🚀 go-pugleaf

Joint Seminar with Mechanical Engineering:
Sensor Based Planning for Coverage and Serpentine Robotic Tasks

Howie Choset
Carnegie Mellon University


SPECIAL PLACE AND TIME

125 Scaife Hall, Refreshments in 322 Scaife Hall
Refreshments 3:00 pm
Talk 3:30 pm


Abstract

Whereas conventional path planning determines a path between two points,
coverage specifically emphasizes the volume swept along the path by the
robot, its detector range, and other effectors. The approach we take to
coverage divides the environment into cells and forms a graph encoding the
adjacent relationships (topology) among cells. Complete coverage of a
target region is achieved by visiting each cell in the decomposition,
i.e., finding a walk through the adjacency graph. Coverage of an unknown
space reduces to dynamically covering each cell while simultaneously
determining the adjacency graph; this enables floor cleaning and
inspection of unknown floor spaces. Our current research involves
extending our work in planar mobile robot coverage to robotic painting.
The challenges here include lifting coverage into three dimensions and
ensuring uniform coverage of a target region, which is imperative for high
quality and low cost operations.

In many situations, such as robotic de-mining, time may not permit
covering a target environment completely. However, if the planner has
access to a probabilistic map of mine locations, it can opportunistically
guide the robot. For mine fields that have been set down in a pattern,
current work involves developing a Bayesian method for efficiently
decoding the parameters that describe the field. Once these are known, the
robot can cover a small fraction of the target region and locate most of
the mines.

Finally, we will briefly discuss our work with hyper-redundant robots
which include serpentine robots and actuator arrays. Serpentine robots
contain many internal degrees-of-freedom, that enable them to thread
through tightly packed volumes and to access locations unreachable by
people and conventional machines. To achieve purposeful motion, these many
degrees of freedom must be coordinated. We also will propose a new
serpentine design that has excellent flexibility and good mechanical
advantage per unit cross-section diameter.


Speaker Biography

Howie Choset was born and raised in New York. After receiving
undergraduate degrees, one in Computer Science and the other in Business
from the University of Pennsylvania, he ventured off to the California
Institute of Technology where he received his MS and PhD in Mechanical
Engineering. Currently, he is an Assistant Professor in Mechanical
Engineering and holds a courtesy appointment in Robotics. His research
interests include design and motion planning for hyper-redundant
mechanisms, simultaneous mapping and localization, and coverage, with
applications towards de-mining and painting. Howie Choset also directs the
undergraduate robotics minor at Carnegie Mellon and has recently
introduced a new undergraduate course on robotics that utilizes the Lego
Mindstorms.

Speaker Appointments

For appointments, please contact the speaker, Howie Choset, at
choset@ri.cmu.edu.