- Astrodynamics, Orbital Mechanics, Space Systems, Spacecraft Attitude Determination and Control, Numerical Analysis, Dynamic Systems and Control, and 13 moreSpacecraft orbit prediction, Control Theory, Spacecraft orbit determination, Aerospace Engineering, Multi Sensor Data Fusion, Navigation, Optimal Control, Orbit Determination, Parameter estimation, Flight Dynamics and Control, Automatic Control, Machine Learning, and Artificial Intelligenceedit
- My current interest is applying machine learning to the challenges associated with space situational awareness (SSA).edit
Research Interests:
Identification and characterisation of the growing population of resident space objects (RSOs) in orbit around Earth is central to current and future Space Traffic Management and Space Situational Awareness (SSA) activities. Research at... more
Identification and characterisation of the growing population of resident space objects (RSOs) in orbit around Earth is central to current and future Space Traffic Management and Space Situational Awareness (SSA) activities. Research at University of New South Wales (UNSW) Canberra Space seeks to assist this effort through combining optical measurements of selected RSOs with numerical astrodynamics modelling techniques to extend the information that can be inferred about an RSO from its photometric light curve signature. The initial phase of this research comprised two three-month observation campaigns, which were completed in July 2018. A collection of photometric light curves was obtained using different nodes of the Falcon Telescope Network (FTN) for the Buccaneer Risk Mitigation Mission (BRMM) 3U CubeSat. The BRMM was launched in 2017 as a joint mission between UNSW Canberra Space and the Defence Science and Technology Group (DST). While BRMM is a pathfinder for the future Bucca...
Research Interests:
Maneuver detection and characterization in near-real time are helpful for identifying anomalous orbital behavior and associated threats, such as close approaches, to the neighboring space assets. In addition, such capability facilitates... more
Maneuver detection and characterization in near-real time are helpful for identifying anomalous orbital behavior and associated threats, such as close approaches, to the neighboring space assets. In addition, such capability facilitates the maintenance of space object catalogs reducing the workload that is necessary for recapturing the lost objects between observations. Although there are studies that investigate the feasibility of detecting and characterizing maneuvers from observations, they are compute-intensive. This paper presents a deep generative model that can learn the normal orbital behavior of resident space objects and detect maneuvers in near-realtime (NRT). The input data are simulated orbital evolutions, with a high fidelity force and realistic space environment modeling, to represent the distribution that is similar to real observations. Because the success of machine learning methods depends on the representation of the input data, the simulated orbital trajectories...
Surveillance of resident space objects (RSOs) is essential for detection, tracking, and cataloguing them to keep active satellites safe from hazards. High altitude space objects are observed using optical telescopes due to their... more
Surveillance of resident space objects (RSOs) is essential for detection, tracking, and cataloguing them to keep active satellites safe from hazards. High altitude space objects are observed using optical telescopes due to their efficiency. However, it is labour-intensive to detect space objects in the images captured by telescopes, and automation is desired. The proposed framework leverages the Feature Pyramid Network (FPN), a convolutional neural network for image segmentation, to automate RSO detection in the telescope images. The backbone used for detecting low-level patterns from images is the pre-trained EfficientNet-B7 on ImageNet. A simple preprocessing is applied to images that are overexposed to scale the input image pixel values, and this thresholding is only conducted using the statistics of the training data. A custom deterministic post-processing method based on vector mathematics is developed to clean the false detections. F1 score of the proposed machine learning framework is 92%, and this performance shows that the convolutional neural networks can be utilised for automating RSO detection from telescope images.
Research Interests:
This paper describes a method to calculate the optimal impulsive maneuver to avoid the collisions using Simplified General Perturbation 4 (SGP4) and Two Line Element (TLE). It also presents a rigorous analysis of the method to investigate... more
This paper describes a method to calculate the optimal impulsive maneuver to avoid the collisions using Simplified General Perturbation 4 (SGP4) and Two Line Element (TLE). It also presents a rigorous analysis of the method to investigate the relative dynamics of the two colliding space objects assuming the encounter is instantaneous. Different collision geometries are used for test cases. For all collision geometries, the relative velocity and the relative miss distance vectors are parallel, which is the worst case because the encounter occurs along the direction of the relative velocity vector. Test cases are satellites with distinctive orbital characteristics that are obtained from the official Spacetrack catalog. A precise numerical orbit propagator is used both to create ephemerides for test cases and to evaluate the accuracy of the proposed method. This work is significant because this is the first comprehensive investigation of the relative dynamics of the optimal impulsive collision avoidance maneuvers for satellites. Moreover, the proposed method isn't compute-intensive because it is a semi-numerical method.
