predefense

The Distributed Spacecraft Attitude Control System Simulator: From Design Concept to Decentralized Control : The Distributed Spacecraft Attitude Control System Simulator: From Design Concept to Decentralized Control Jana Schwartz jana@vt.edu November 14, 2003 PREDEFENSE


Overview : Overview


Overview : Overview Schwartz, Peck & Hall, JGCD 03 VanDyke, Schwartz & Hall, AAS 04 Schwartz & Hall, AAS 03 Schwartz & Hall, AAS 04 Schwartz & Hall, FMS 03


Overview : Overview


Whorl Development : Whorl Development Stable Whorl-I running Reaction wheel and momentum wheel controllers “working” CPU speed limits algorithm complexity Insufficient attitude information Whorl-II under construction Computer + sensors operational soon Whorl-I following Whorl-II Kudos to everyone who works in the lab: this is not a one-person project!


Whorl Equations of Motion : Whorl Equations of Motion Easy when considering torques only from cg offset & actuators Effect of the non-inertial lab Dynamics Drag torque order of magnitude greater than rotating Earth torque Sensors Rate gyro error constant, will be calibrated out Accelerometer error has constant bias and attitude dependent terms Bias will be calibrated out Variations O(10-5 g), well within noise band


Whorl Parameter Estimation : Whorl Parameter Estimation Want to improve values for moments of inertia and cg vector obtained from CAD model analysis Commercial components modeled with uniform density Wiring harness prohibitively complex to include Linear least squares Simulations perform well Torque method, natural and integrated Energy balance Momentum integral, natural and integrated Present experimental results poor Run with momentum wheel controller Improve data from motor Improve attitude filter Iterative procedure: alternate estimation of I and rcg


Whorl Parameter Estimation (2) : Whorl Parameter Estimation (2) Sequential techniques EKF simulations not promising SPKF under development Total least squares (Psiaki, FMS 03) Not yet implemented Should be robust to uncertainties from poor data Algorithm assumes perfect attitude knowledge (momentum integral technique)


Hardware Summary : Hardware Summary Unresolved problems Attitude sensors CPU speed Unresolved questions Parameter estimation x3 Break for discussion at this point? Time for another cookie?


And now for something completely different : And now for something completely different Glossing over single-satellite theory On to the good stuff!


Relative Orbital Dynamics : Relative Orbital Dynamics A wide array of literature Two-body Perturbations Earth oblateness (Ross, JGCD 03) Differential drag (Carter & Humi, JGCD 02; Lovell et al., AAS 03) Time-varying differential drag Attitude Mass (due to thrust profile) Altitude


Relative Orbital Dynamics (2) : Relative Orbital Dynamics (2) A wide array of literature Operational constraints Formation establishment Out of scope Fuel optimal formation maintenance As an independent problem, out of scope Fuel distribution within formation (Vadali, Vaddi & Alfriend, IJRNC 02) Scientific effectiveness (Hughes & Hall, JAS 02)


Relative Orbital Dynamics (3) : Relative Orbital Dynamics (3) Vadali, Vaddi & Alfriend, IJRNC 02 Minimize total fuel consumption Maintain equal average fuel consumption per spacecraft Non-optimal, but uses 33% less fuel Hughes & Hall, JAS 02 Explores constant shape formations Apply scientific performance criteria to above fuel optimal formations


Relative Attitude Dynamics : Relative Attitude Dynamics Xing & Parvez, JGCD 01 Relative attitude control law decoupled from orbital dynamics Reduce tracking problem to regulator problem, ( ds, dw )  (0,0) Wang & Hadaegh, JAS 96 Partially decentralized orbit control scheme Leader-follower and DW formations only Attitude control based on perfect orbit control


Relative Attitude Dynamics (2) : Relative Attitude Dynamics (2) Side note: tracking control laws (Long & Hall, FMS 99; Tsiotras, Shen & Hall, AAS 99) Formation “subsatellite” point Stationary target tracking Moving target tracking Requires knowledge of target trajectory Probably out of scope


Distributed Control : Distributed Control Centralized controller Instructions for entire formation determined by chief satellite Low communications cost, high risk Decentralized controller High communications requirement, low risk Typical terrestrial missions share position and velocity data  orbit, attitude, rates Try sharing target information instead Substantially lower communications requirement Perhaps more onboard computing required In each case, what kind of controller will the algebra suggest?


Distributed Control (2) : Distributed Control (2) Decentralized control architectures Virtual structure / perceptive framework Centroid trajectory precomputed, each node knows how to compute its trajectory via geometric offsets Behavioral approach A fancy way of saying “minimize a cost function” Seems to provide more room for true autonomy Maintain fuel-saving drifting formation Maintain constant shape formation (same distance between you and two closest neighbors) Observe target midway between your two closest neighbors’ targets Minimize deviation from nadir pointing attitude Orient thruster in velocity vector direction


Theory Summary : Theory Summary Lots to do! Time-varying differential drag derivation Combining interesting orbit control schemes Relative attitude analysis algebra Defining behaviors for distributed control


Discuss! : Discuss!