Resuts are shown here.
A distributed computer simulation environment
of earthquake disasters including maps has
been prepared by the organizer, where buildings
collapse, streets are blocked, fire spreads,
and traffic condition is affected according
to given seismic intensity maps, reproducing
virtual comprehensive urban disaster.
Intelligent action brigades (agents) of software
try to minimize the disaster damage in this
virtual space. Rescue parties save
victims from the destroyed buildings, firefighter
companies extinguish the fire, and police
parties open the blocked roads. The
magnitude of damage is determined by the
behavior of these agents and their intelligence.
The autonomous agents have abilities similar to robots: sensing (see, hear, conversation, etc.) and action
(move, talk, rescue, extinguish, repair roads,
etc.) by their own decision and strategies.
In this competition, each participating team
develops 5 rescue parties, 10 fire companies,
10 police parties, 1 rescue center, 1 fire
station, and 1 poice station which behave
autonomously. Behavioral intelligences
compete each other in disaster fields 500
x 500m. The winner of competition
should rescue the maximum number of victims
minimizing disaster damage.
A pair of agent teams try a disaster field at the same time in order that the audience can compare them. They do not mutually fight each other, but they challenge the same situation independently such as golf and athletic sports. In this disaster domain, obstructing the opponent is not desirable, and fighting style competition is not suitable.
In order to minimize possibility of accidental win, all the teams try multiple disaster maps in the preliminary games and the total points determin the finalists. These maps are created by the participants. In the final round, total points of two maps prepared by the organizer judges the championship.
The point is counted by the number of remain of lives. Area of burnt and remaining agents' HP are considered supplementally.
The RoboCup-Rescue WC 2001 rule is announced here.
The following computer environment will be prepared. All the programs including simulators, agents, and logViewers should use these computers.
The participants should make their source code open after the competition.
Each participating team prepares a set of agent parties. Modification of program, parameters, etc. is not allowed after starting the competition.
| No. | Team Name | Team Members (Affiliations) |
Technical Appeal Points |
| 1 | Arian Simulation Team | Jafar Habibi, Mazda Ahmadi, Mohammad Badry,
Hossein HadiPour, Farshid Marbooti, Ali Nouri
(Sharif University of Technology, Iran) |
??? |
| 2 | Bachau (CANCELED) |
P Ravi Prakash, Rahul D Vakil (National Centre for Software Technology, India) |
Intelligent agents, planning and scheduling |
| 3 | DAI-HARD (CANCELED) |
Neelima Sajja, Steven Nuchia, Rajatish Mukherjee,
Sandip Sen (University of Tulsa, USA) |
We are interested in multi-agent co-ordination and distributed artificial intelligence. Dr. Sandip Sen has been working in the area of multi-agent systems for the past 10 years and is currently focussing his research on team-work and multi-agent co-ordination. We are mainly interested in the learning aspect of robo-cup, where dynamic task allocation and planning is a research issue. |
| 4 | Gemini-R | Masayuki Ohta (Tokyo Institute of Technology, Japan) |
Agent strategies are determined according to the optimal order of rescue activities and allocation of man-power that has been automatically learned by simulation results. |
| 5 | Rescue-ISI-JAIST | Takayuki Ito, Milind Tambe, Ranjit Nair,
Stacy Marsella (Information Science Institute / University of Southern California, USA; Japan Advanced Institute of Science and Technology, Japan) |
Cooperative and autonomous rescue activities are generated. |
| 6 | NITRescue | Tetsuya Ezaki, Taku Sakushima, Nobuhiro Ito,
Yoshiki Asai, (Nagoya Institute of Technology, Japan) |
Cooperation of multiple agents is important in a dynamic environment as the RoboCup-Rescue simulation. Cooperative behabior of agents is defined as a group behavior of agents. A dynamic grouping algorithm is developed. |
| 7 | no-fear (CANCELED) |
Norifumi Oda (Toyohashi University of Technology, Japan) |
A learning algorithm of AI gives optimal rescue action. |
| 8 | RMIT-ON-FIRE | Lin Padgham, James Harland, John Thangarajah,
Naveen Ruwanpura, Chandaka Fernando (Royal Melbourne Institute of Technology University, Australia) |
BDI intelligent agent system. Development primarily by final year undergraduates using JACK BDI agent development environment. |
| 9 | Survivor (CANCELED) |
Von-Wun Soo, Ken Chun Chen, Maw Yuan Hsu,
Biing-Yi Lin (National Tsing Hua University, Taiwan) |
??? |
| 10 | YabAI | Takeshi Morimoto (University of Electro-Communications, Japan) |
Behavior of agents is switched according to the distribution of disater. |
| 11 | JaistR | Kosuke Shinoda (Japan Advanced Institute of Sience and Technology, Hokuriku, Japan) |
Each Agents has learning mechanism based on Organizational Learning. |
The detail is announced here.