Research Interests:
Research Interests:
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... more
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 residual mean squared errors below 1 km for space objects with varying area-to-mass ratios and orbital characteristics. Second, gradient boosting decision trees and fully-connected neural networks are trained to map the orbital evolution of space object...
Research Interests:
Research Interests:
Abstract Accurate prediction of aerodynamic forces in Low Earth Orbit (LEO) remains a key challenge for space situational awareness (SSA) and space traffic management (STM) activities. A neglected aspect of the LEO aerodynamics problem is... more
Abstract Accurate prediction of aerodynamic forces in Low Earth Orbit (LEO) remains a key challenge for space situational awareness (SSA) and space traffic management (STM) activities. A neglected aspect of the LEO aerodynamics problem is the force resulting from the charged aerodynamic interaction of LEO objects with the ionosphere, i.e. ionospheric aerodynamics. This work studies the effect accounting for ionospheric drag may have on the motion of LEO objects. This work aims to assess the influence of ionospheric aerodynamics on atmospheric density estimation and orbit prediction capabilities essential to SSA and STM services. The approach taken in this work was to apply Particle-in-Cell (PIC) simulations to develop a surrogate model that describes the variation of a charged drag coefficient ( C D , C ) as a function of plasma scaling parameters. This surrogate model was then incorporated into an orbit propagator, and the influence of ionospheric drag on body motion is studied for a range of conditions. Results indicate that, when inferring atmospheric neutral density from orbit data, neglecting the contribution of ions without accounting for electrodynamic phenomena, may cause an over-prediction of neutral density ranging between 1% and 45% for the space weather condition considered ( F 10.7 = 150 , ap = 5 ). Including electrodynamic phenomena was seen to increase this over-prediction for all cases. Objects with thick plasma sheaths were shown to be particularly sensitive to ionospheric drag forces; thick plasma sheaths caused by either a reduction in object scale ( r B ) or increase in surface potential ( ϕ ). This result has important implications for modelling space debris populations as thick plasma sheaths may arise at natural floating potentials ( - 0.75 V ) when debris fragmentation occurs. For example, accounting ionospheric drag on a spherical debris object with a radius of 0.005 m, area-to-mass ratio of 0.0157 m 2 / kg and floating potential of - 0.75 V in a circular equatorial orbit at 350 km altitude was predicted to cause a change in along-track position of 299 m (and a reduction in semi-major axis of 4 m) over a 24-h period. Results also have important implications for modern satellites, where the trend is toward nano/micro-satellite platforms (e.g. CubeSats) with high-voltage powers systems that may inadvertently cause large artificial surface potentials and therefore enhanced ionospheric aerodynamic forces.
Research Interests:
Research Interests:
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... more
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 ...
Research Interests:
Research Interests:
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... more
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 fully-connected neural networks are trained to map orbital evolution of space objects to the ...
Research Interests:
A channeled Stokes polarimeter that recovers polarimetric signatures across the scene from the modulation induced channels is preferrable for many polarimetric sensing applications. Conventional channeled systems that isolate the intended... more
A channeled Stokes polarimeter that recovers polarimetric signatures across the scene from the modulation induced channels is preferrable for many polarimetric sensing applications. Conventional channeled systems that isolate the intended channels with low-pass filters are sensitive to channel crosstalk effects, and the filters have to be optimized based on the bandwidth profile of scene of interest before applying to each particular scenes to be measured. Here, we introduce a machine learning based channel filtering framework for channeled polarimeters. The machines are trained to predict anti-aliasing filters according to the distribution of the measured data adaptively. A conventional snapshot Stokes polarimeter is simulated to present our machine learning based channel filtering framework. Finally, we demonstrate the advantage of our filtering framework through the comparison of reconstructed polarimetric images with the conventional image reconstruction procedure.
