ESA GNC Conference Papers Repository
Structured H-infinity control design for the VEGA launch vehicle: recovery of the legacy control behaviour
The design of the ascent-flight control system of a launch vehicle for the atmospheric phase is a challenging task. Along this first phase of the mission, the launch vehicle is considerably affected by several undesired effects such as high aerodynamic loads, wind disturbances and rapid system dynamic changes. Further complications are introduced by the elastic behaviour of the launch vehicle, which may cause instability. In the face of all these adverse effects, the control system must satisfy very demanding and tight performance requirements and still be robust against a large range of substantial parameter dispersion. This robustness is generally verified in terms of classical stability margins and validated by exhaustive Monte-Carlo time-domain simulations using a high-fidelity model. As demonstrated by the current state-of-practice, there is a rich heritage and experience in applying classical control solutions to the launcher problem. In addition, in order to deal with the wide dynamic variation along the atmospheric phase the classical approach uses the so-called Gain Scheduling (GS) scheme. It consists of linearizing the vehicle around several representative points along the flight trajectory and designing a frozen-time controller at each point. These individual controllers are then interpolated based on a parameter (e.g. time or non-gravitational-speed) resulting in a scheduled controller. This is the design strategy used by the small European VEGA launcher, which uses a classical (proportional-derivative plus bending filters) controller for the Thrust Vector Control (TVC) system. This strategy has been proven successful for the eight flights VEGA has performed so far but several practical limitations are recognized. First, since each point design is a complex process which has to account for different concurrent requirements, the GS strategy generally results in an expensive (in terms of both cost and time) design and validation process. Second, the classical framework only takes uncertainties in an implicit fashion using classical stability margins (gain and phase margins), and thus it must rely in a good analysis coverage (i.e. again costly and time intensive). Third, the performance and the robustness are not guaranteed for the flight instants between the design points of the scheduled controller. This feature is especially critical when the dynamics change rapidly between design points. And fourth, the tuning process is partially automated, with limited or no connection to other points or across missions. In order to address these limitations, advanced control synthesis techniques such as 8 and the Structured H8 approaches, as well as linear parameter varying and adaptive methods are being considered within the frame of an ESA Networking Partnering Initiative (NPI No. 4000114460) participated by ESA-ESTEC, ELV and the University of Bristol with the aim to provide a methodological framework for launcher atmospheric control design and transfer it to industry. In this paper, the recently developed Structured H8 approach  is applied on the VEGA launch vehicle for the design of the atmospheric ascent-flight control. This new approach, which is based on 8 theory, allows defining a specific order and structure for the controller. This technique has shown great promise and has led to intense study by the community, even resulting already in relevant Space flown missions, such as ESA Rosetta and CNES Microscope. Nonetheless, these applications are being performed by H8 experts, since unlike classical control, H8 requirements are expressed in terms of weighting functions in the frequency domain. The authors feel that a detailed understanding of the H8 metrics, based on the so-called sensitivity functions, reconciled with physical system effects, is required for adequate transfer to industrial control engineers with a more classical control background. First, the analytical model used for the controller design is described. Secondly, in order to facilitate the weight selection, the main closed-loop transfer functions will be presented and their physical meaning described. This step is key to be able to express system requirements such as tracking, drift/drift-rate control, disturbance rejection and actuation effort to the frequency-domain. Then, based on the actual VEGA vv05 mission controller and data, the H8 metrics and the classical VEGA TVC structure are used to guide the Structure H8 design to obtain the same controllers used as baseline along different design points. The results clearly show that it is possible to recover the behaviour of the classically, and designed in ad-hoc semi-automated manner, VEGA controller but in an automated manner. This is quite promising as it shows a way to exploit, and build upon, the legacy know-how in launcher control design while addressing current shortcomings of the industrial approach. In addition, this work paves the way to explore the potentials for improvement by using a robust control design framework, which provides better robustness and performance capabilities, as well as powerful robust analysis tools.