ESA GNC Conference Papers Repository
Tapping the Potential of Adaptable Control and Estimation Methods for Satellite AOCS
In the last decades, linear parameter-varying (LPV) control methods have found their way into the aerospace domain. Their theoretical basis is now well understood and practical design tools are now readily available. Examples of successful application of LPV methods include aircraft wing flutter control and the control of space launch vehicles. Especially the latter can be considered a more than suitable application due to the small time constants and wide range of dynamical properties. For satellites, on the other hand, the challenges posed by parameter variations are less overt. A major reason lies in the comparatively long time constants of the spacecraft dynamics and environmental conditions along with modest ranges of parameter variations, hence the satellite dynamics can be treated as quasi-time invariant. A typical example are solar arrays, which usually rotate at a rate of one revolution per orbit for Earth observation satellites (i.e., less than 0.1deg/s even for low Earth orbits). As past experiences have shown, the induced changes of the spacecrafts dynamical properties can be readily covered by classical robust LTI controller synthesis methods while still meeting the stability and performance requirements of current satellite missions. Nevertheless, accounting for parameter variations in the controller design process can have benefits, some of which lie beyond typical closed-loop performance metrics. Motivated by the on-going ESA-funded ACE study (Adaptable Control and Estimation with Guaranteed Robust Performance), this paper will therefore discuss several challenges and opportunities of current practical interest that are connected with time-varying parameters. Some examples will be outlined in the following. The current trend of commercialisation has introduced product lines into satellite development and production. Among a product line, individual satellites are mostly similar (or even identical), but may still differ from each other in terms of e.g. payload. As a result, many AOCS/GNC relevant properties are constant along the product line, while some parameters such as the moment of inertia vary within certain bounds. A natural goal is therefore to aim for a single adaptable controller for the entire product line with suitable scheduling parameters to adapt the controller for each individual satellite. The main benefit is the increased efficiency of the controller design process itself, which becomes significant already in the extreme case of a product line of two satellites such as the MetOp-SG spacecraft. This product line approach can even be transferred to the design process of a single spacecraft where relevant properties such as mass, centre of mass, and inertia (MCI) are typically known only with large uncertainties in early design phases, but become more and more certain as the design matures. Compared with designing the controllers robust against the initial large uncertainty, the achievable performance could be increased by requiring robustness only against the final small uncertainties and treating the nominal parameter values in a parameter-varying framework. On mission level, the ability to explicitly account for parameter variations enables additional operational modes. For example, some science spacecraft such as the envisaged Athena mission contain several instruments that share the same optics, requiring large movable parts (and thus, masses) to align the correct detector to the line of sight. Reorienting both the spacecraft and the movable parts simultaneously can save a significant amount of manoeuvring time, thus increasing the scientific output. Laser communication terminals require high pointing stability, but are subjected to structural vibrations with varying frequencies (e.g. reaction wheel induced microvibrations that depend on the wheel speeds). For this application, an adaptable disturbance estimation and cancellation scheme can allow using less performant (and thus, less expensive) hardware while still fulfilling all performance requirements. In the full paper, these approaches will be discussed in more detail including examples on how to recast the relevant spacecraft dynamics in an adaptable or LPV framework. Afterwards, some points regarding the industrial application of adaptable and parameter-varying methods will be addressed.