Unsupervised Change Detection for Space Habitats Using 3D Point Clouds (2312.02396v3)

Abstract: This work presents an algorithm for scene change detection from point clouds to enable autonomous robotic caretaking in future space habitats. Autonomous robotic systems will help maintain future deep-space habitats, such as the Gateway space station, which will be uncrewed for extended periods. Existing scene analysis software used on the International Space Station (ISS) relies on manually-labeled images for detecting changes. In contrast, the algorithm presented in this work uses raw, unlabeled point clouds as inputs. The algorithm first applies modified Expectation-Maximization Gaussian Mixture Model (GMM) clustering to two input point clouds. It then performs change detection by comparing the GMMs using the Earth Mover's Distance. The algorithm is validated quantitatively and qualitatively using a test dataset collected by an Astrobee robot in the NASA Ames Granite Lab comprising single frame depth images taken directly by Astrobee and full-scene reconstructed maps built with RGB-D and pose data from Astrobee. The runtimes of the approach are also analyzed in depth. The source code is publicly released to promote further development.

• The paper presents an unsupervised change detection algorithm that leverages unlabeled 3D point clouds to autonomously identify alterations in space habitats.
• It employs a two-phase approach using a modified GMM clustering and Earth Mover’s Distance to effectively compare data from different time points.
• Evaluation on NASA Ames data shows robust detection capabilities, despite challenges with object appearances and boundary false positives.

Introduction

The development of autonomous robotic systems for the maintenance of deep-space habitats, such as those envisioned for the Gateway space station, is a critical component for future space exploration. The International Space Station currently utilizes scene analysis software that relies on a human-intensive process of manually labeling images to detect changes within the habitat. However, the paper introduces an innovative change detection algorithm that leverages raw, unlabeled 3D point clouds, thereby enhancing the capabilities of robot caretakers in identifying changes without human intervention.

Methodology

At the core of the proposed change detection method is a two-phase process. The first phase involves using a modified Expectation-Maximization Gaussian Mixture Model (GMM) clustering technique to create probabilistic models from input point cloud data. Then, the second phase employs the Earth Mover’s Distance (EMD) to compare the GMM clusters from two different points in time to pinpoint areas of change.

The veracity of the algorithm is put to the test using data from NASA Ames Granite Lab, where an Astrobee robot – a research platform on the ISS – gathers depth images and constructs full-scene 3D maps using additional RGB-D and pose data. Distinctively, the source code for the algorithm has been made publicly available, promoting open research and development in this field.

Results and Performance

Upon evaluation, the algorithm shows a promising capacity to detect multiple object changes, demonstrated through both single-frame point clouds and full-scene reconstructed map data. To further assess the fidelity and computational efficiency of the method, modifications such as varying the initial number of Gaussian clusters and utilizing Principal Component Analysis for pre-processing the data were explored. These variations underscore the balance between detection precision and algorithm runtimes.

Conclusion and Future Directions

Despite showing potential for unsupervised change detection in space habitats, current limitations of the algorithm include its inability to simultaneously detect object appearances and disappearances and the propensity for false positives at the boundaries of scenes. Ongoing development aims to incorporate semantic understanding into the architecture and apply the algorithm to ISS data to handle the complex environment.

Moving forward, the continual advancement of anomaly detection algorithms will equip robotic assistants to autonomously caretake off-world facilities, supporting NASA's ambitions for sustainable extraterrestrial presence and exploration.