Speaker: Dr. Nare Karapetyan
From: University of Maryland
Abstract
In the pursuit of understanding the world around us, significant efforts are dedicated to
developing dependable autonomous systems capable of planning out exploration tasks. These
tasks encompass a range of applications, such as scientific sampling, environmental monitoring,
surveillance, and search and rescue operations. Robotic systems deployed in real-world
scenarios must take into account the dynamic changes and uncertainties of their environments,
which exponentially increase the decision state space and make these problems intractable. As
such, exploiting the inherent geometric properties of the areas allows for redefining planning as
a combinatorial optimization task. This helps to reduce the overhead complexity of the problem
and break it into manageable subproblems for handling uncertainties.
In this talk, we will explore a hierarchical approach to generating a comprehensive pipeline of
robust exploration and coverage plans, including: (i) creating global trajectory plans, and (ii)
incorporating dynamic changes to adjust trajectories as needed. We will specifically examine
various coverage strategies for both single and multi-robot systems that consider obstacles, the
environment's implicit geological properties, and effective data collection suited to the deployed
sensors. Additionally, we will discuss methods for integrating uncertainties within these plans.
Furthermore, we will showcase real-world applications of these techniques, automating scientific
sampling operations in the marine domain and beyond..
For more info, please follow this link.
Read More