In collaboration with: Leen-Kiat Soh
Meet with Soh and me before submitting a proposal on this
project
A multiagent system consists of autonomous agents that are capable of sensing the environment, making decisions, and carrying out actions which in turn modify the environment. The agents can be cooperative or competitive. In a cooperative multiagent system, it is possible for agents to form ad hoc teams to solve a particular task without pre-coordination and then disband after the task is solved. Because of uncertainty and dynamic changes in the environment, working as a team without pre-coordination may lead to having incompatible teammates, resulting in inefficient or ineffective task solutions. Thus, it is important for an agent to model its environment, including task characteristics and agent behaviors. To do so, agents turn to reinforcement learning to learn the utilities of performing certain actions or decisions given certain situations. These learned utilities subsequently guides agents in making decisions. This project will make use of an existing multiagent system simulation and involve Java programming to implement solutions and carrying out large-scale experiments.
Return to the CSCE 478/878 Home Page