competitive data science

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform various operators who can then plan risk mitigation measures. Such measures can be aided by the development of suitable machine learning (ML) models that predict, for example, the evolution of the collision risk over time. In October 2019, in an attempt to study this opportunity, the European Space Agency released a large curated dataset containing information about close approach events in the form of conjunction data messages (CDMs), which was collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, which was an ML competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying ML methods to this problem domain.

As the threat of space debris grows larger, it is becoming increasingly important to develop methods to conduct spacecraft collision risk analysis, and subsequently automate spacecraft collision avoidance. To contribute towards this goal, I explored and tried to identify the most ideal method to conduct spacecraft risk analysis. In this paper, I present the result of this research, and showcase the importance of utilising Bayesian Machine Learning. A machine learning model using Bayesian Machine Learning has been trained and tested, and the results displayed.

After decades of space travel, low Earth orbit is a junkyard of discarded rocket bodies, dead satellites, and millions of pieces of debris from collisions and explosions. Objects in high enough altitudes do not re-enter and burn up in the atmosphere, but stay in orbit around Earth for a long time. With a speed of 28,000 km/h, collisions in these orbits can generate fragments and potentially trigger a cascade of more collisions known as the Kessler syndrome. This could pose a planetary challenge, because the phenomenon could escalate to the point of hindering future space operations and damaging satellite infrastructure critical for space and Earth science applications. As commercial entities place mega-constellations of satellites in orbit, the burden on operators conducting collision avoidance manoeuvres will increase. For this reason, development of automated tools that predict potential collision events (conjunctions) is critical. We introduce a Bayesian deep learning approach to this problem, and develop recurrent neural network architectures (LSTMs) that work with time series of conjunction data messages (CDMs), a standard data format used by the space community. We show that our method can be used to model all CDM features simultaneously, including the time of arrival of future CDMs, providing predictions of conjunction event evolution with associated uncertainties.

Over 34,000 objects bigger than 10 cm in length are known to orbit Earth. Among them, only a small percentage are active satellites, while the rest of the population is made of dead satellites, rocket bodies, and debris that pose a collision threat to operational spacecraft. Furthermore, the predicted growth of the space sector and the planned launch of megaconstellations will add even more complexity, therefore causing the collision risk and the burden on space operators to increase. Managing this complex framework with internationally agreed methods is pivotal and urgent. In this context, we build a novel physics-based probabilistic generative model for synthetically generating conjunction data messages, calibrated using real data. By conditioning on observations, we use the model to obtain posterior distributions via Bayesian inference. We show that the probabilistic programming approach to conjunction assessment can help in making predictions and in finding the parameters that explain the observed data in conjunction data messages, thus shedding more light on key variables and orbital characteristics that more likely lead to conjunction events. Moreover, our technique enables the generation of physically accurate synthetic datasets of collisions, answering a fundamental need of the space and machine learning communities working in this area.

Most of existing algorithms deal with the problem of collision risk assuming spherical objects. This assumption works fine if the size of the satellite is much smaller than the standard deviation of the position uncertainty and it shows approximately the same area for any attitude, as the collision risk will depend on the total area exposed and not to the precise shape of the objects. As the orbit determination algorithms and the surveillance systems improve, this assumption may fail for some satellites. This situation can be solved using complex geometries definition, where a satellite can be defined differently to an sphere. A method implemented in the ESA operational CORAM tool is presented in this paper. The method is based on addressing complex geometries by assuming a satellite composed of several oriented boxes. The algorithm is based on the computation of the collision volume accounting for the geometry of the two objects. While in the spherical case the hard-body object (co...

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform various operators who can then plan risk mitigation measures. Such measures can be aided by the development of suitable machine learning (ML) models that predict, for example, the evolution of the collision risk over time. In October 2019, in an attempt to study this opportunity, the European Space Agency released a large curated dataset containing information about close approach events in the form of conjunction data messages (CDMs), which was collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, which was an ML competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying ...

We investigate the performance of a generalized linear mixed model in predicting the Probabilities of Collision (Pc) for conjunction events. Specifically, we apply this model to the log(sub 10) transformation of these probabilities and argue that this transformation yields values that can be considered bounded in practice. Additionally, this bounded random variable, after scaling, is zero-inflated. Consequently, we model these values using the zero-inflated Beta distribution, and utilize the Bayesian paradigm and the mixed model framework to borrow information from past and current events. This provides a natural way to model the data and provides a basis for answering questions of interest, such as what is the likelihood of observing a probability of collision equal to the effective value of zero on a subsequent observation.

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform various operators who can then plan risk mitigation measures. Such measures can be aided by the development of suitable machine learning (ML) models that predict, for example, the evolution of the collision risk over time. In October 2019, in an attempt to study this opportunity, the European Space Agency released a large curated dataset containing information about close approach events in the form of conjunction data messages (CDMs), which was collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, which was an ML competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying ...

How predictable are life trajectories? We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study. Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model. Within each outcome, prediction error was strongly associated with the family being predicted and weakly associated with the technique used to generate the prediction. Overall, these results suggest practical limits to the predictability of life outcomes in some settings and illustrate the value of mass collaborations in the social sciences.

Rankings compare the performance of organizations. In many cases, rankings provide a good assessment of successful or-ganizations. However, rankings often generate controversy and debate since they support the making decisions. A ranking is a weighted linear combination of indicators, and the weights assigned to each of the indicators can lead to different rank orders. In most cases, rankings are used as a tool to support making decisions, such as resource allocation; therefore, these decisions can be affected by the assignment of such weights. In this article, we analyze the behavior of a ranking and the weights; simulations are used to calculate the change in the order of the equally weighted ranking and of the randomly weighted ranking. In this regard, we present a discussion and ranking design alternatives.

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform various operators who can then plan risk mitigation measures. Such measures can be aided by the development of suitable machine learning (ML) models that predict, for example, the evolution of the collision risk over time. In October 2019, in an attempt to study this opportunity, the European Space Agency released a large curated dataset containing information about close approach events in the form of conjunction data messages (CDMs), which was collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, which was an ML competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying ...

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform various operators who can then plan risk mitigation measures. Such measures can be aided by the development of suitable machine learning (ML) models that predict, for example, the evolution of the collision risk over time. In October 2019, in an attempt to study this opportunity, the European Space Agency released a large curated dataset containing information about close approach events in the form of conjunction data messages (CDMs), which was collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, which was an ML competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying ...

In near future, commercial companies have proposed to launch the extreme large constellation of small to medium-sized satellites over new 20,000 satellites in orbit. As a result, satellite operators will face a new challenge and have to prepare some solutions to mitigate huge potential collision risks. Under current practices, operators receive collision warning messages (Conjunction Data Messages: CDMs) generated by the Combined Space Operations Center (CSpOC). It is possibly to receive huge conjunctions that exceed warning threshold control each day but many of them are false conjunctions because the result of a combination of inaccuracies in the sensor measurements, predicting error and especially inadequacy of satellite information (e.g., GPS data and satellite size) To reduce the false CDMs and effectively mitigate the potential risk in time, space traffic management system can minimize the problems. The key system capabilities are able to automatically screen potential risky objects and provide an analysis tool to decide effective maneuver planning. As a result, GISTDA initiates the project to develop the space traffic management system known as "ZIRCON" to monitor and mitigate the risk of GISTDA's satellites and future satellites (e.g., THEOS2 project). This paper presents the operational collision avoidance process and algorithm analysis of space traffic management system. The accuracy of analysis results is validated by comparing the analyzed results of CSpOC and Space Data Association (SDA) serving the commercial space sector to monitor and provide warning message for membership. The last section describes the future development planning and services for satellite operators.

Two-line elements are widely used for space operations to predict the orbit with a moderate accuracy for 2-3 days. Local optimization methods, such as the nonlinear least squares method with differential corrections, can estimate a TLE as long as there exists an initial estimate that provides the desired precision. Global optimization methods to estimate TLEs are computationally intensive, and estimating a large number of them is prohibitive. In this paper, the feasibility of estimating TLEs using machine learning methods is investigated. First, a Monte-Carlo approach to estimate a TLE, when there are no initial estimates that provide the desired precision, is introduced. The proposed Monte-Carlo method is shown to estimate TLEs with root mean square errors below 1 km for space objects with varying area-to-mass ratios and orbital characteristics. Second, gradient boosting decision trees and fullyconnected neural networks are trained to map orbital evolution of space objects to the associated TLEs using 8 million publicly available TLEs from the US space catalog. The desired precision in the mapping to estimate a TLE is achieved for one of the three test cases, which is a low area-to-mass ratio space object.

How predictable are life trajectories? We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study. Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model. Within each outcome, prediction error was strongly associated with the family being predicted and weakly associated with the technique used to generate the prediction. Overall, these results suggest practical limits to the predictability of life outcomes in some settings and illustrate the value of mass collaborations in the social sciences.

How predictable are life trajectories? We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study. Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model. Within each outcome, prediction error was strongly associated with the family being predicted and weakly associated with the technique used to generate the prediction. Overall, these results suggest practical limits to the predictability of life outcomes in some settings and illustrate the value of mass collaborations in the social sciences.

How predictable are life trajectories? We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study. Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model. Within each outcome, prediction error was strongly associated with the family being predicted and weakly associated with the technique used to generate the prediction. Overall, these results suggest practical limits to the predictability of life outcomes in some settings and illustrate the value of mass collaborations in the social sciences.

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform the various operators who can then plan risk mitigation measures. Such measures could be aided by the development of suitable machine learning models predicting, for example, the evolution of the collision risk in time. In an attempt to study this opportunity, the European Space Agency released, in October 2019, a large curated dataset containing information about close approach events, in the form of Conjunction Data Messages (CDMs), collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, a machine learning competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying machine ...

How predictable are life trajectories? We investigated this question with a scientific mass collaboration using the common task method; 160 teams built predictive models for six life outcomes using data from the Fragile Families and Child Wellbeing Study, a high-quality birth cohort study. Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model. Within each outcome, prediction error was strongly associated with the family being predicted and weakly associated with the technique used to generate the prediction. Overall, these results suggest practical limits to the predictability of life outcomes in some settings and illustrate the value of mass collaborations in the social sciences.

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform the various operators who can then plan risk mitigation measures. Such measures could be aided by the development of suitable machine learning models predicting, for example, the evolution of the collision risk in time. In an attempt to study this opportunity, the European Space Agency released, in October 2019, a large curated dataset containing information about close approach events, in the form of Conjunction Data Messages (CDMs), collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, a machine learning competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying machine ...

Today, active collision avoidance has become a routine task in space operations, relying on validated, accurate and timely space surveillance data. For a typical satellite in LEO, hundreds of conjunction alerts can be expected every week. Processing and filtering these still leaves about two actionable alerts per spacecraft and week requiring detailed follow-up by an analyst. At ESA, more than one collision avoidance maneuver can be expected per satellite and year. It is clear that such an approach requiring 24/7 expert availability to analyze more than 20 parameters and constraints generates high operational costs. The future with an accelerating launch rate and deployments of smaller satellites and constellations, as well as improved space surveillance networks delivering catalogues of up to 200k objects will render efforts by any operator following this approach an unmanageable task. ESA's proposal for a Space Safety Programme to start in 2020 includes also a cornerstone &quo...

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform the various operators who can then plan risk mitigation measures. Such measures could be aided by the development of suitable machine learning models predicting, for example, the evolution of the collision risk in time. In an attempt to study this opportunity, the European Space Agency released, in October 2019, a large curated dataset containing information about close approach events, in the form of Conjunction Data Messages (CDMs), collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, a machine learning competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying machine ...

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform the various operators who can then plan risk mitigation measures. Such measures could be aided by the development of suitable machine learning models predicting, for example, the evolution of the collision risk in time. In an attempt to study this opportunity, the European Space Agency released, in October 2019, a large curated dataset containing information about close approach events, in the form of Conjunction Data Messages (CDMs), collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, a machine learning competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying machine ...