Research Interests:
UNSW Canberra has a program of experiments onboard the M2 formation flying CubeSat mission to provide truth data for available space situational awareness (SSA) sensors and modelling algorithms. The paper outlines the program of... more
UNSW Canberra has a program of experiments onboard the M2 formation flying CubeSat mission to provide truth data for available space situational awareness (SSA) sensors and modelling algorithms. The paper outlines the program of experiments and deployments planned throughout the early, main, and extended operation phases of the mission that provide opportunities for SSA observations. The mission comprises 2x6U CubeSats. Each satellite uses a 3-axis attitude control system to exploit differential atmospheric drag forces between the spacecraft to control the along-track formation. The differential aerodynamic formation control enables the satellites to remain within an acceptable alongtrack offset to perform the main mission experiments. Several important opportunities to collect benchmark SSA data are present throughout the mission. The CubeSat pair are initially conjoined as a 12U satellite and, following a scheduled command from the UNSW Canberra ground station, will be impulsively...
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... more
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 fully-connected 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.
Research Interests:
: A new method for orbit prediction, which is as accurate as numerical methods and as fast as analytical methods, in terms of computational time, is desirable. This paper presents Kolmogorov Arnol'd Moser (KAM) torus orbit prediction... more
: A new method for orbit prediction, which is as accurate as numerical methods and as fast as analytical methods, in terms of computational time, is desirable. This paper presents Kolmogorov Arnol'd Moser (KAM) torus orbit prediction using Simplified General Perturbations 4 (SGP4) and Two-Line Element (TLE) data. First, a periodic orbit and its Floquet solution is calculated. After that, perturbations, which are on the order of 10��5 and smaller, are added to the periodic orbit plus Floquet solution. Then, the low eccentricity KAM torus is least squares fitted to the SGP4 and TLE data. The performance of the theory is presented in various ways. The new method is approximately five times more accurate for the best fits and three times more accurate for mean fits comparing to SGP4 and TLE. History of TLEs and KAM torus theory can be used to make accurate orbit predictions, which is conceptually similar to extrapolation. In addition, the new method may rival numerical methods and i...
Research Interests:
Research Interests:
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... more
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.
Research Interests:
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... more
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 fully-connected 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.
Research Interests:
Identification and characterisation of the growing population of resident space objects (RSOs) in orbit around Earth is central to current and future Space Traffic Management and Space Situational Awareness (SSA) activities. Research at... more
Identification and characterisation of the growing population of resident space objects (RSOs) in orbit around
Earth is central to current and future Space Traffic Management and Space Situational Awareness (SSA)
activities. Research at University of New South Wales (UNSW) Canberra Space seeks to assist this effort
through combining optical measurements of selected RSOs with numerical astrodynamics modelling techniques
to extend the information that can be inferred about an RSO from its photometric light curve signature.
The initial phase of this research comprised two three-month observation campaigns, which were completed in
July 2018. A collection of photometric light curves was obtained using different nodes of the Falcon Telescope
Network (FTN) for the Buccaneer Risk Mitigation Mission (BRMM) 3U CubeSat. The BRMM was launched in
2017 as a joint mission between UNSW Canberra Space and the Defence Science and Technology Group
(DST). While BRMM is a pathfinder for the future Buccaneer Main Mission whose primary objective will be
the calibration of the Jindalee Operational Radar Network (JORN), it also serves as a stepping stone in building
Australian space capability. For BRMM one of the mission objectives was to perform photometric experiments
to contribute to SSA research and development efforts via dynamic on-orbit manoeuvres. This paper reports on
the initial analysis of the photometric light curves central to the SSA mission goal. The material properties and
dynamic attitude motion of the BRMM during the FTN observations are known, with rotational body rates
commanded from 0.2 to 5 degrees per second about multiple combinations of body axes to build a
comprehensive database of light curves for analysis. Further variation in the light curve database is provided by
observations obtained prior to solar panel and antenna deployment.
