Post on 25-Jan-2022
Carnegie Mellon University THE ROBOTICS INSTITUTE
Thesis Proposal Sankalp Arora
Monday, April 24, 2017NSH 110911:00 a.m.
Sebastian Scherer Chair
William (Red) L. Whittaker
David Wettergreen
Kostas Alexis University of Nevada, Reno
Thesis Committee
Safe, Efficient Data Gathering in Physical Spaces
Abstract
Reliable and efficient acquisi0on of data from physical spaces will have countless applica0ons in industry, policy and defense. The capability of gaining informa0on at different scales makes Micro-‐Aerial Vehicles (MAVs) excellent for aforemen0oned applica0ons. However, reasoning about informa0on gathering at mul0ple resolu0on is NP-‐Hard and the state of the art methods are too slow to present an approximate solu0on online. Moreover, a robust data gathering system needs to reason about mul0-‐resolu0on nature of informa0on gathering while being safe, and cognizant of its sensory and baKery limita0ons.
This thesis addresses three key aspects of enabling safe, efficient, mul0-‐resolu0on data gathering: online budgeted mul0-‐resolu0on informa0ve path planning (IPP), guaranteeing safety and, op0miza0on of sensing bandwidth for implicit and explicit data gathering requirements.
Firstly, we present an online naviga0on algorithm to guarantee the safety of the robot through an Emergency Maneuver Library (EML). We discuss an efficient method to construct EML while exploi0ng vehicle's dynamics capabili0es. We then present an informa0on gathering approach that op0mizes the sensory ac0ons to ensure vehicle safety and gain informa0on relevant for mission progress. We validate these methods by deploying them on-‐board a full scale helicopter, demonstra0ng significant performance increase. We address the IPP problem through Randomized Any0me Orienteering (RAOr), an any0me, asympto0cally near-‐op0mal algorithm, that enables the planning for informa0on gathering online.
We will focus our future work on three sub-‐problems that will lead to a safe, efficient data gathering framework. The first is developing a receding horizon planner that enables the vehicle to stay safe while maximizing the informa0on gathered, through embedding safety constraint and informa0on theore0c reward func0ons in sampling based planning framework. The second is learning a set of heuris0cs to enable faster mul0-‐resolu0on informa0ve path planning through RAOr. The third is to use the safe data gathering framework to improve vehicle's long-‐term performance through improving its assump0ons about the environment.
We will evaluate the performance of our informa0on gathering framework on an autonomous MAV. We expect that our framework will enable long term deployment of autonomous mul0-‐resolu0on data gathering systems, while guaranteeing their safety, enabling MAVs to realize their poten0al as efficient data gatherers.