Simulation Project { As Test-Bed for. Multi-agent Organizational Behavior in

Transcription

1 Simulator Complex for RoboCup Rescue Simulation Project { As Test-Bed for Multi-agent Organizational Behavior in Emergency Case of Large-Scale Disaster? Toshiyuki KANEDA 1,Fumitoshi MATSUNO 2, Hironao TAKAHASHI 3, Takeshi MATSUI 4, Masayasu ATSUMI 5, Michinori HATAYAMA 6, Kenji TAYAMA 2,Ryousuke CHIBA 7, Kazunori TAKEUCHI 7 1 Faculty of Engineering, Nagoya Institute of Technology 2 Interdisciplinary Graduate School of Sci. and Eng., Tokyo Institute of Technology 3 Port and Harbor Research Institute, Ministry of Transport 4 Division of Information System, Applied Technology Co. 5 Faculty of Engineering, Soka University 6 Disaster Prevention Research Institute, Kyoto University 7 Graduate School of Engineering, University oftokyo Abstract. In the RoboCup Rescue Simulation Project, several kinds of simulator such as Building-Collapse and Road-Blockage Simulator, Fire Spread Simulator and Trac Flow Simulator are expected to provide a complicated situation in the case of the large-scale disaster through their synergistic eect. It is called Simulator Complex. This article addresses, rst, system components of the prototype version of this Simulator Complex, then, explains each of the simulators, and nally refers on the Space-Time GIS(Geographical Information System) which is expected to play a role of DBMS(DataBase Management System) in the whole project. 1 Introduction RoboCup-Rescue Simulation Project (RCRS) is being prepared as an opencompetition style research activity. In RCRS, several kinds of simulators, such as Building-Collapse and Road-Blockage Simulator, Fire Spread Simulator and Trac Flow Simulator, are expected to provide a complicated situation of a large-scale disaster through their synergistic eect. It is called Simulator Complex. The prototype versions of simulators are developed and the plug-in specications are published by the RCRS organizer, so it makes a open-style development possible. On the other hand, competitors' role is to develop teams of multi-agent system, such as Fire Brigade, Ambulance Team and Rubble Cleaner, in order to simulate their emergency behaviors in cases of the large-scale disaster, i.e. Hanshin Awaji Earthquake.? The Fourth International Workshop on RoboCup, The Melbourne Exhibition Center, Melbourne, Australia, August 30-31, 2000.

2 Open competitions for performance evaluation of multi-agent systems are to be held, in which the competitors make their agent teams to interact in this Simulator Complex, not only for evaluating the agent development, but also for testing and evolving appropriate strategies and tactics for organizational behaviors in the various kinds of emergency cases. Accordingly, the role of Simulator Complex is assigned to provide the competitors a test-bed for development of multi-agent systems. This article addresses, rst, the system organization of the prototype version of RoboCup Rescue Simulation Project, then, explains the Building Collapse and Road-Blockage Simulator, Fire Spread Simulator and Trac Flow Simulator, and nally refers on Space-Time GIS (Geographical Information System) which is designed to play a role of DBMS(DataBase Management System) in our RoboCup Rescue Simulation Project. 2 System Organization of the Prototype Version of the Simulator Complex Figure 1 shows the system organization of the prototype version of the Simulator Complex. All simulators, the GIS, viewers and agents are to be plugged in to the kernel, therefore, all information are communicated through the kernel. Fig. 1. System organization of the prototype version The case area is chosen to km width in Nagata Ward in Kobe city, where is an typical Japanese inner-city area, and one of the most damaged area

3 in 1995's Hanshin Awaji Earthquake. The simulation covers the rst three days after the occurrence of an earthquake, by simulating phenomenon and activity for each minute by each step. 3 Building Collapse and Road-Blockage Simulator(BRS) BRS calculates the collapse of buildings in the case of earthquake, and based on this result, it generates situations of the road-blockage. BRS is developed by the Port and Harbor Research Institute of the Ministry of Transportation. Buildings and poles collapse and occupy the roads. Falling objects attached to buildings, such as external walls, windows, and roof tiles, and objects attached to poles, such assignboards and hanging power lines block trac. As the seismic motion becomes stronger, the more collapse occurs. Thus, the occurrence of road blockade can be explained based on the degree of collapse and also road width. Figure 2 (a) shows ratio of collapse as a function of the strength of the seismic motion. As for road width, the wider the road is, there is more possibility that D (TXLYDOHQW VHLVPLF LQWHQVLW\ E (TXLYDOHQW VHLVPLF LQWHQVLW\ Fig. 2. Seismic intensity eects for collapse and blockade a passage at the center of these roads will remain intact for trac. The role of the blockade occurrence model is to compute the probability of vehicle trac through the roads, which are separated into seven classes based on their width. Figure 2 (b) shows probability of vehicle trac as a function of the seismic intensity in case of 0-4m road width [4]. 4 Fire Spread Simulator(FSS) FSS calculates many res at the same time in the case of the earthquake and simulates the spread process of the res, and change the situations also depending