A set of 70 light curves was obtained during the observational campaigns, with each light curve signature
containing a bulk change in intensity over time due to the change in range as BRMM approaches the FTN node
on its Low Earth Orbit (LEO) trajectory. Superimposed on the mean intensity change are characteristic peaks
and troughs produced by reflections from individual facets of the spacecraft, the magnitude and frequency of
which are highly dependent upon the spacecraft’s attitude and body rate. Samples from the light curve database
are presented with the attitude data downlinked from the spacecraft to assess light curve variations with attitude
and spin rate of the spacecraft.
Supporting the optical data are numerically simulated light curves, generated by applying the Ashikhmin-
Premože Bidirectional Reflectance Distribution Function (BRDF) Model for the BRMM geometry using a high-
fidelity 6 Degree of Freedom (DOF) orbit propagator supported by the Orekit orbit propagation library for
computations related to time systems, coordinate frames, and gravitational perturbations. The performance of
the numerical simulation was evaluated by superimposing the attitude profile reported by the spacecraft
telemetry on top of the propagated orbit to provide a one-to-one comparison between the measured and
simulated light curves for select cases.
A preliminary investigation into the feasibility of using the simulation tool to infer attitude dynamics from a
given light curve signature is also presented. A candidate set of simulated light curves was generated by
numerically propagating a set of initial attitude states and constant body rates through the observation window.
The results were searched to find the case that provided the best fit to the observed light curve. A further study
was initiated to investigate the errors introduced by the assumption of a constant body rate throughout the
observation for the simulated light data.
Earth is central to current and future Space Traffic Management and Space Situational Awareness (SSA)
activities. Research at University of New South Wales (UNSW) Canberra Space seeks to assist this effort
through combining optical measurements of selected RSOs with numerical astrodynamics modelling techniques
to extend the information that can be inferred about an RSO from its photometric light curve signature.
The initial phase of this research comprised two three-month observation campaigns, which were completed in
July 2018. A collection of photometric light curves was obtained using different nodes of the Falcon Telescope
Network (FTN) for the Buccaneer Risk Mitigation Mission (BRMM) 3U CubeSat. The BRMM was launched in
2017 as a joint mission between UNSW Canberra Space and the Defence Science and Technology Group
(DST). While BRMM is a pathfinder for the future Buccaneer Main Mission whose primary objective will be
the calibration of the Jindalee Operational Radar Network (JORN), it also serves as a stepping stone in building
Australian space capability. For BRMM one of the mission objectives was to perform photometric experiments
to contribute to SSA research and development efforts via dynamic on-orbit manoeuvres. This paper reports on
the initial analysis of the photometric light curves central to the SSA mission goal. The material properties and
dynamic attitude motion of the BRMM during the FTN observations are known, with rotational body rates
commanded from 0.2 to 5 degrees per second about multiple combinations of body axes to build a
comprehensive database of light curves for analysis. Further variation in the light curve database is provided by
observations obtained prior to solar panel and antenna deployment.
A set of 70 light curves was obtained during the observational campaigns, with each light curve signature
containing a bulk change in intensity over time due to the change in range as BRMM approaches the FTN node
on its Low Earth Orbit (LEO) trajectory. Superimposed on the mean intensity change are characteristic peaks
and troughs produced by reflections from individual facets of the spacecraft, the magnitude and frequency of
which are highly dependent upon the spacecraft’s attitude and body rate. Samples from the light curve database
are presented with the attitude data downlinked from the spacecraft to assess light curve variations with attitude
and spin rate of the spacecraft.
Supporting the optical data are numerically simulated light curves, generated by applying the Ashikhmin-
Premože Bidirectional Reflectance Distribution Function (BRDF) Model for the BRMM geometry using a high-
fidelity 6 Degree of Freedom (DOF) orbit propagator supported by the Orekit orbit propagation library for
computations related to time systems, coordinate frames, and gravitational perturbations. The performance of
the numerical simulation was evaluated by superimposing the attitude profile reported by the spacecraft
telemetry on top of the propagated orbit to provide a one-to-one comparison between the measured and
simulated light curves for select cases.
A preliminary investigation into the feasibility of using the simulation tool to infer attitude dynamics from a
given light curve signature is also presented. A candidate set of simulated light curves was generated by
numerically propagating a set of initial attitude states and constant body rates through the observation window.
