Advanced Intelligence Research Center: Autonomous systems research

 

Study on autonomous technology in generic model and JAXA’s test facilities

Description:
Before going to the spacecraft in which reliability and computational cost are severe, those autonomous technologies should be demonstrated in generic spacecraft model and JAXA’s test facilities. Engine test stands at Kakuda Space Center, and Low Speed Wind Tunnel (LWT1) at Chofu Aerospace Center are candidates for this study. There are also requirements for improving the efficiency of maintenance and operation of test facilities in JAXA, so the technologies obtained in this study are useful not only for the autonomous spacecraft but also for test facilities. Branches for operating those facilities in JAXA are interested in this study. Generic spacecraft model is also employed for demonstrating autonomous health management technique. It is a fact that there is no appropriate metrics for measuring the robustness of autonomous technology based on machine learning for critical system. We have organized the metrics through literature reviews and try them in this research topic.
Objectives:
1) Demonstration of up-to-date autonomous technology   2) Event-driven processing to detect trigger conditions, such as science events, “interesting”    features, and changes relative to previous observations   3) Generate novel command sequences and repairs.

 

Definition of levels of autonomy in spacecraft

Description:
This topic focuses on the defining the LoA in spacecraft. Unlike autonomous cars, there is no metric to measure autonomous level of spacecraft. Such metric will help us set goals for our study. The aim of this study is to define LoA based on the review of autonomous car, drone and aircraft, and robotics, and knowledge of satellite development so far. It will be nice if the LoA defined here becomes a de facto standard in the world.
Objectives:
・ Introduce LoA in spacecraft

 

Study on computational performance of state estimation model in FPGA

Description:
Due to the increasing demands of onboard sensor and autonomous processing, one of the principal needs and challenges for future spacecraft is onboard computing. An FPGA is a candidate to be used in the spacecraft, and the performance of the reduced-order physical simulation and autonomous technique should be investigated to realize autonomous technology in spacecraft. This study aims to get insights on computer resources required. Preliminary will be conducted as a joint work with Toyohashi University of Technology (TUT) and RIKEN.   In addition, direction of new wave of AI technology is surveyed in this topic. Candidates at this moment is neuromorphic computing based on spiking neural network which is considered to achieve computation with very low power like the human brain (only 20W!).
Objectives:
・ Investigate if the technique such as Gaussian Processing Regression have the potential    to be an effective tool for the development of digital twins on FPGA.   ・ Evaluate hardware requirements for physical implementation   ・ Survey new wave of AI technology