Unmanned Aerial Vehicle Fault Tolerant Control Test Bed
Brief Description of Project:
This summer research work is linked to a NASA System-Wide Safety Assurance project funded by NASA from February 2021. Our group has been developing online risk analysis and fault-tolerant control mechanisms to determine UAV flight safety, which can be impacted by weather conditions, collision with other flying objects or static obstacles, and faults and degradation in the UAV itself.
The focus of this project will be to continue the development of a testbed in the Gazebo environment to facilitate and the testing, validation, and visualization of the performance of our fault-tolerant control algorithms in a simulation environment. The student will work closely with a Research Scientist and Graduate students working on this project to support the development of the Gazebo environment, develop octocopter UAV models, and implement fault-tolerant control, and replanning algorithms to demonstrate the effectiveness and to prepare for future flight tests at NASA Langley.
The student should have completed CS 3251 - Intermediate Software Design, and CS 3250 - Algorithms.
Students who have taken CS 4260: Undergraduate AI and/or CS 4262: Undergrad Machine Learning will be preferred
Nature of Supervision:
The faculty mentor, Gautam Biswas, will mentor the student introducing him/her to mathematical models and reinforcement learning methods for fault-tolerant control. He will provide the background reading and the papers the student needs to get started on the project.
The student will work very closely with the Research Scientist and Graduate Students to understand the Gazebo environment and the current status of the implementation of the testbed for UAV safety analysis.
The PIs will meet with the student at least once a week to review progress, provide guidance, and set directions for subsequent research and development work.
The student will attend all research meetings, present his progress at these meetings, and use the feedback received to build on his understanding of the research and development work.
A Brief Research Plan (period is for 10 weeks):
Weeks 1 and 2: The student will review background material and research papers to become familiar with the research topic. The student will also spend time with the graduate student to become familiar with the Gazebo environment and the current development status of the simulation environment.
Weeks 3 and 4: Work closely with the Research Scientist and Graduate to update and refine the existing UAV (octocopter) and environment models as required; Develop a detailed understanding of the Reinforcement Learning (RL) algorithms, and work to integrate the algorithms implemented in Open AI Gym into the Gazebo framework.
Weeks 5 and 6: Extensive testing of the Fault Tolerant Control algorithm under different scenarios, and develop visualizations in Gazebo to illustrate the significant characteristics of the fault-tolerant control algorithm.
Weeks 7 and 8: Gain an understanding of the replanning algorithms to ensure a safe flight, and implement them in the Gazebo environment. Develop scenarios to test the replanning algorithm, and illustrate using visualizations in the Gazebo environment.
Weeks 9 and 10: Extensive testing of the developed testbed environment and report writing.
Number of Open Slots: 1
Name: Gautam Biswas
Department: Computer Science