320 likes | 442 Views
Nonlinear High-Fidelity Solutions for Relative Orbital Motion. 20 May 2014. Dr. T . Alan Lovell AFRL/RVSV Space Vehicles Directorate. Outline. Goals, Motivation, Previous/Recent Work FY14 Effort Projected FY15 Effort. Goals.
E N D
Nonlinear High-Fidelity Solutions for Relative Orbital Motion 20 May 2014 Dr. T. Alan Lovell AFRL/RVSV Space Vehicles Directorate
Outline • Goals, Motivation, Previous/Recent Work • FY14 Effort • Projected FY15 Effort
Goals • To derive explicit (closed-form) analytical expressions for translational relative orbital motion that capture • Nonlinearities • Perturbation forces • To apply these solutions to space applications of interest to USAF, DoD, & the space community in general • Relative navigation/orbit determination • Orbital targeting/maneuver planning
Relative S/C Dynamics Radial (x) Chief r 0 rC r LVLH frame Deputy Along-track (y) r rD rC • Tschauner-Hempel (T-H) Model • Linear Time-Varying ODE’s • r << r • Any eccentricity Chief • e between chief/deputy small • Hill-Clohessy-Wiltshire (HCW) Model • Linear Time-Invariant ODE’s • r << r • Chief near-circular (e << 1) • e between chief/deputy small Inertial frame • Full 2-Body Dynamics • Nonlinear, Time-Invariant • Any chief/deputy eccentricities • Analytically intractable
Motivation • Relative motion has been subject of many decades of research • Most relative motion models can be categorized based on assumptions inherent in themodel: • What natural forces on each object are to be accounted for? • Are the objects in close proximity, i.e., on closely neighboring orbits? (linear vs nonlinear) • Is the chief on a circular or near-circular orbit? (time-varying vs time-invariant)
Motivation • Summary/categorization of several existing relative motion models • No widely accepted closed-form Cartesian (x, y, z) solution that contains nonlinear terms AND accounts for forces other than two-bodygravity
One Approach • One nonlinear modeling technique involves use of multi-dimensional convolution integrals (e.g. Volterra series)
Recent Solution • Newman (2012) applied Volterraapproach to nonlinear ODEs of 2-body relative orbital motion (assuming circular chief orbit): • Set up problem to retain terms up to 2nd order—quadratic (e.g. x2) and bilinear (e.g. xy) • Resulting solution consists of expressions for x(t), y(t), and z(t) and their rates • Each expression can contain up to 27 terms • For linear solution (e.g. HCW), each expression can contain up to only 6 terms
Recent Solution • Radial (x) component of solution is as follows (15 terms): • y, z expressions are similar (16 & 8 terms, respectively)
Recent Solution • Volterra relative motion solution shows much better agreement w/ 2-body truth compared to linear (HCW) solution • Reveals characteristics not evident in HCW: • coupling between the cross-track (z) motion and the radial/in-track (x-y) motion • secular drift terms in radial direction (HCW indicates in-track driftonly)
FY14 Effort • TASK 1: Apply Volterra solution • Relative navigation/orbit determination • TASK 2: Derive nonlinear closed-form solution by different approach curvilinear coordinates • Compared to Volterra solution
TASK 1: Apply Volterra Solution • An excellent application of the Volterrasolution is the problem of angles-only relative navigation (or relative orbit determination) • Navigation answers the question “WHERE AM I?” • Relative navigation answers the question “WHERE AM I RELATIVE TO THAT OTHER OBJECT?” (or “WHERE IS IT RELATIVE TO ME?”) • Typically applies to close-proximity scenarios, but can be applied over a LONGER baseline as well
TASK 1: Apply Volterra Solution • This is fundamentally an estimation problem • Critical issue in our scenario is observability • Whether (& how accurately) the states (x, y, z, x-dot, y-dot, z-dot) can be determined from the measurements (LOS) • Unobservabilityimplies ambiguity, i.e. we can’t uniquely determine the relative orbit given our set of measurements • For our scenario, it turns out unobservabilityisguaranteedunder the following conditions: • Linear dynamics assumed • Cartesian coordinate frame • Angles only • No maneuvers • What does this mean? • Two (relative) state vectors that differ by a constant multiple possess identical LOS • When propagated forward with LINEAR, CARTESIAN dynamics, they produce identical LOS histories • We can’t determine is the size of thetrajectory RANGE AMBIGUITY
TASK 1: Apply Volterra Solution • To alleviate range ambiguity, need to relax at least one of the restricting conditions one of which was assumption of linear dynamics • Thus, incorporating nonlinear effects in the estimation model should induce observability • Two initial states propagated forward with nonlinear dynamics • LOS histories deviate more with increase in downrange (along-track) separation • (If propagated with linear dynamics, LOS histories would be identical) t1 x t2 x t3 x t1 t2 o o t3 o t4 o t4 x Nonlinear dynamics won’t entirely alleviate ambiguity problem observability may be WEAK
TASK 1: Apply Volterra Solution • Volterrasolution consists of polynomial expressions in x0, y0, etc • Applying it to “initial relative orbit determination” (IROD) yields SIX 2nd-order polynomial eqn’s in SIX states • Solved via Macaulay resultant matrix (1916)
TASK 1: Apply Volterra Solution • Volterra-based ROD solution provides initial guess for precise/refined solution
TASK 2: Curvilinear Solution • Consider characterizing relative motion in curvilinear coordinates (cylindrical dr, dq, dzor spherical dr, dq, df) • Essentially inertial state differences; no concept of relative frame • Even linearized dynamics accurately capture curvature of drifting motion
TASK 2: Curvilinear Solution • Begin with linearized relative ODEs in spherical coordinates: • These eqn’s possess same form as the linearized Cartesian (“HCW”) equations and are easily solved:
TASK 2: Curvilinear Solution • We can then use the NONLINEAR kinematic relationships between Cartesian & spherical coordinates (& vice versa)…
TASK 2: Curvilinear Solution • …to express the curvilinear solution as a Cartesian solution via “double transformation”:
TASK 2: Curvilinear Solution • Result is a closed-form, nonlinear Cartesian solution for relative motion expressed as a function of the initial states and time: • Possesses many similar properties to Volterra solution… • coupling between cross-track (z) and radial/in-track (x-y) motion • radial secular drift • …but are NOT polynomial expressions in x0, y0, etc
TASK 2: Curvilinear Solution • Generate polynomial expressions from our new nonlinear solution via Taylor series expansion: • Rewriting using basic trig terms yields same form as previously shown for Volterra solution • This allows direct term-by-term comparison of the two solutions
TASK 2: Curvilinear Solution • Comparison of Volterrasolution & curvilinear solution in radial (x) direction: • BLACK: left & right side identical • BLUE: left & right side contain sameelements (e.g. cos(nt), sin(2nt), n2t2) • RED: left & right side do not contain same elements
TASK 2: Curvilinear Solution • Comparison of Volterrasolution & curvilinear solution in along-track (y) direction:
TASK 2: Curvilinear Solution • Comparison of Volterrasolution & curvilinear solution in cross-track (z) direction:
TASK 2: Curvilinear Solution • Plotted results: Case 1—short-term drifting motion xy Orbit Track xy Orbit Track
TASK 2: Curvilinear Solution • Plotted results: Case 1—short-term drifting motion x Error y Error
TASK 2: Curvilinear Solution • Plotted results: Case 2—long-term drifting motion xy Orbit Track xy Orbit Track
TASK 2: Curvilinear Solution • Plotted results: Case 2—long-term drifting motion x Error y Error
TASK 2: Curvilinear Solution • Plotted results: Case 2—very long-term drifting motion xy Orbit Track
TASK 2: Curvilinear Solution • So which of these solutions is preferable? • Answer will likely depend on the application • QV seems to have best short-term propagation accuracy • DTS seems to have best long-term propagation accuracy • Polynomial solutions (QV & ADTS) lend themselves to closed-form IROD solution • DTS/ADTS solutions required far fewer man-hours toderive (approx 1/20th the time)
Projected FY15 Effort • Further comparison of Volterra-based & curvilinear-based models • Derive closed-form nonlinear relative motion solution incorporating J2 gravity (+ perhaps other perturbation forces) • Will consider both candidate methods (Volterra & curvilinear) • Implement curvilinear-based model (ADTS) in IROD • Derive geometric parameter set (“relative orbit elements”) based on 1 or more existing solutions