This document proposes an approach for coordinating human-agent teams during urban search and rescue operations after an earthquake. It analyzes the structure and environment of search and rescue tasks, and identifies coordination as a key challenge due to geographic and uncertain conditions. The document then describes designing a model for distributed spatial task allocation using a contract net approach. Intelligent assistant agents are proposed to help humans assign tasks to teams efficiently in space and time. The approach is implemented and validated in a geospatial simulation.

1. Introduction to Spatially Distributed Intelligent
Assistant Agents for Coordination of
Human-Agent Teams’ Actions
Reza Nourjou 1
Informatics Graduate School
Kyoto University
Kyoto, Japan
nourjour@gmail.com
Michinori Hatayama 2
DPRI, Kyoto University
Kyoto, Japan
hatayama@imdr.dpri.kyoto-u.
ac.jp
Hirokazu Tatano 3
DPRI, Kyoto University
Kyoto, Japan
tatano@imdr.dpri.kyoto-u.ac.jp
Abstract Earthquake emergency response requires different
number of teams to cooperate with each other and coordinate their
activities to achieve global objectives. Coordination is the key
challenging problem that field teams face in urban search and rescue
(USAR) emergency response because of the geographic and
uncertain environment. Therefore, it is necessary to optimize
coordination of teams by allocating tasks to teams in time and space
efficiently and sufficiently. The present paper aims to propose an
efficient approach that allows humans to collaborate with coordinator
assistant agents to assign tasks to teams. To achieve this purpose,
we did the following stages: (1) the structure of the USAR task
force and environment characteristics were analyzed, (2) we
analyzed the spatial coordination problem and then modeled it, (3)
features and properties of required approach were assessed, (4) we
designed a contract net based model for distributed spatial task
allocation and proposed the architecture of spatially distributed
intelligent assistant agents (SpDI2A), and (5) finally, our approach
was implemented and validated in a geospatial simulation which was
developed using AnyLogic software and Java programming. The
final result composes of 10 features satisfying the features of the
required system.
Keywords: coordination, multi-agent systems, GIS, geospatial,
simulation, urban search and rescue, task allocation, human-agent,
earthquake emergency response
I. INTRODUCTION
each other and coordinate
their actions to do complex tasks in order to achieve global
utility and objective. USAR has become an important issue in
the recent decades because a number of earthquake disasters
have occurred and had many deaths and injured especially in
the developing countries. USAR operations are vital in
earthquake emergency response especially during the first 72
hours. Their main objective is to maximize number of rescued
victims in the shortest time. To reach this goal, one
appropriate solution is to make decision on allocating tasks to
teams in time and space efficiently and sufficiently. Because
of complex environment characteristics of earthquake disaster,
coordination is a key challenging problem that field teams
face during field operations.
Five reasons require coordination among teams in
multi-agent systems: (1) preventing anarchy or chaos, (2)
efficiency, (3) meeting global constraints, (4) distributed
information, expertise or resources, and (5) dependencies
between the teams’ activities [17]. In coordination theory,
coordination is the act of managing interdependencies
between activities performed to achieve a goal [15]. In the
recent decade, multi-agent systems have been used for
disaster management systems and emergency response
simulations [9, 12] that we can mention DEFACTO [19],
ALADDIN [11], RoboCup rescue simulation [13], and
Coordinators [14]. The main methods [8] of coordination
problem solving are classified into the task allocation,
coalition formation, multi-agent planning, communication and
data sharing, and negotiation.
Characteristics of environment of emergency response make
coordination of teams so difficult and complex. Emergency
response requires an appropriate approach in order to
overcome this challenge and optimize the objective of USAR.
After required properties of this approach were analyzed in
this paper, we found the Geospatial feature vital. The present
paper aims at proposing this efficient approach for solving the
coordination problem of USAR in the geographic and
uncertain environment. This approach provides a framework
that allows a human to collaborate with his coordinator
assistant agents to assign tasks to field teams. We called this
approach SpDI2As or agent. The paper proposes the roles of
the agents and the way for designing them. The paper also
illustrates its implementation ability and validity by
developing a geospatial simulation.
