From Santa Fe Institute Events Wiki
I am Roozbeh, a second year PhD student in Texas A&M University, department of Electrical and Computer Engineering.
My research interests are:
- Complex Systems
- Self Organization
- Multi Agent/Multi Robot Systems
My MS thesis was about “Using Environment as a Means of Communication in Distributed Systems”. In this research we considered the methods that the agents in an environment use for implicit communication. This led our research to systems with high number of agents that have limited capabilities. They had uncertainties in their sensory inputs and had bounded rationality in their decision makings. We developed some approaches for overcoming these limitations. As an example, we showed that how a set of virtual springs connecting the agents together could result in robust group movements. Using RoboCup benchmark, we also showed that when we have limited amount of processing resources, our agents perform better than sophisticated agents. We also set some experiments with multiple light weight robots and observed the group behaviors.
In a part of my PhD studies, we studied some naturally occurring behaviors in groups of individuals. We modeled and simulated the behaviors of the individuals and observed the group behaviors that emerged from their interactions. We inspired from some natural behavior like foraging, pursuing and patrolling.
My research interests are in the areas of Complex Systems and Emergence. In one hand, I am interested in studying the ubiquitous characteristics behind all these systems from a theoretical perspective. On the other hand, I want to exploit my knowledge about complexity for different applications. One of them is a robotics application that we have a large group of mobile robots with limited capabilities and we want to make some desired higher level group behaviors. Another sample application is controlling a crowd of people with individual level controls.
Currently we are studying a system that contains multiple robots with limited capabilities. Our goal is to raise a group level intelligence in the group. We have created a vector field to guide the robot in the environment. This vector field is made by blending a set of linear dynamical systems so that as a whole, it shows the desired non-linear behavior. Each linear system has an equilibrium, which is be a stable or unstable and is a node or a focus. The blending function is a differentiable combination of these linear functions that guides the agent in the environment. In this system the obstacles are presented with unstable nodes or foci and targets are presented with stable nodes or foci (the choice of node or focus depends on the type). The agents are simple and limited, but they can still achieve a set of goals.