Speakers
Description
The growing population of objects in orbit and the accumulation of space debris over the past decades make it necessary to develop activities and initiatives focused on space debris removal. The first and most fundamental role of Space Surveillance & Tracking (SST) in those space debris removal activities is evident, enabling knowledge and prediction of object orbits and their associated uncertainty through cataloguing methodologies, leveraging SST measurements from networks of sensors (telescopes, radars and laser ranging stations). However, to extend the support to those activities, it is imperative not only to know the orbits, but also multiple aspects related to the characterisation of the objects to be removed such as their size, shape, rotation and attitude. Therefore, advanced techniques further exploiting those SST measurements beyond traditional orbit estimation and prediction methods are essential to achieve a fully comprehensive understanding of the objects. The presentation aims to describe object characterisation methodologies developed at GMV, including size and shape estimation, rotation determination and attitude characterization and estimation, with the objective of enhancing the effectiveness of space debris removal.
The methodologies developed make use of SST sensor data that complement the typical SST measurements (range, range-rate, azimuth, elevation, right-ascension, declination…), such as light-curves (i.e. apparent magnitude) in the case of optical sensors, and Radar Cross Section (RCS) measurements in the case of radars. Additionally, they can also be combined with initial estimates of size, shape, materials, and even attitude, which some other sensor technologies are able to provide.
Our first method involves classifying objects based on their rotational state, distinguishing between tumbling objects and stabilized objects. This is achieved through Machine Learning techniques that use RCS measurements and light-curves as input. The results of this method have been highly successful, surpassing a 90% accuracy rate, particularly when both types of measurements are combined.
Following that, another method involves determining the apparent rotation period of rotating objects using Lomb-Scargle periodograms and epoch folding techniques. This algorithm has been validated with objects of known rotation, yielding errors lower than 1%.
However, the most promising method by far is the attitude estimation based on light-curves. Our implementation relies on an LSM filter that aims to minimize the residuals between the real light-curves provided and those computed from the estimated attitude. To accomplish this, the development of a light-curve simulator, called GRIAL, has been necessary. GRIAL uses OpenGL to model the object and considers many aspects related to light reflection, such as shape, size, materials, solar panels, or the shadows that occur between different parts of the object. This simulator has been successfully validated with known real light-curves of well-known objects such as the SMOS and ESA’s Sentinel satellites. The attitude estimation method can be combined with input data about the object (size, shape, estimated attitude, etc.), whether they are known or estimated, to be used as an initial solution during the estimation process. These data can be provided by external sources, or be outputs of the previous algorithms. The method allows to estimate the objects attitude mode, the refinement of the initial attitude, the determination of the rotation axis and the fine-tuning of the rotation rate of rotating objects. In addition, it is possible to include a scale factor, allowing to deal with uncertainties in the size or reflection coefficient of the materials. The results obtained vary depending on the initial level of knowledge about the object. For example,. when there is good initial knowledge and enough amount of data, the orientation calculation accuracy reaches errors below 15º, although on the other side, the algorithm even diverges in cases with very limited initial knowledge (e.g., no initial information about shape, size, or attitude) or when there is few data available.
Furthermore, the development of the simulator is currently ongoing to expand its capabilities to simulate RCS measurements. This will allow to perform attitude estimation based on this type of measurement as well.
In addition, to deal with cases where no prior information is known about the object, a simple auxiliary method has been recently added that allows a reasonable estimation of the object size using light-curves and RCS. While it is only an approximation, it provides an initial starting point when no further data is available.
As mentioned earlier, successful space debris removal operations need to have knowledge of the orbits of the objects for the effective initial approaches. But in addition, these missions typically demand a detailed understanding of the object characterisation, particularly in aspects such as attitude, rotation, size and shape, to conduct the activity in an efficient and safe manner. Therefore, a combined integration of our methods can be extremely useful for these activities, facilitating the necessary prior knowledge of the object before taking action.
The presentation gives more details on the methodologies indicated and on the results obtained, and presents some practical cases of application. For example, information is provided on the training process and results of the Machine Learning method; it is detailed how we use the periodograms to obtain an initial estimate of the rotation based on light-curves and RCS, and then epoch folding process for fine-tuning; it explains how GRIAL works as a light-curve simulator using OpenGL and 3D modelling, and gives some real validation examples; and it also shows the implementation of the LSM filter for attitude estimation and how GRIAL is integrated in this method, detailing the main results obtained in the different cases tested. Finally, the level of maturity achieved with these methods is analysed, as well as their current limitations.
In conclusion, this presentation serves to show the development and performance of characterisation methodologies that are useful in supporting activities aimed at maintaining a clean and sustainable space environment, also highlighting the main challenges and expectations for the future in this field.
