a. kleiner, albert-ludwigs-universität freiburg robocup rescue austrian robocup workshop 2007...

A. Kleiner, Albert-Ludwigs-Universität Freiburg

RoboCup Rescue Austrian RoboCup Workshop 2007

Motivation Rescue Robot League

RRFreiburg: Behavior Maps Rescue Simulation League

Virtual Competition RRFreiburg: RFID Technology-based Exploration

Agent Competition


RoboCup Rescue - Austrian RoboCup Workshop 2

MotivationThe time problem after an incident



MotivationWhere is the benefit from robotics technology?

After a disaster many places are unreachable for humans

Robots can access places humans can’t (e.g. small holes or spaces under the floor)

Robots can send video and thermo images from hazardous places

Destroyed infrastructure: Problem of self-localization

Quality of disaster response strongly depends on information, such as maps with victim locations

Tom Haus (firemen at 9/11): “We need a tracking system that tells us where we are, where we have been, and where we have to go to”

Technology from Robotics can be deployed for information gathering and world modeling

Autonomous systems: Reduction of “cognitive load”



Rescue Robot LeagueResearch Challenges

High degree of mobility Simultaneous Localization

And Mapping (SLAM)

Victim detection

Autonomy!Autonomy!



Rescue Robot LeagueBuild robots that are ready to leave the lab!

You might be too small … … or you might be too big



Rescue Robot LeagueGoals and directions

Cooperative development with simulation league Step-wise increase of difficulty (e.g. like golf

courses) Building of standards for mapping and data

exchange between heterogeneous units Towards “mixed-initiative” solutions, i.e. humans

and robots build one team for efficient disaster response



Competition settingThree types of arenas

YELLOW ARENARANDOM MAZEPITCH & ROLL RAMP FLOORING (10°) DIRECTIONAL VICTIM BOXES (FOR AUTONOMOUS ROBOTS)

ORANGE ARENAPITCH & ROLL RAMP FLOORING (10°, 15°)

HALF CUBIC STEPFIELDSCONFINED SPACES (UNDER ELEVATED FLOORS)

VICTIM BOXES WITH HOLES

RED ARENAFULL CUBIC STEPFIELDS

STAIRS (40°, 20CM RISERS)RAMP (45° WITH CARPET)

PIPE STEPS (20CM)DIRECTIONAL VICTIM BOXES

REGIONAL/PRELIMINARY ARENAS SHOWN, CHAMPIONSHIP ARENAS WILL BE TWICE THIS SIZE



Competition settingSimulated victims

Signs of life: form,

motion, heat,

sound, CO2

THERMAL IMAGE

VISUAL IMAGE



Competition settingRules at a glance

GeoTIFF map formats will be used to allow comparison of maps to ground truth arena configurations.

Best-In-Class awards for autonomy and mobility will be given to robots that find the most victims in the Yellow and Red arenas respectively over all missions.

Random mazes with non-flat flooring Stepfield pallets (Orange: half-cubic, Red: full-cubic) Stairs (40°, 20cm riser, 25cm tread depth) Ramp (45° to test torque and center of gravity) Confined spaces (ceiling blocks under elevated floors) Visual acuity (tumbling E eye charts, hazmat labels) Directed perception boxes with victims/targets inside Simulated Victims: 4 per arena, 12 total Signs of life: form, heat, motion, sound, and/or CO2



Competition settingRules at a glance II (Missions)

15/20/25 minute missions include robot placement at the start point and operator station setup. Each team is responsible for making sure victims are functional (heat, batteries, tags) prior to their mission start.

Teams are allowed one operator during missions. Start points will be in the Yellow arena with all robots facing the same direction

(“north” on your map). Yellow arena victims can be scored only by robots with autonomous navigation

and victim identification. Operators may take over control at any time to move into the Orange and Red arenas but must return to the start point to resume autonomous searches.

Teleoperative robots can only score Orange or Red arena victims, which are placed on both sides of the Yellow arena to encourage complete mapping.



Teams at GermanOpen 2007



Behavior MapsElevation mapping and classification of rough terrain

Basic idea: 1) Robot with tilted Laser

scanner and IMU sensor explores rough terrain.

2) Generation of elevation map, and classification with Markov Random Fields (MRFs)

3) Detection of skill pre-conditions, e.g. starting position and angle

4) Planning and execution of skills

Lurker robot



Rough terrain Elevation Map

Classified Map Behavior Map

Behavior Maps cont.Elevation mapping and classification of rough terrain



Behavior Maps cont.Elevation mapping and classification of rough terrain


Rescue Roboticsand the RoboCup Rescue Challenge

MotivationRescue Robot League

RRFreiburg: Behavior MapsRescue Simulation League

Virtual CompetitionRRFreiburg: RFID Technology-based Exploration

Agent Competition



Rescue Virtual CompetitionUSARSim

Based on the Unreal game engine (UT2004, Epic Games)

Realistic models of USAR environments, robots (Pioneer2 DX, Sony AIBO), and sensors (Laser Range Finder, Color Camera, IMU, Wheel Odometry)

Multiple heterogeneous agents can be placed in the simulation environment

High fidelity simulation of up to 12 robots

Agents connect via a TCP/IP interface NEW: Communication Server



Rescue Virtual CompetitionIntroduction cont.

Unreal Client

Unreal Server

Command

Sensor data

Sonar Sensor message



Rescue Virtual CompetitionPerformance Metrics

Victim discovery Victims are detected by a simplified sensor retuning ID/state depending on

distance Victim ID (10pt), Victim status (20pt), victim location (10pt), additional information

(20pt) Self Localization and Mapping (SLAM)

Metric quality (50pt): How close are reported locations to ground truth? Multi-robot fusion: Bonus for maps generated by multiple robots

Exploration Max. 50pt if exploring the whole area NEW: Explored areas have to be marked as “cleared”

Penalization Robot-Victim collision (-5pt) Teleoperation: Division of total score by (1+N)2

NEW: One mandatory operator for each team



Rescue Virtual CompetitionMore new features

Improved robot models for

realistic mobility

GeoTIFF format for maps Maps will be overlaid on and compared

to ground truth Teams must specify areas “cleared”

Points deducted for victims in “cleared” areas



Rescue Virtual Competition RRFreiburg solution at RoboCup’06

Basic Idea: RFID Technology-based Exploration Robots generate local grid maps, generated from Laser Range data A* based planning on local grid Each robot distributes autonomously RFID tags and counts locally in the

memory of tags the relative locations already visited If in perception range, robots receive the data of tags and optimize their

search by avoiding frequently visited places Extension: Global planner that resets the local search if beneficial

Cheap computation on each robot due to a local world model Locations are stored relatively to the tag, hence do not suffer under

positioning errors Efficient coordination without need for communication



Rescue Virtual CompetitionResults from RoboCup’06 cont.

Area explored by all teams during the finals



Rescue Virtual CompetitionResults from RoboCup’06: Exploration Trajectories

Area explored by our team (red trajectory) compared to all others

Area explored by each single robot of our team

Semi-final (1276m2)

Final (1203m2)



Rescue Virtual CompetitionVideo from the final


Rescue Roboticsand the RoboCup Rescue Challenge

MotivationRescue Robot League

RRFreiburg: Behavior MapsRescue Simulation League

Virtual CompetitionRRFreiburg: RFID Technology-based Exploration

Agent Competition



Rescue Agent Competition Introduction

Large scale disaster simulation Simulators for earthquake, fire, civilians,

and traffic The task is to develop software agents

with different roles, that make roads passable (police) extinguish the fires (fire brigades) rescue all civilians (ambulances)

Difference to Soccer Simulation: A challenging Multi-Agent Problem since Agents must cooperate

Simulator components are developed within the “Infrastructure Competition”



Conclusion RoboCup Rescue offers a rich set of problems to AI and Robotics Due to the difficulty for robots to cooperate in harsh environments,

research is just at the beginning Developed solutions are socially significant!

Links: Rescue Robot League:

Homepage: http://robotarenas.nist.gov/competitions.htm Rescue Simulation League:

Homepage: http://www.robocuprescue.org USARSim (code base): http://sourceforge.net/projects/usarsim Rescue Agent (code base): http://sourceforge.net/projects/roborescue



Thanks for your attention!

a. kleiner, albert-ludwigs-universität freiburg robocup rescue austrian robocup workshop 2007...

Documents

red arena victims

yellow arena victims

incident slide

size slide

big slide

humans robots

holes red arena

floor robots