What is it about?
Safe autonomous navigation is a challenging problem, especially for unmanned aerial systems. Reachable sets provide an alternative to safe control of autonomous vehicles, but are challenging to compute themselves. We use machine learning to compute the reachable sets for safe autonomous control.
Featured Image
Photo by Dose Media on Unsplash
Why is it important?
Safe control of autonomous vehicles (esp. unmanned aerial systems (UAS)) is particularly challenging due to: 1) safety critical operating scenarios, 2) public perception and demands of safety from autonomous systems, 3) black-box techniques applied to control of UASs. As a result, control methods that allow embedding safety requirements into learning-based control design schemes are of importance. Primarily because it is difficult to impose (fuzzy) safety constraints into learning-based control design schemes. Proposed method presents one way to do so.
Perspectives
Using learning-based methods for safety critical control of autonomous vehicles is a very challenging, but inevitable task as learning-based methods are finding applications in safety critical operating environments. The proposed method is a promising exploration into imposing safety requirements of a specific kind into safe autonomous control.
Omanshu Thapliyal
Hitachi America Ltd.
Read the Original
This page is a summary of: Embedding Safety Requirements into Learning-Based Controllers for Urban Air Mobility Applications, January 2024, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2024-2395.
You can read the full text:
Contributors
The following have contributed to this page
