This thesis presents a visual servoing system that was developed to track the atmospheric re-entry of the Hayabusa spacecraft in June of 2010. The primary objective of this research was to experimentally measure the spectral emissions radiated from the surface of the Sample Return Capsule (SRC) and the superheated gases within the shock layer during its descent. In achieving this, a ground-based observation was undertaken to autonomously track and record spectral emissions in the visible and near IR.
Although a significant volume of research has already been conducted in visual servoing over the last two decades, attempts to track a spacecraft re-entry for the purposes of recording its spectral emissions has not been published. The use of visual servoing for the tracking component of this mission is the key in robustly recording spectra at high spatial resolution, avoiding the reliance on manual (hand-held) techniques.
The ground based setup implemented a visual servoing scheme that employed a classical image-based approach to identify and track the SRC. The methodology included Lucas-Kanade pyramidal approach to identify the feature on the image plane combined with visual feedback control strategy to allow automated tracking of the target. The acquisition of the SRC and the radiation emitted from its surface and the shock layer was recorded by a co-aligned setup consisting of a two Point Grey cameras. The tracking camera was attached with a 6mm lens providing a 30o field of view (FOV) while the co-aligned spectrometer was attached with a 25mm lens providing a FOV of 9o. The spectrometer consisted of a 300l/mm near infra-red transmission grating attached before the camera lens which was spatially and spectrally calibrated to capture a wavelength range of 450-900nm (with peak efficiency at 700nm) and a maximum spatial resolution of approximately 25m/pixel (vertically resolved).
This visual servoing setup successfully tracked the re-entry with a mean error of approximately 15 pixels and 5 pixels (operating at a 1024x768 pixel resolution) and a maximum angle deviation of 0.9o and 0.3o in the pan and tilt axes respectively. In addition, spectra were recorded for a period of 43 seconds with the SRC spectra being spatially distinguishable from the spacecraft bus for a period of 27 seconds. Preliminary analysis of the emissions data has shown strong peaks of atomic oxygen (O) at 777nm & 844nm and atomic nitrogen (N) at 821nm & 868nm with the estimated peak average surface temperature of the heat shield in the order of 3100 100K.
The resolution of the optics utilized to observe Hayabusa was sufficient to deduce the spectrum and conditions at the surface of the shock layer as a whole but did not allow spatial resolution of the SRC surface. To obtain the physical conditions at the shock layer as a function of surface area; higher focal length lens were warranted. This would effectively reduce the observed FOV and therefore reduce the margin for error in tracking. To meet this demanding requirement we subsequently improved the control that was implemented for the Hayabusa re-entry by introducing a feed-forward strategy which adds a simplified dynamic model of the system. We demonstrated the improvements in tracking performance by undertaking an additional experiment which involved the tracking of the International Space Station (ISS) using a narrow FOV lens which essentially provided a target , with image-based velocity characteristics, similar in magnitude to a hypersonic re-entry. The results show that an object with a velocity of Hayabusa can be tracked reliably within a 0.2o FOV in each axis and thus provide a spatial resolution of approximately 27-45cm/pixel for a 60-100km range. While this is still too large to spatially resolve a re-entry vehicle of the size of the Hayabusa capsule which was only 40 cm in diameter, it represents a considerable improvement over existing capabilities and would enable the spatial resolution of larger vehicles during future re-entry observation missions.