Yuta Nakajima

SDS-4 Attitude Control System: In-Flight Results of Three Axis Attitude Control for Small Satellites

Abstract The Small Demonstration Satellite (SDS) program is a JAXA technology demonstration program, targeting the in-orbit demonstration of advanced technologies using small satellites. Following the success of SDS-1, the SDS-4 spacecraft was successfully launched on 18 May 2012 and continues the in-flight demonstration of a space-based Automatic Identification System (AIS) as well as numerous flight experiments. This paper describes the in-flight results of the Attitude Control System (ACS) for 50kg class small satellites using 3-axis zero-momentum stabilization.

A Data-Driven Health Monitoring Method for Satellite Housekeeping Data Based on Probabilistic Clustering and Dimensionality Reduction

In the operation of artificial satellites, it is very important to monitor the health status of the systems and detect any symptoms of anomalies in the housekeeping data as soon as possible. Recently, the data-driven approach to the system monitoring problem, in which statistical machine learning techniques are applied to the large amount of measurement data collected in the past, has attracted considerable attention. In this paper, we propose a new data-driven health monitoring and anomaly detection method for artificial satellites based on probabilistic dimensionality reduction and clustering, taking into consideration the miscellaneous characteristics of the spacecraft housekeeping data. We applied our method to the telemetry data of the small demonstration satellite 4 (SDS-4) of the Japan Aerospace Exploration Agency (JAXA) and evaluated its effectiveness. The results show that the proposed system provides satellite operators with valuable information for understanding the health status of the system and inferring the causes of anomalies.

Semantic classification of spacecraft's status: integrating system intelligence and human knowledge

In this paper, we introduce a novel approach where the system involves human knowledge in the classification task using decision trees. Machine learning techniques are now applied to a variety of tasks in real-world problems. The computer performs complex computations better than humans. However, in many real-world applications, humans have background domain knowledge about the problem that the computer often does not have. For instance, in a spacecraft status classification task, humans have a sense for which factors are likely to correlate with the classes of interest. Without this knowledge, machines may overfit to training data. We propose to combine two models: one based on human reasoning, common sense, or heuristics, and the other learned by a machine learning algorithm in a data-driven manner. In our experiments, we use decision trees and categorical features so that the model consists of rules which are semantic and interpretable for humans. Our proposed approach results in an improvement in classification performance over either models alone. Our work illustrates the possibility of integrating human knowledge and artificial intelligence.

Dynamic Grouped Mixture Models for Intermittent Multivariate Sensor Data

For secure and efficient operation of engineering systems, it is of great importance to watch daily logs generated by them, which mainly consist of multivariate time-series obtained with many sensors. This work focuses on challenges in practical analyses of those sensor data: temporal unevenness and sparseness. To handle the unevenly and sparsely spaced multivariate time-series, this work presents a novel method, which roughly models temporal information that still remains in the data. The proposed model is a mixture model with dynamic hierarchical structure that considers dependency between temporally close batches of observations, instead of every single observation. We conducted experiments with synthetic and real dataset, and confirmed validity of the proposed model quantitatively and qualitatively.

Receding-horizon unscented kalman filter using successive unscented transformation for spacecraft attitude estimation

A new attitude estimation method for a spacecraft is derived. This method employs a receding-horizon strategy to use a time window and constraint. A conventional constrained filter, the receding-horizon nonlinear Kalman filter (RNKF), propagates the state value in the prediction step, and minimizes the cost function with a constraint in the filtering step. It is desirable for the optimization to be a quadratic programming (QP) problem, whose constraint is linear, in terms of computational complexity. If the RNKF is applied to the attitude estimation problem, the appropriate attitude representation is the quaternion, which has no singular point, in the prediction step. However, the quaternion does not define a QP problem in the filtering step because the quaternion needs to satisfy a single constraint of a unit norm. Therefore, this paper proposes the receding-horizon unscented Kalman filter (RUKF), which is an improvement of the RNKF, to deal with appropriate attitude representation in each step. In the RUKF, each attitude of a time window is represented by generalized Rodrigues parameters (GRPs) in the filtering step employing the successive unscented transformation. The GRPs is an attitude representation with no constraint. Simulation revealed that the RUKF is more accurate than the extended Kalman filter.