Semester of Graduation
Master of Science in Mechanical Engineering (MSME)
Mechanical and Industrial Engineering
Multi-agent systems have potential applications in a multitude of fields due to their ability to perform complex tasks with a high degree of robustness. Decentralized systems are multi-agent systems where sensing, control, and communication responsibilities are distributed amongst the agents giving this scheme a particularly high potential for robustness, adaptability, and complexity; however, this comes with its own list of challenges. Formation control is a topic in multi-agent systems that aims to have the agents moving as a virtual rigid body through space. In the literature, there have been successful implementation of formation control algorithms on robot agents in both the 2D and 3D case; however, these have all been done using a centralized setup.
The aim of this project was to create a 3D decentralized platform for testing formation controllers on mini quadcopters. This was done using three Parrot Mambo quadcopters and Simulink/MATLAB for the control and communication programming as well as simulation. The distance-based formation control scheme, which is inherently decentralized, was used to test the platform. Implementation of formation acquisition was shown in simulation and real-life testing with perfect acquisition shown in simulation and up to 10% steady state error on the hardware. This error in the real-life test is likely attributable to noisy and inaccurate sensors, communications delays, and unmodeled dynamics. The platform was successfully created and demonstrated and could serve as a testbed for further research. Overall, the hardware results demonstrate working but potentially improvable results.
Snellgrove, Matthew, "A Multi-Aerial Vehicle Platform for Testing Collaborative Mobility Formation Controllers" (2023). LSU Master's Theses. 5836.
De Queiroz, Marcio