Supervision and Operation Decision Support Platform for Urban Rail Transit

878 JOURNAL OF SOFTWARE, VOL. 8, NO. 4, APRIL 2013

Ming Zhang Institute of Computing Technologies, China Academy of Railway Sciences, Beijing, China Email: [email protected]

Fuzhang Wang and Ping Li and Yingjie Wang Institute of Computing Technologies, China Academy of Railway Sciences, Beijing, China

Abstract—This paper proposed approach of integrating as well as information need to be integrated. Once supervision alarms and trains located on the map to support accident happened, dynamic changes of passengers would emergency management of urban rail transit network. The lead to congest in short time if operation decision scheme characteristics of dynamic passenger flow are analyzed to is not be make out efficiently. Adjustment of timetable in figure out transfer coordination and evacuation schema, urgent period of accident is necessary, especially in related algorithm is developed. Information sharing is discussed and method of connection timetable optimization transfer hub, which requires excellent algorithms and data for metro and other traffic tools are introduced in hub. analysis of running trains. To meet these challenges, Explored technologies are applied in supervision and advanced supervision in real-time is introduced in the operation decision support platform, with introducing process of accident treatment, and the platform aiming at hierarchical structure for data exchange and applications. supervision information sharing and decision support is Implementation of the technical system shows its presented. effectiveness in emergency treatment and rescue resource Typical problems needed to be confronted in the collocation in time. information disposing platform can be collected as follows: Index Terms—urban rail transit; emergency management; (1) More transfer hubs caused dynamic passenger flow information sharing; supervision; decision support between metro lines, which further leads to difficulties of transport capacity allocation. (2) Corresponding to I. INTRODUCTION multiple metro agencies, cooperation and resource sharing serve as another requirement in the new mode of In recent years, multiple metro lines are connected and management, e.g. sharing of power substation for several more complex urban rail transit (URT) network is URT lines. (3) How to utilize information collection and emerging. