IEEE Robotics & Automation Magazine - December 2010 - 10

C O M P E T I T I O N S

The Mobile Manipulation Challenge
Steve Balakirsky

n this column, we have a report on the 2010 IEEE International Conference on Robotics and Automation (ICRA)
Robot Challenge, held during ICRA 2010 in Anchorage,
Alaska. The challenge consisted of four events and attracted
16 teams that demonstrated state-of-the-art work in mobility
and manipulation, at both human- and micro- (under 600 lm)
scales, planning for industrial automation, and the application of
modular robotics. The report was authored by Steve Balakirsky
[National Institute of Standards and Technology (NIST)], with
contributions from Matei Ciocarlie (Willow Garage), Jason
Gorman (NIST), and Mark Yim (University of Pennsylvania).
The mobile manipulation challenge showed human-scale
mobility and manipulation. Dynamaid, from the University of
Bonn, demonstrated its ability to offer drinks and snacks to human
guests, when requested by a pointing gesture or by showing the
robot an empty container. Lisa, from the University of KoblenzLandau, showed its ability to find objects on the floor and on a
table and manipulate them by combining a gripper mounted at its
base with its robotic arm, in addition to mapping an unknown
environment. The Care-O-bot, from Frauhofer IPA and Siemens,
successfully mapped and navigated an unknown environment
with dynamic obstacles, recognized and grasped objects common
in daily living, and transported them to desired locations. Finally,
PR2 robot from Willow Garage Inc. demonstrated robust pickand-place operations on both known and previously unseen
objects, using its stereo cameras for object detection. Overall, the
challenge participants succeeded in demonstrating reliable autonomous behavior in a real-life scenario while identifying a number of
reusable building blocks that can be shared by future higher
level applications. The team highlight sessions were very well
attended, a trend we hope to continue for future editions. Discussion also focused on active research areas, such as manipulation in
clutter, versatile end effectors, and safe manipulation in human
environments. For future editions, the challenge is likely to center
on more complex, lifelike environments such as homes or offices.
Complementing the human-scale challenge was the NIST
Mobile Microrobot Challenge. An essential part of the challenge rules is that the microrobots must be untethered, and each
microrobot must be smaller than 600 lm in all of its dimensions.
In the 2010 event, there were three tasks, designed to highlight
known technical limitations: the 2-mm dash, microassembly,
and the freestyle task. In the 2-mm dash, the microrobot had to
travel 2 mm and then come to a complete stop as quickly as possible, which provided a measure of the robot's maximum speed
and motion repeatability. The microassembly task was based on
the planar case of the classic peg-in-hole assembly problem
using pegs that were approximately 100 lm wide (equivalent to

the width of a human hair). The freestyle task allowed each
team to select a technical demonstration of their choice. Six
teams participated in the challenge, which showcased several
major advances in microrobot technologies. Most impressively,
a piezoelectric microrobot developed by FEMTO-ST and the
Institut des Systemes Intelligents et de Robotique in France completed the 2-mm dash in 32 ms, and a magnetic microrobot from
ETH Z€
urich was able to complete four peg-in-hole assemblies in
each of three consecutive trials without a single failure. Because of
the outstanding performances in 2010, the 2011 event will see
significant increases in difficulty in these two tasks. The first task will
require autonomous navigation around a figure eight track to
demonstrate 2-D motion control. The second task will be focused
on the manipulation of microscale objects in narrow channels, as
motivated by biomedical applications where objects must be transported through blood vessels for diagnostics or surgery.
The virtual manufacturing and automation challenge (VMAC)
focused on simulated factory automation. This year's events were
designed to focus attention on the problems of mixed-palletizing
and intrafactory package delivery. Mixed palletizing is an example
of the three-dimensional cutting stock problem, a variant of the
combinatorial nondeterministic polynomial-time hard knapsack
problem. Teams were required to generate a pallet stacking plan
from a provided order file and constraints. This plan was then
evaluated by NIST's emerging metrics for pallet quality, built in
simulation, and then built on a 1/3 scale palletizing cell. Two
teams from the Georgia Institute of Technology competed against
the state-of-the art commercial solution in this event by applying
both a neural network learning-based approach and a deterministic planning approach. The teams were able to create mixed pallets
of goods and build the pallets on actual robotic hardware. The second event aimed to address the need for factory robots to operate
in unstructured environments amongst dynamic obstacles. Teams
were tasked with docking a rear-wheeled Ackerman steered
robot with a conveyor system, picking up packages for transport, and then delivering these packages to unloading stations in
the factory. The highlight of the event was the University of
Zagreb controlling three autonomous vehicles with their automated delivery system. This system uses a layered approach with
the lower layer implementing vehicle localization, waypoint
tracking, docking, and undocking. At the higher layer, tasks for
individual vehicles are assigned and scheduled using a version of
the Banker's algorithm. The VMAC competition will continue
next year with both mixed palletizing and forklift operations.
The final challenge event examined robotic planetary contingencies with an eye toward modular robotics. For this event, teams
developed solutions to an emergency problem at a Mars base

Digital Object Identifier 10.1109/MRA.2010.938833

IEEE Robotics & Automation Magazine

DECEMBER 2010

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2010