On-board parameter learning using a model reference adaptive position and attitude controller

The Synchronized Position Hold Engage Reorient Experimental Satellites (SPHERES) is a testbed aboard the International Space Station (ISS) developed by MIT's Space System Laboratory (SSL) to test control and autonomy algorithms in a full six degrees of freedom (6DOF) microgravity environment. With the recent launch of the Halo Expansion Device, an attachment for the SPHERES testbed that accommodates up to six peripheral devices - such as docking ports, stereo-vision cameras, robotic arms, etc. - the available hardware configuration space for the SPHERES facility has been greatly expanded. However, this capability increase comes with an increment in complexity; without being able to fully recreate the space environment on the ground, it is not possible to determine all the parameters needed to entirely describe the system dynamics without completing system identification tests on-orbit (aboard the ISS, in this case). Additionally, given the size of the space spanned by all the possible peripheral configurations, the task of performing identification methods on-orbit starts to become intractable, significantly hindering the available control authority to be exerted upon the system. This problem is not particular to SPHERES, but it is also a potential concern for spacecraft performing actions such as autonomous docking, reconfiguration, or assembly (e.g., tugging a defunct satellite and docking to an uncooperative tumbling target). In order to overcome and be robust against the unknown properties and poor models caused by said issues, a position and attitude Model Reference Adaptive Controller (MRAC) was designed and implemented to perform on-board learning of the unknown model parameters. Ground tests using distinct hardware configurations were conducted using the SPHERES platform and its glass table facility to assess the algorithm's performance.