Research

My research interests include multi-robot coordination, robot programming systems, and the use of robots in science and engineering education. Current and past research projects are given below. Look here for my publications.

Current projects (see below for past projects):

Leaving Flatland: Fast Autonomous Navigation For UGVs in Challenging Terrain

We envision a new research program on autonomous modeling and navigation, with the goal of enabling small unmanned ground vehicles (UGVs) to move alongside their human counterparts through virtually any environment. Such UGVs will operate indoor and outdoor, with no distinction, and at a speed compatible with human motion. This is a joint project with Stanford University and Boston Dynamics.

Player

The Player Project develops Open Source robot control and simulation software. The primary products of this project are Player, a robot device interface that provides a powerful, flexible, language- and platform-neutral interface to a variety of sensors and actuators; and Stage and Gazebo, two highly parameterizable sensor-based multiple robot simulators. Player, Stage, and Gazebo are widely used in labs and classrooms around the world. I am a founding developer on the project and I head development of Player.

Learning Applied to Ground Robots (LAGR)

We are developing novel algorithms for high-speed vision-based navigation in outdoor environments. We are focusing on: real-time visual odometry; accurate obstacle detection and map construction; learning for long-range sensing and path detection; globally optimal path-planning; and high-speed trajectory generation with dynamics. We periodically ship our code to be tested against the other LAGR teams by government evaluators. They run our system on their version of the LAGR robot in environments that we've never seen.

Update: We placed 1st in the final 2 tests of Phase I !

Past projects (see above for current projects):

	Commbots We seek algorithms for the distributed control of small, radio-equipped mobile robots that form a multi-hop ad hoc communcation network. The task of the Commbots is to self-organize physically in order to maximize network service in indoor environments. To achieve this task, the robots must adapt quickly and effectively to changing wireless environments and service demands.
	Pursuit-evasion We study a form of the pursuit-evasion problem, in which one or more searchers must move through a given environment so as to guarantee detection of any and all evaders, which can move arbitrarily fast. We introduce a new class of searcher, with limited field of view, which can be readily instantiated as a physical mobile robot. We have established the complexity of this problem and have developed the first complete search algorithm for a single searcher.
	Fundamentals of multi-robot task allocation Important theoretical aspects of mechanisms for multi-robot task allocation have, to date, been largely ignored. In this project, we are trying to address part of this negligence by formally studying the problem within an organizational framework developed in the Operations Research community. In particular, we are currently exploring multi-robot task allocation as an instance of the well-known Optimal Assignment Problem. In this light, we have recently analyzed and compared the algorithmic characteristics of several existing approaches to the problem.
	Auction-based Multi-Robot Coordination The key to utilizing the potential of multi-robot systems is coordination. In this project, we are exploring economically-inspired approaches to achieving robust multi-robot coordination. In particular, we have developed MURDOCH, a highly-scalable, distributed, auction-based multi-robot coordination system. A variant of the well-known Contract Net Protocol, MURDOCH has been experimentally validated in a variety of task domains with physical robots.
	Mathematical Modeling of Multi-Agent Systems (Led by Kristina Lerman) Our research goals are two-fold: show that a distributed mechanism based on purely local interactions can lead to the desired group behavior in several different agent-based systems; model and analyze these systems mathematically.
	Hi-Scale User-System Interaction (HiSUSI) (Led by Ashley Tews) This project is concerned with investigating the high-level interaction dynamics between people and multiple robotic and embedded systems. The key question is how to connect possibly hundreds of users to hundreds of systems and maintain personal interaction. An interaction infrastructure has been developed for this purpose that allows interaction at both extremes. This project is part of the larger Human-System Interaction project at USC Robotics.

Last updated 20 November 2007 17:36:19