ADITYA MAHADEVAN

HRBB 407

MENTOR: DR. NANCY AMATO

SUPERVISOR: PHILLIP COLEMAN

DEPT: ECEN

PROGRAM: CSE REU




Research Topic

Implementing efficient strategies for searching areas using multiple agents


Description

This research aims to develop and implement robust, efficient strategies for using a group of agents to find targets, objects or other agents located in two- or three-dimensional spaces filled with obstacles of different shapes, sizes and complexities.


Purpose

I chose this research topic because it is relevant to multi-agent robotics and robotic group behaviors. There is a need to better direct multiple agents as one group rather than individually. The behaviors available at present tend to treat agents of a group as individuals rather than as a collective entity. This research will attempt to address this and exploit the entire group so that a task such as area searching can be completed according to the specific requirements of different scenarios. For instance, one strategy might involve finding a stationary object as quickly as possible, with minimal overlap in the area covered by different robots. On the other hand, another strategy might require a large overlap so that an area is searched more thoroughly.


Project Goals and Implications

I expect that this project will lead to a good understanding of how to better direct large groups of agents to accomplish suitable tasks. The results from this could be scaled to included hundreds or thousands of agents in situations such as large-scale farming or traffic control. In addition, since the agents are desired to behave emergently, i.e. taking into account local conditions and variables, the results of this research could be applied not only to robots but to humans as well. One obvious real-world application of this is in simulating bomb-disposal teams looking for a bomb in a building, or police officers pursuing a fugitive.


Personal Goals

This summer research will get me acquainted with the software and code used by the Group Behaviors group in the Parasol lab so that I can work on an Honors thesis for my final undergraduate year. I see both this summer and the coming academic year as a stepping stone to higher education and further research. I hope it will help me get into a relevant and interesting graduate program once I graduate.


Approach

I will generate several group-based area search strategies based on prior research and techniques borrowed from roadmap-based path planning. I will then modify the C++ based behavioral framework to include these strategies as functions called by one or more agents or the entire group. Finally, I will simulate agents in different spaces with different obstacles to evaluate the effectiveness of these search strategies. The obstacles and spaces chosen will simulate real-world situations, such as a floor in a building containing a bomb. I aim to have spaces with a variety of different numbers and types of objects, so that a range of strategies can be applied.


The criteria for evaluating the strategies will include:

These evaluations will be done by calling evaluation functions incorporated into the behavioral framework. Each criterion will affect the evaluation differently depending on the specific requirements of each strategy.


Methods

I will incorporate the rule-based behavioral framework developed by the Group Behaviors group and the PRM functions developed by the path planning group in programming the agents using the C++ language. I will use a UNIX environment to program and simulate the agents.


Work Schedule

I am expected to be physically present in the Parasol lab from 10am-6pm except when attending seminars or organized events, but I am also free to continue working on the project outside these times.


Deliverables and Due Dates


June 10th – Research plan and basic personal website

July 22nd – Research abstract

July 16th, 23rd, 30th – Student presentations

August 3rd – Research poster

August 5th – Research paper and complete personal website