4 on re-ghting activities of Fire-Brigade agents. FSS is developed by Applied Technology Co. Ltd. in collaboration with some academic researchers. FSS is designed as a general-purpose system that can be applied to the re not only at earthquake cases but ordinary cases. The ow chart of computation is shown in Figure 3, which consists of the following processes: 1)combustion process, 2)spread process, 3)ignition process, and 4)the inuence of the ghting a re power. Fig. 3. Flow-chart of re spread model [2] 5 Trac Flow Simulator(TFS) TFS simulates the moves of the agents on the roads especially blocked by rubbles of buildings. Briey saying, the role of TFS should be assigned to provide a trac situation in a inner-city area in case of a large-scale disaster. The following items are considered for the model design. 1. inuences on road blockage mainly caused by the collapse of buildings

5 2. inuences on blackout of trac signals 3. problems on passing of emergency cars such as re brigades 4. problems on refuge in heaps of rubble The trac network consists of roads and nodes (intersections). A road has the following properties: (a) nodes in connection with the road,(b) road width (mm), (c) blockage width of rubble,(d) the number of lanes of 'upward' line, (e) the number of lanes of'downward' line, and (f) width of sidewalk (mm). A node has the following properties: (a) roads in connection with the node, (b) existence of signal, (c) 'split' pattern of the signal (if it has a signal), (d) the number of lanes of short-cut for left-turn, (e) the number of lanes of pocket for right-turn, and (f) lane length of the pocket. At the start of simulation and each simulation step, the simulator receives information from the kernel and renews information on location of ambulance, re-brigade, police and refuge agents. Inside the trac simulator, trac is simulated every one second, which is the same as TRANSIMS [5] that is a cellular automata-based micro-simulator. The simulator sends back to the kernel simulated results in each step. As further direction of development, we should incorporate mechanisms to simulate eects of emergency trac control [1]. The following are requirements imposed on the current trac ow simulator to realize the trac control { implementation of various channels for information gathering by trac control agents, { implementation of trac signal control mechanism, trac signal control interface with trac control agents, and trac ow calculation under the trac signal control, { implementation of various trac restrictions instructed by trac control agents and various channels for communicating trac restriction information to car agents, { implementation of damage management mechanism on trac control equipment and its monitoring interface by trac control agents. 6 Space-Time GIS as DBMS In order to manage spatial temporal information of a concerned simulation area, Space-Time GIS(Geographical Information System) is devised as a Database Management System (DBMS). We call this system DiMSIS [3]. The essential feature of DiMSIS is that spatial information is described based on topology calculation and temporal information is managed based on the space-time approach. To manage temporal information, there exist the following two approaches (see also Figure 4). Snapshot View Model: In this approach, each spatial data set is connected with the temporal information. Almost all the commercialized spatial temporal GISs have adopted this approach.

6 Space-time Approach Model: In this approach, each element of the spatial data is connected with the temporal information. The spatial temporal GISs adopting this approach start studying recently. 7 W 7 W 7 W 6QDSVKRW 9LHZ 0RGHO 6SDFH7LPH $SSURDFK 0RGHO WJHQHUDWLRQ WLPH WJHQHUDWLRQ WLPHWGLVDSSHDUDQFH WLPH WJHQHUDWLRQ WLPH Fig. 4. Snapshot view and space-time approach for GIS data structure We compare the performances on memory use and computational complexity of the following four types: Explicit Topology & Snapshot View, Topology Calculation & Snapshot View, Explicit Topology & Space-Time Appraoch and Topology Calculation & Space-Time Approach. Results of theoretical evaluations suggested that our proposing Topology Calculation & Space-Time Approach was superior to other types. 7 Concluding Remarks The design principles and logics of models of three kinds of simulator, Building Collapse and Road- Blockage Simulator, Fire Spread Simulator and Trac Flow Simulator, and the Space-Time GIS are explained. The prototype version of this Simulator Complex has now achieved the speed of one minute simulation per second, that enables us to hold open-style agent competitions. References 1. Atsumi, M.: Adaptive Coordination of Distributed Truth Maintenance among Agents - Experimental Evaluation with Expressway Control Agency Testbed - IPSJ SIG-AI, 95-AI-102 (in Japanese). (1995) Fire department, Kobe city oce (eds.): Ocial Report on Fire in Hanshin Awaji Earthquake. Tokyo Horei Shuppan (in Japanese). 3. Kakumoto, K., Hatayama, M., Kameda, H., and Taniguchi, T.: Development of Disaster Management Spatial Information System (DiMSIS). GIS'97 Conference Proceedings. (1997) 595{ Kotani, M., et.al.: Analysis on Blockage of Neighborhood Road-Networks in Hanshin Awaji Earthquake. Journal of Trac Engineering Association (in Japanese). (1996) Nagel, K., Beckman, R., Barrett, C.: TRANSIMS for Transportation Planning. 6th Int. Conf. on Computers in Urban Planning and Urban Management '99. (1999)