The results were searched to find the case that provided the best fit to the observed light curve. A further study
was initiated to investigate the errors introduced by the assumption of a constant body rate throughout the
observation for the simulated light data.
Research Interests:
This work introduces a novel open source collision avoidance maneuver planning tool for satellites. OpenCAMPT computes the suboptimal collision avoidance maneuvers between two space objects for short-term encounters. OpenCAMPT is the... more
This work introduces a novel open source collision avoidance maneuver planning tool for satellites. OpenCAMPT computes the suboptimal collision avoidance maneuvers between two space objects for short-term encounters. OpenCAMPT is the first tool which provides robust informing suboptimal maneuver decisions. It is fast, scalable and a reliable tool. OpenCAMPT is a standalone software suite written in Cython, C, and C++. It uses numerical orbit propagator to correct the initial estimates of the maneuver provided by XGBoost, a machine learning routine. XGBoost is used to augment the computational speed of the maneuver optimization module of OpenCAMPT. The ensemble tree structure of the XGBoost is known to provide accurate estimates with limited amount of training data. This opens the possibilities of inexpensive onboard calculations of collision avoidance maneuvers and formation flying applications. Adding the contribution of the air drag to the relative encounter dynamics increases the accuracy of the initial estimates of the maneuver. Present work also proposes an approach which maps infinite possibilities of the orbits of the secondary object to a finite number of encounter geometries by defining 26 encounter geometries.
Research Interests:
This paper describes a method to calculate the optimal impulsive maneuver to avoid the collisions using Simplified General Perturbation 4 (SGP4) and Two Line Element (TLE). It also presents a rigorous analysis of the method to investigate... more
This paper describes a method to calculate the optimal impulsive maneuver to avoid the collisions using Simplified General Perturbation 4 (SGP4) and Two Line Element (TLE). It also presents a rigorous analysis of the method to investigate the relative dynamics of the two colliding space objects assuming the encounter is instantaneous. Different collision geometries are used for test cases. For all collision geometries, the relative velocity and the relative miss distance vectors are parallel, which is the worst case because the encounter occurs along the direction of the relative velocity vector. Test cases are satellites with distinctive orbital characteristics that are obtained from the official Spacetrack catalog. A precise numerical orbit propagator is used both to create ephemerides for test cases and to evaluate the accuracy of the proposed method. This work is significant because this is the first comprehensive investigation of the relative dynamics of the optimal impulsive collision avoidance maneuvers for satellites. Moreover, the proposed method isn't compute-intensive because it is a semi-numerical method.
Research Interests:
Her geçen gün bir çok yeni ülkenin uzaya uydu gönderme becerisi kazanmasıyla beraber özellikle de alçak irtifa uyduların yörüngelerindeki çarpışma riski artmaktadır. Uzay'da aktif uydular, roket parçaları ve pasif uydular, görevine devam... more
Her geçen gün bir çok yeni ülkenin uzaya uydu gönderme becerisi kazanmasıyla beraber özellikle de alçak irtifa uyduların yörüngelerindeki çarpışma riski artmaktadır. Uzay'da aktif uydular, roket parçaları ve pasif uydular, görevine devam edemediği için kontrol edilmeyen uydular, bulunmaktadır. Amerika Birleşik Devletleri Hava Kuvvetleri tüm bu cisimleri tanımlar, takip eder ve kataloglar. Bu işlemleri gerçekleştiriken SGP4 isimli analitik yörünge tahmin metodu ve dünya üzerinde farklı konumlarda bulunan optik teleskop ve radarlar aracılığıyla yapar. Aynı zamanda uzay ortamı farkındalığı için tüm aktif uydular için çarpışma analizi yapar ve daha önce tanımladığı kriterlere göre ilgili uydu sahiplerini çarpışma anıyla ilgili detayları göndermek suretiyle ikaz eder. Fakat, manevra kararı ve sorumluluğu tamamen uydu sahibindedir. Bu çalışma, yapılan ikaz metninde yer alan bilgileri kullanarak optimum manevra parametrelerini hesaplamayı amaçlar. Böylelikle verilen manevra kararı uyduyu muhtemel çarpışma alanından çıkarır