A Hybrid Method for Distributed Multi-Agent Mission Planning System
Abstract
This thesis presents work concerning a distributed heterogeneous multi-agent robotic team for outdoor applications such as search and rescue or surveillance. The goal of this research is to develop a method of control for a team of unmanned aerial and ground robots that is resilient, robust, and scalable given both complete and incomplete information about the environment. The method developed and presented in this paper integrates approximate and optimal methods of path planning integrated with a market-based task allocation strategy.This thesis also presents a solution to unmanned ground vehicle path planning within the developed mission planning system framework under incomplete information. Methods such as genetic algorithm and deep reinforcement learning are proposed to solve movement through unknown terrain environment. The final demonstration for Advantage-Actor Critic deep reinforcement learning model elicits successful implementation of the proposed model.
Degree
M.Sc.
Advisors
Mou, Purdue University.
Subject Area
Robotics|Artificial intelligence|Aerospace engineering|Operations research|Pedagogy|Public administration|Transportation
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.