II. ISSUES IN URBAN SEARCH AND RESCUE OPERATIONS
A. Structure of USAR Task Force in Earthquake Disaster
USAR Task Forces [7, 21] are the key responsible
organization to coordinate and carry out tasks and Fig. 1
Proceedings of the 2011 IEEE International Symposium on
Safety,Security and Rescue Robotics
Kyoto,Japan,November 1-5 2011
251
978-1-61284-769-6/11/$26.00 ©2011 IEEE

2. shows its organization. They include the multiple teams of
planning (situation assessment), hazmat, search, rescue,
medical, and logistics. An incident commander (IC)
commands their activities and makes tactical decisions which
determine what team has to do what task, where, when, and
with whom.
The goal of situation assessment teams is to locate damaged
buildings, estimate number of potential victims trapped in
debris, and report situation information to the IC. Search
teams departure to the disaster-affected area and move to
assigned buildings to do search tasks, considering the tactical
decisions made by the IC. They try to find and locate trapped
victims, and then they report situation information
(information of location and exact number of found victims,
state of trapped victims, and required rescue levels (low,
medium, and heavy)) to the IC. In fact this information is
output of the done search task. Based on the made tactical
decision, rescue teams move to location of assigned tasks. By
accessing information gathered from those buildings, rescue
teams release and rescue victims. They disseminate
information of number of released victims, number of injured
victims, and type of injure to the IC. This procedure is
repeated over time until 72 hours. Because our assumption is
to focus on only two types of tasks (search and rescue tasks)
and two types of teams (search and rescue teams), we do not
explain other tasks and teams in this paper.
B. Environment Characteristics of Emergency Management
To understand the coordination problem of USAR teams
clearly, we aimed to study and analyze environment
characteristics of earthquake emergency response. We
extended the results of other researchers [4, 12] to describe
the following characteristics:
1) Key elements of environment: Tasks, teams, damaged
buildings, and tactical decisions constitute key elements,
which are shown with their related information in Fig. 2. In
addition, treatment centers, operational areas, city blocks, and
street network seem to be important.
2) Uncertainty and ambiguous information: This
characteristic causes many challenging problems especially
during the first 72 hours. For example, different information is
reported to database; some tasks are revealed, done, or
canceled unexpectedly; teams can not follow made decisions
schedules; and database is not updated with new situation
information.
3) Geographic aspect: Earthquake disaster- affected area is
an extensive geographical environment whose elements are
spatial entities and are distributed in the environment.
4) Spatial decision making: Geographical distance or travel
time among elements is an important parameter in action
planning and decision making.
5) Dynamic in time and space: Teams have dynamic
positions, and also information of elements varies over time.
6) Heterogeneous teams: Search teams and rescue teams,
who are the 2 key teams, have different expertise and goals.
7) Multiple agents: There are more than one rescue teams or
search teams at a USAR Task Force.
8) Real time decision making: Because of dynamic
information, decision making should be done in real time
way.
9) Interdependency Among activities: Tow types of
interdependencies were studies in this paper: “Enable” and
“Equal”. The “Enable” relationship specifies that when an
action is carried out, it makes possibility of performing
another action; and equality relationship means that certain
actions are not liked to a specific team, and can be carried out
Sum of
all task
Search Team
(spatial Intelligent Object)
Real time location
State
Real time location
State
Rescue level
Number of rescued victims
Amount of total time
Geo-referencedlocation
Search task Rescue Task
Damaged Building
(spatial entity on the map)
Building Damage
level
Estimated number
of trapped victims
Start time
Location and
number of found
trapped victims
Required rescue level
State of trapped v.
Tme used for search
Who did it
Number and sate of
rescued victims
Number of injured
released victims
State of rescue task
Time used for rescue
task
Finish time
Who did it
Does
Rescue Team
(spatial Intelligent Object)
Fig. 2. Geographical distribution of human-agent teams in area and
collaboration human with SpDI2A via PDA; and spatial coordination
problem of USAR modeled for one damaged building;
Fig. 1. Organization of the USAR Task Force with 6 different field teams[7]
252