ve bunu gerçekleştiriken mevcut kaynakların kullanımını minize eder ve görevin devamlılığına katkıda bulunur. SGP4 yörünge tahmin metodu ile diferansiyel doğrulama metodu ile aktif uydu çarpışma anından manevra kararı verilecek zamana geriye doğru yörünge ötelenir. Hesaplanan başlangıç noktasından itibaren hız vektöründe değişiklik yapılarak tekrar çarpışma anına ileri ötelenir ve çarpışma olasılığında yaratılan değişim gözlenir. Bu çalışma yer istasyonlarında görev alan uydu operatörleri için otonom optimum manevra tavsiyeleri yapabilir. Bunun yanında uydu sahipleri kendi uydularının hassas verilerini de kullanıp bu uygulamanın doğruluğunu arttırabilir. ABSTRACT Collision risk is increasing at Low-Earth Orbit (LEO) altitude for satellites as many new countries can build and launch satellites in orbit. There are active satellites, spent rocket stages and dead satellites in space. US Air Force (USAF) detects, tracks and catalogs these space objects. USAF uses radars and optical telescopes located at different locations on Earth and an analytical orbit determination method called Simplified General Perturbations (SGP4). Moreover, USAF calculates collision probabilities for all active satellites and warns satellite operators using predefined criteria for the sake of Space Situational Awareness (SSA). However, maneuver decision and responsibility belong to the satellite owners. In this paper, a method for calculating optimal maneuver direction, magnitude and time is investigated using the data within Conjunction Summary Message (CSM) published by USAF. Optimal maneuver decision not only avoids the collision by minimizing collision probability but also minimizes the fuel requirement. Because the smallest velocity requirement for the collision avoidance is calculated, the satellite operations will be minimally interrupted. Method is initiated by propagating the orbit of maneuverable satellite from close approach time to maneuver time. Optimal maneuver direction for velocity vector is calculated by defining the problem as a constraint optimization problem and the orbit is propagated to close approach time with the new velocity vector. The probability of collision is calculated for many different maneuver times and magnitudes is calculated and stored for analysis. Satellite controllers can use the method to calculate optimal maneuver for a given collision warning. In addition, satellite owners can use their precise orbit data to increase the accuracy of the method.
Research Interests:
Her geçen gün bir çok yeni ülkenin uzaya uydu gönderme becerisi kazanmasıyla beraber özellikle de alçak irtifa uyduların yörüngelerindeki çarpışma riski artmaktadır. Uzay'da aktif uydular, roket parçaları ve pasif uydular, görevine devam... more
Her geçen gün bir çok yeni ülkenin uzaya uydu gönderme becerisi kazanmasıyla beraber özellikle de alçak irtifa uyduların yörüngelerindeki çarpışma riski artmaktadır. Uzay'da aktif uydular, roket parçaları ve pasif uydular, görevine devam edemediği için kontrol edilmeyen uydular, bulunmaktadır. Amerika Birleşik Devletleri Hava Kuvvetleri tüm bu cisimleri tanımlar, takip eder ve kataloglar. Bu işlemleri gerçekleştiriken SGP4 isimli analitik yörünge tahmin metodu ve dünya üzerinde farklı konumlarda bulunan optik teleskop ve radarlar aracılığıyla yapar. Aynı zamanda uzay ortamı farkındalığı için tüm aktif uydular için çarpışma analizi yapar ve daha önce tanımladığı kriterlere göre ilgili uydu sahiplerini çarpışma anıyla ilgili detayları göndermek suretiyle ikaz eder. Fakat, manevra kararı ve sorumluluğu tamamen uydu sahibindedir. Bu çalışma, yapılan ikaz metninde yer alan bilgileri kullanarak optimum manevra parametrelerini hesaplamayı amaçlar. Böylelikle verilen manevra kararı uyduyu muhtemel çarpışma alanından çıkarır ve bunu gerçekleştiriken mevcut kaynakların kullanımını minize eder ve görevin devamlılığına katkıda bulunur. SGP4 yörünge tahmin metodu ile diferansiyel doğrulama metodu ile aktif uydu çarpışma anından manevra kararı verilecek zamana geriye doğru yörünge ötelenir. Hesaplanan başlangıç noktasından itibaren hız vektöründe değişiklik yapılarak tekrar çarpışma anına ileri ötelenir ve çarpışma olasılığında yaratılan değişim gözlenir. Bu çalışma yer istasyonlarında görev alan uydu operatörleri için otonom optimum manevra tavsiyeleri yapabilir. Bunun yanında uydu sahipleri kendi uydularının hassas verilerini de kullanıp bu uygulamanın doğruluğunu arttırabilir.
