Title: Robotic Manipulation in Human Environments: Challenges and
Lessons Learned from DARPA ARM Project

Abstract:
========

DARPA’s Autonomous Robotic Manipulation (ARM) program aimed to develop
software that enables a robot to autonomously manipulate, grasp, and
perform complicated tasks with humans providing only high-level
supervision. Over the course of the program (2010-13), six funded
teams (JPL, iRobot, USC, SRI, HRL, and CMU) competed and developed
algorithms that enable the DARPA robot to accomplish a number of
everyday manipulation and grasping tasks.

In his talk, Dr. Moslem Kazemi, a former member of the CMU ARM team,
will provide an overview of the program, the challenges, and the
approach taken by CMU team to address the challenges. Their approach
was centered around the tight integration of several core
functionalities, including perception, planning and control, with the
logical structuring of tasks driven by a Behavior Tree
architecture. The talk summarizes the lessons learned throughout the
project and gives an overview of related research problems and the
existing hard challenges facing robotic manipulation in human
environments.