The number of trackable resident space objects (RSOs) is increasing due to new launches and novel space surveillance sensors. Therefore, it is desired to automate the collision risk prediction and mitigation (CREAM) to keep pace with the number of conjunction data messages (CDMs). Recently, the feasibility of leveraging machine learning models has been investigated to reduce the false detections in the literature, and ESA collision risk prediction challenge hosted on Kelvins platform provided the publicly available dataset. The proposed work benchmarks the machine learning models and frameworks that have been studied for collision risk prediction to evaluate their suitability for real world deployment. Therefore, machine learning models are benchmarked against the naive solution, which considers the risk value at cutoff day as risk at close approach time, using JSpOC CDMs issued for satellites that ESA maintains. This work shows that Bayesian neural networks, Siamese-embedding, Boosting decision trees models, and Linear Discriminant Analysis (LDA) are promising machine learning approaches for collision risk prediction.

The General Mission Analysis Tool (GMAT) is an open-source astrodynamics tool, developed in partnership between NASA's Goddard Space Flight Center (GSFC), industry partners, and academic institutions. 1 GMAT is in operational use in the GSFC Flight Dynamics Facility, providing guidance and navigation support for active missions, and as a mission design tool at NASA and industrial organizations. The GMAT R2020a release includes a new GMAT Application Programming Interface (API) that provides access to core GMAT capabilities from Python and Java, and, through the Java interface, from MATLAB. GMAT's API is a tested and documented beta-quality feature of the 2020 release. It includes sample scripts that can be executed in Python or MATLAB, tutorial walk-throughs presented as interactive notebooks, and more than 70 pages of user documentation. The GMAT API can also be run interactively. When working with the tool in this mode, users have access to a live help system that presents available options to the user to simplify use of the system. Configuration for the GMAT API is straightforward on Linux, Windows, and Mac workstations. This paper presents these features and several use cases for the GMAT API. One core feature of the GMAT API is the access to GMAT's core components for direct manipulation and use. Access to GMAT's components enables the development of tutorial materials, presented as Jupyter notebooks, that introduce users to GMAT's components and walk users through interactive use of these components using the native user interface elements of Python or MATLAB. The GMAT API provides a toolkit to build graphical tools hosted in language native environments. Access to GMAT's computational models for propagation and navigation models enable use of those features for analysis of a user's data, including live streaming of measurements into custom tools to validate those data streams in real time. This API provides users with easy access to a library of astrodynam-ics utilities that are rigorously tested and trusted operationally by a wide range of NASA and private missions, reducing the need for individual users to implement and test their own utilities. These utilities range from time system conversions, to coordinate system conversions, all the way to simulating measurements including corrections such as light-time delays and relativistic corrections. All of the GMAT scripting capabilities are accessible using the GMAT API. The script interfaces enable simple configuration of large scale analysis problems, ranging from scans of parameters to explore a solution space to more complex problems like Monte-Carlo analyses. These capabilities are presented with sample problems designed to illustrate their use. Current work on the GMAT API includes integration of GMAT and Monte into a unified environment, driven from Monte's Python based tool set. The status of this work is also described in this paper.

Spacecraft collision avoidance procedures have become an essential part of satellite operations. Complex and constantly updated estimates of the collision risk between orbiting objects inform the various operators who can then plan risk mitigation measures. Such measures could be aided by the development of suitable machine learning models predicting, for example, the evolution of the collision risk in time. In an attempt to study this opportunity, the European Space Agency released, in October 2019, a large curated dataset containing information about close approach events, in the form of Conjunction Data Messages (CDMs), collected from 2015 to 2019. This dataset was used in the Spacecraft Collision Avoidance Challenge, a machine learning competition where participants had to build models to predict the final collision risk between orbiting objects. This paper describes the design and results of the competition and discusses the challenges and lessons learned when applying machine learning methods to this problem domain.

This Letter to 'The Sciences' (the magazine of the New York Academy of Sciences) argues that the danger of polluting planetary atmospheres with Earth contaminants and micro-organisms has not been sufficiently studied and that more thorough cleaning of spacecraft is an urgent matter.

An evaluation is made of the application of novel, AI-capabilities-related technologies to aerospace systems. Attention is given to expert-system shells for Space Shuttle Orbiter mission control, manpower and processing cost reductions at the NASA Kennedy Space Center's 'firing rooms' for liftoff monitoring, the automation of planetary exploration systems such as semiautonomous mobile robots, and AI for battlefield staff-related functions.