A Model Predictive Approach to Satellite Formation Control

It is notoriously costly to send a large satellite full of on-board instruments to space forplanetary exploration and distributed physical experiments. What if, instead, we could split onelarge instrument into several smaller parts, and then launch each smaller part with a smallersatellite? The problem of having to launch big satellites would be solved; moreover it would alsobe cheaper to manufacture and launch several small satellites that one big satellite.But, unfortunately, there is no free lunch. Splitting a big single satellite into a swarm ofsmall micro satellites introduces the problem of maintaining these micro satellites in a certainposition with respect to each other: indeed if we want our instruments to work together, on-boardinstruments should maintain a certain relative distance and orientation between themselves. Inother words, there is the need to achieve and maintain formations, and continuously counteractthe slight differences in disturbances that will act on each satellite.This thesis then considers this need, and proposes control algorithms that guarantee swarmsof satellites to maintain a certain position in space and counteract spatial disturbances.Once again there is no free lunch: to maintain a formation and be robust with respect todisturbances there is the need for using on-board fuel. But on-board fuel is finite: as soon as itruns out, the satellites won’t be able to change their orbit anymore. Our mission is then to finda control strategy that not only guarantees the satellites to maintain a formation irrespectivelyof the disturbances, but also to minimize fuel usage, so to increase the life of the satellites inorbit.For all these purposes we derive and simulate a tailored stochastic centralized Model Pre-dictive Control (MPC) approach for keeping satellites in formation while considering referencepositions, fuel cost and relative position between satellites in its formulation.More specifically we do comparison between different control approaches and study theirperformances. Preliminary simulations have shown that MPC is suitable for satellite formationcontrol and can be used to control a formation on an orbit. There are still some problems, morespecifically we have a centralized controller, which is not robust with respect to failures.Validerat; 20160630 (global_studentproject_submitter