2020 Virtual undergraduate Research symposium
Task Allocation in Disconnected and Dynamic Systems
PROJECT NUMBER: 50
AUTHOR: Hayden Sather, Computer Science | MENTOR: Qi Han, Computer Science
ABSTRACT
—
The Distributed Hungarian Task Allocation Algorithm (DHTA) is an effective method of Task Allocation in a distributed Multi-Agent team. The Request And Share Then Assign (RASTA) distributed task allocation algorithm greatly improves on DHTA by reducing the number of messages required to find an optimal solution significantly. However, both of these algorithms are designed to work in a static and connected environment which is unlike a real-world scenario. My work applies the RASTA algorithm to a dynamic and disconnected environment and analyses the efficiency of multiple different techniques to do so. I improve upon the original RASTA algorithm by reducing the dissemination of unneeded information between the agents. I analyze the effect of multiple different methods of information dissemination on the efficiency and performance of the system. The conclusion which I came to is that to a certain point, the amount of information shared in the system can be reduced without affecting the system’s ability to come to an optimal configuration. However, if there isn’t enough information shared then conflicts arise where two agents assign themselves to the same task when only one is needed, or agents travel a longer distance than optimal. I provide several techniques of information dissemination based on RASTA and explore the advantages and disadvantages of each.
VISUAL PRESENTATION
—
AUTHOR BIOGRAPHY
—
Hayden is a Senior in the Honor’s Research Computer Science program. He is also a student in the BS + MS CS program and a MURF recipient. He is an athlete on the track team and competes in the decathlon. He has completed research with Dr. Yang in embedded systems, Mr. Culbreth in business analytics, and Dr. Han in IoT smart irrigation, and now task allocation in Multi-Agent teams. This research project expands on an existing task allocation algorithm and applies it towards a disconnected and dynamic system. Hayden wants to focus on Machine Learning and Computer vision in the future, especially for the application of self-driving cars.
Very interesting work, Hayden. I would have loved to see your presentation today. Good work.