Collision Avoidance Challenge dataset

Collision Avoidance Challenge 数据集是 ESA 为“Collision Avoidance Challenge”举办的 Kelvins 竞赛期间使用的官方数据集。该数据集是 ESA 从 2015 年到 2019 年收到的联合数据消息 (CDM) 的集合。CDM 已匿名分发。最初的原始数据以及在比赛期间保密的标签也被发布。
欧空局感谢美国太空监视网络提供支持欧空局航天器安全运行的监视数据。此外，我们感谢允许公开发布当前数据集的协议。
数据集表示为一个表，其中每一行对应一个 CDM，每个 CDM 包含 103 个记录的特征/特征。因此有 103 列，这些列在竞赛页面中进行了描述。该数据集由几个独特的碰撞/接近事件组成，这些事件在 event_id 列中标识。反过来，每个碰撞事件都由随时间记录的多个 CDM 组成。因此，可以将单个碰撞事件视为 CDM 的时间序列。从这些 CDM 中，对于每个碰撞事件，我们有兴趣预测在时间序列的最后一个 CDM 中计算的最终风险（即每个碰撞事件最后一行的风险值）。

The Collision Avoidance Challenge dataset is the official dataset employed by the European Space Agency (ESA) during the Kelvins Competition held for the "Collision Avoidance Challenge". This dataset comprises a collection of Coordinated Data Messages (CDMs) received by ESA between 2015 and 2019. The CDMs have been anonymized prior to distribution. Both the original raw data and the confidential labels used throughout the competition have also been publicly released.

ESA extends its gratitude to the U.S. Space Surveillance Network for providing surveillance data that supports the safe operation of ESA spacecraft. Furthermore, we thank the relevant agreements that enabled the public release of this dataset.

The dataset is structured as a table, where each row corresponds to one CDM, and each CDM contains 103 recorded features/attributes. Accordingly, the table has 103 columns, which are detailed on the competition's official webpage. This dataset encompasses multiple distinct collision or close approach events, which are identified using the event_id column. Each collision event, in turn, consists of multiple CDMs recorded over a period of time. As such, a single collision event can be regarded as a time series of CDMs. For each collision event, the objective is to predict the final risk calculated in the last CDM of the time series, i.e., the risk value present in the last row of each collision event.

搜集汇总

背景与挑战

背景概述

Collision Avoidance Challenge dataset是ESA提供的官方数据集，包含2015-2019年的匿名联合数据消息(CDM)，以表格形式呈现，每行对应一个CDM，共103列特征，用于预测碰撞事件的最终风险。数据集由多个独特的碰撞事件组成，每个事件包含随时间记录的多个CDM，形成时间序列数据。