Background

Within the past two decades, the number of resident space objects (RSOs - artificial objects that are in orbit around the Earth) has nearly doubled, from around 11000 objects in the year 2000 to around 19500 objects in 2019. This number is expected to rise even higher as more satellites are put into space, thanks to improvements in satellite technology and lower costs of production. On the other hand, the increase in the number of RSOs also indirectly increases the risk of collision between them. The important issue here is the reliable and accurate orbit tracking of satellites over sufficiently long periods of time.

Failure to address this issue has led to incidents such as the collision between the active US Iridium-33 communication satellite, and the inactive Russian Kosmos-2251 communication satellite in February 2009. In fact, this accident increased the amount of space debris by 13%, as shown in the figure below:

More accidents will result in more debris being produced, and through a chain reaction of collisions (if left unchecked), may lead to a dire situation in which it becomes difficult or downright impossible to put a satellite into orbit due to the large accumulation of space debris surrounding the Earth. This scenario is known as the Kessler Syndrome. Thus, considering the gravity of the situation at hand, it is imperative to prevent such catastrophic collisions from ever happening again.

Context

The aim is to use machine learning or other forecasting algorithms to predict the positions and speeds of 600 satellites in orbit around the Earth. The original datasets were obtained from the International Data Analytics Olympiad 2020 (IDAO 2020) Competition, provided by the Russian Astronomical Science Centre. Being a competition, IDAO does not provide the answer key for their test dataset. However, I prepared their train dataset in such a way that testing one's algorithm can easily be done via the answer_key. This preparation is outlined in the notebook associated with this dataset ("Data Preparation to Produce Derived Datasets").

Satellite positions and speeds (henceforth, they will be collectively referred to as the "kinematic states") can be measured using different methods, including simulations. In this dataset, there are two kinds of simulators: the precise simulator and the imprecise simulator. We refer to measurements made using the precise simulator as the "true" kinematic states of the satellite and measurements made using the imprecise simulator as the "simulated" kinematic states.

The aim is to make predictions for the true kinematic states of 600 satellites in the final 7 days of January 2014.

Datasets Description

• The jan_train dataset contains data on both true and simulated kinematic states of 600 satellites for the 24-day period of time from 01-Jan-2014 00:00 to 24-Jan-2014 23:59.
• The jan_test dataset contains data on only the simulated kinematic states of the same 600 satellites for the final 7-day period of time from 25-Jan-2014 00:00 to 30-Jan-2014 23:59.
• The answer_key dataset contains data on only the true kinematic states of the same 600 satellites for the final 7-day period of time from 25-Jan-2014 00:00 to 30-Jan-2014 23:59. NOTE: The answer_key should only be used for the purpose of evaluation, NOT for training.

Variables Description

• id (integer): unique row identifier
• epoch (datetime): datetime (at the instant of measurement) in "%Y-%m-%d %H:%M:%S.%f" format (e.g. 2014-01-27 18:28:18.284)
• sat_id (integer): unique satellite identifier, ranging from 0 to 599
• x, y, z (float): the true position coordinates of a satellite (km)
• Vx, Vy, Vz (float): the true speeds of a satellite, measured along the respective axes (km/s)
• x_sim, y_sim, z_sim (float): the simulated position coordinates of a satellite (km)
• Vx_sim, Vy_sim, Vz_sim (float): the simulated speeds of a satellite, measured along the respective axes (km/s)

Tasks

1. Start by exploring the dataset as a whole, getting familiar with the data contained within.
2. Choose one satellite to explore in more detail, looking at both true and simulated kinematic states
3. Look at the periodicity and pattern of each kinematic state, and think about whether or not they will be useful in a predictive model.
4. Develop a machine learning model to forecast over the final 7 days for each kinematic state of the chosen satellite.
5. Repeat the process for the remaining 599 satellites, keeping in mind that there could be differences between them. Otherwise, choose the satellites that are most interesting or those that are quite different from each other.

Acknowledgements

All ownership belongs to the team, experts and organizers of the IDAO 2020 competition and the Russian Astronomical Science Centre.

Original Owner of Datasets (IDAO 2020): https://idao.world/ License (Yandex Disk): https://yandex.com/legal/disk_termsofuse/ Database (where original data is stored): https://yadi.sk/d/0zYx00gSraxZ3w