Cargando…
Synthetic database of space objects encounter events subject to epistemic uncertainty
The databases included on this article refers to variables and parameters belonging to the Space Traffic Management (STM), Evidence Theory and Machine Learning (ML) fields. They have been used for implementing ML for autonomously predict risk associated to a close encounter between two space (Sanche...
Autores principales: | , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
Elsevier
2020
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7495056/ https://www.ncbi.nlm.nih.gov/pubmed/32984484 http://dx.doi.org/10.1016/j.dib.2020.106298 |
Sumario: | The databases included on this article refers to variables and parameters belonging to the Space Traffic Management (STM), Evidence Theory and Machine Learning (ML) fields. They have been used for implementing ML for autonomously predict risk associated to a close encounter between two space (Sanchez and Vasile, On the Use of Machine Learning and Evidence Theory to Improve Collision Risk Management, Acta Astronautica, Special Issue for ICSSA2020, In Press [1]). The position of the objected is assumed to be affected by epistemic uncertainty, which has been modeled according to Dempster-Shafer Evidence theory (DSt) [2]. Six datasets are presented. Two (DB1 and DB2, respectively) include samples of space object close encounters subject to epistemic uncertainty on the relative position. Other two databases (DB3 and DB4, respectively) include the values of the Cumulative Plausibility and Belief Curves (CPC and CBC, respectively) of each sample included in DB1. The remaining databases (DB5 and DB6), contain the value of the CPC and CBC of each sample included in DB2. All of them are synthetic databases created using computer simulation to obtain the results presented in [1]. DB1 database is constituted by 9,000 samples and 45 columns and a header, while DB2 is formed by 28,800 samples and 45 columns and a header. These databases come from a set of, respectively, 5 and 14 different families of encounter geometries defined by the range of values that can be assigned to the bounds of the intervals for the uncertain variables, assumed to be affected by epistemic uncertainty, considered to have been provided by two sources of information. The uncertain variables are: the miss distance, [µ(x), µ(y)], on the impact plane (B plane), the standard deviation of the relative position projected on the B plane, [σ(x), σ(y)], and the Hard Body Radius of the combined objects, HBR. The dataset is completed with STM related parameters: miss distance and covariance matrix of the uncertain ellipse projected on the B plane enclosing all samples defined by the uncertainty intervals, the Probability of Collision (P(C)) of this ellipse or the elapsed time to the Time of Closest Approach (TCA); with DSt related parameters: Belief and Plausibility of certain values of Pc; and the class of the event according to the classification detailed in [1]. DB3 and DB4 are constituted by 34 columns and 9000 rows containing the Plausibility and Belief for Pc values and the corresponding Probabilities of Collision necessary to build the CPC and CBC of the events in DB1, while DB5 and DB6 are constituted by 34 columns and 28,800 rows containing the Plausibility and Belief for Pc values and the corresponding Probabilities of Collision values necessary to build the CPC and CBC of the events in DB2. These databases have a potential usage by the ML community interested in STM as well as for the space community, especially, space operators interested in introduce epistemic uncertainty on collision risk assessment. These databases contribute to build a scarce field such as the databases of encounter events [3]. |
---|