COHORT - CoOperation for HeterOgeneous Robot Teams

Loading...
Thumbnail Image

Authors

Kulkarni, Mihir

Issue Date

2021

Type

Thesis

Language

Keywords

Aerial Robots , Multi-Robot Systems , Path Planning , Robotics , Simulation , Subterranean

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

This thesis presents a multi-robot exploration path planning framework for underground environments, the development of associated simulation models, alongside further contributions to the winning performance of Team CERBERUS at the DARPA Subterranean Challenge Final Event.First, simulation models are detailed for multiple aerial platforms and a ground system developed in order to enable the systematic evaluation of contributions in the domains of control, perception, and path planning. The models have been submitted to open-source repositories for utilization on the Ignition and Gazebo simulators. We also reconstruct meshes from LiDAR point clouds collected from robotic deployments within complex field environments which are then used in simulation. The methodology of generation of the simulation models is further presented.Secondly, a centralized multi-robot coordination framework that facilitates synergistic exploration of large-scale, multi-branching and diverse subterranean settings is developed. This framework allows for incrementally built submaps to be sent to a central ground station whenever a communication link is available. The central ground station identifies global frontiers of exploration space to direct each system to an appropriate location in order for it to best continue its mission. The framework is demonstrated through a simulation study with 3 flying robots exploring a multi-branched environment with simulated network connectivity constraints.Finally, additional contributions to the artifact detection pipeline used by Team CERBERUS in the DARPA Subterranean Challenge, as well as modifications to one of our team's ground systems are presented.

Description

Citation

Publisher

License

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN