

Implementation: quadrature Discrete forces (k+1)h kh f δq dt h f k+1 + f k 2 δq k+1 + δq k 2 = h 4 (f k+1 + f k ) δq k + h } {{ } 4 (f k+1 + f k ) } {{ } =:f =:f + k k δq k+1 Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.17

Example: Low thrust orbital transfer Satellite with mass m, to be transferred from one circular orbit to one in the same plane with a larger radius. Number of revolutions around the Earth is fixed. In 2d-polar coordinates q = (r, ϕ) L(q, q) = 1 2 m(ṙ 2 + r 2 ϕ 2 ) + γ Mm, r M: mass of the earth. Force u in the direction of motion of the satellite. Goal minimize the control effort J(q, u) = T 0 u(t) 2 dt. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.18

Comparison to traditional scheme 0.24 0.22 1 rotation Euler Variational 0.2 0.18 objective function value 0.16 0.14 0.12 0.1 0.08 0.06 0.04 10 15 20 25 30 35 40 45 50 number of nodes Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.19

Comparison to traditional scheme 0.18 0.16 1 rotation Midpointrule Variational 0.14 objective function value 0.12 0.1 0.08 0.06 0.04 5 10 15 20 25 30 35 40 45 50 number of nodes Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.20

Comparison to the true solution 8 7 Euler Variational 6 deviation of final state 5 4 3 2 1 0 5 10 15 20 25 30 35 40 45 50 number of nodes Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.21

Comparison to the true solution 0.4 0.35 Midpointrule Variational 0.3 deviation of final state 0.25 0.2 0.15 0.1 0.05 0 5 10 15 20 25 30 35 40 45 50 number of nodes Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.22

Example: Reconfiguration of a group of hovercraft Hovercraft y θ f 2 f 1 r Configuration manifold: Q = R 2 S 1 Underactuated system, but configuration controllable. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.23 x

Reconfiguration of a group of Hovercraft Lagrangian L(q, q) = 1 2 (mẋ 2 + mẏ 2 + J θ 2 ), q = (x, y, θ), m the mass of the hovercraft, J moment of inertia. Forced discrete Euler-Lagrange equations ( ) 1 M ( q h k 1 + 2q k q k+1 ) + h fk 1 +f k + f k+f k+1 = 0, (1) 2 2 2 k = 1,..., N. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.24

Goal Minimize the control effort while attaining a desired final formation: (a) a fixed final orientation ϕ i of each hovercraft, (b) equal distances r between the final positions, (c) the center M = (M x, M y ) of the formation is prescribed, (d) fixed final velocities. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.25

Results Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.26

Application: Spacecraft Formation Flying ESA: Darwin NASA: Terrestrial Planet Finder SFB 376 Massive Parallelism, Project C10: Efficient Control of Formation Flying Spacecraft Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.27

The Model A group of n identical spacecraft, single spacecraft: rigid body with six degrees of freedom (position and orientation), control via force-torque pair (F, τ) acting on its center of mass. dynamics (as required for Darwin/TPF): circular restricted three body problem Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.28

The Lagrangian Potential energy: 1 µ V (x) = x (1 µ, 0, 0) µ x ( µ, 0, 0), µ = m 1 /(m 1 + m 2 ) normalized mass. kinetic energy: K trans (x, ẋ) = 1 2 (( x 1 ωx 2 ) 2 + ( x 2 + ωx 1 ) 2 + ẋ3 2 ) + K rot (Ω) = 1 2 ΩT JΩ, Ω: angular velocity, J: inertia tensor. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.29

The Control Problem Goal: compute control laws (F (i) (t), τ (i) (t)), i = 1,..., n, such that within a prescribed time interval, the group moves from a given initial state into a prescribed target manifold, minimizing a given cost functional (related to the fuel consumption). Target manifold: 1. all spacecraft are located in a plane with prescribed normal, 2. the spacecraft are located at the vertices of a regular polygon with a prescribed center on a Halo orbit, 3. each spacecraft is rotated according to a prescribed rotation, 4. all spacecraft have the same prescribed linear velocity and zero angular velocity. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.30

Collision avoidance Artificial potential 4 3.5 3 2.5 2 1.5 1 0 1 2 3 4 5 L = K trans + K rot V n V a ( q (i) q (j) ). i,j=1 i j Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.31

Halo orbits x 10 3 4 2 0 L 2 z 2 4 E 6 8 10 12 0.01 y 0 0.01 0.999 1.0016 1.0042 x 1.0068 1.0094 1.012 Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.32

Result Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.33

Hierarchical decomposition same model for every vehicle n identical subsystems, coupling through constraint on final state Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.34

Hierarchical optimal control problem min ϕ 1,...,ϕ n n J i (ϕ i ) s.t. g(ϕ) = 0, i=1 (g describes final state) with J(ϕ i ) = optimal solution of 2-point bvp with ϕ i parametrizing the final state Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.35

Parallelization solve inner problems in parallel synchronization (communication) for iteration step in solving the outer problem implementation: PUB ( Paderborn University BSP-Library ) library to support development of parallel algorithms based on the Bulk-Synchronous-Parallel-Model. Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.36

Result Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.37

Conclusion new approach to the discretization of mechanical optimal control problems direct discretization of the underlying variational principle faithful energy behaviour by construction Outlook convergence backward error analysis hierarchical decomposition in time: discontinuous ( weak ) solutions, Pontryagin-d Alembert principle generalization to spatially distributed systems (PDEs) Oliver Junge Discrete mechanics, optimal control and formation flying spacecraft p.38