Research Interests:
A new method for orbit prediction, which is as accurate as numerical methods and as fast as analytical methods, in terms of computational time, is needed. Kolmogorov-Arnold- Moser (KAM) torus orbit prediction method is a modern orbit... more
A new method for orbit prediction, which is as
accurate as numerical methods and as fast as analytical methods,
in terms of computational time, is needed. Kolmogorov-Arnold-
Moser (KAM) torus orbit prediction method is a modern orbit
determination that can meet the aforementioned needs. This
paper presents a differential correction technique to create
parameters needed by the new theory and an approximate
accuracy analysis of the new orbit determination method by
using Simplified General Perturbations 4 (SGP4) and Two-Line
Element Set (TLE) as observational data.
accurate as numerical methods and as fast as analytical methods,
in terms of computational time, is needed. Kolmogorov-Arnold-
Moser (KAM) torus orbit prediction method is a modern orbit
determination that can meet the aforementioned needs. This
paper presents a differential correction technique to create
parameters needed by the new theory and an approximate
accuracy analysis of the new orbit determination method by
using Simplified General Perturbations 4 (SGP4) and Two-Line
Element Set (TLE) as observational data.
Identifying the type of an approaching aircraft, should it be a helicopter, a ̄ghter jet or a passenger plane, is an important task in both military and civilian practices. The task in question is normally done by using radar or RF... more
Identifying the type of an approaching aircraft, should it be a helicopter, a ̄ghter jet or a
passenger plane, is an important task in both military and civilian practices. The task in
question is normally done by using radar or RF signals. In this study, we suggest an alternative
method that introduces the use of a still image instead of RF or radar data. The image was
transformed to a binary black and white image, using a Matlab script which utilizes Image
Processing Toolbox commands of Matlab, in order to extract the necessary features. The
extracted image data of four di®erent types of aircraft was fed into a three-layered feed forward
arti ̄cial neural network for classi ̄cation. Satisfactory results were achieved as the rate of
successful classi ̄cation turned out to be 97% on average.
passenger plane, is an important task in both military and civilian practices. The task in
question is normally done by using radar or RF signals. In this study, we suggest an alternative
method that introduces the use of a still image instead of RF or radar data. The image was
transformed to a binary black and white image, using a Matlab script which utilizes Image
Processing Toolbox commands of Matlab, in order to extract the necessary features. The
extracted image data of four di®erent types of aircraft was fed into a three-layered feed forward
arti ̄cial neural network for classi ̄cation. Satisfactory results were achieved as the rate of
successful classi ̄cation turned out to be 97% on average.
Research Interests:
Surveillance of resident space objects (RSOs) is essential for detection, tracking, and cataloguing them to keep active satellites safe from hazards. High altitude space objects are observed using optical telescopes due to their... more
Surveillance of resident space objects (RSOs) is essential for detection, tracking, and cataloguing them to keep active satellites safe from hazards. High altitude space objects are observed using optical telescopes due to their efficiency. However, it is labour-intensive to detect space objects in the images captured by telescopes, and automation is desired. The proposed framework leverages the Feature Pyramid Network (FPN), a convolutional neural network for image segmentation, to automate RSO detection in the telescope images. The backbone used for detecting low-level patterns from images is the pre-trained EfficientNet-B7 on ImageNet. A simple preprocessing is applied to images that are overexposed to scale the input image pixel values, and this thresholding is only conducted using the statistics of the training data. A custom deterministic post-processing method based on vector mathematics is developed to clean the false detections. F1 score of the proposed machine learning framework is 92%, and this performance shows that the convolutional neural networks can be utilised for automating RSO detection from telescope images.
Research Interests:
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... more
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.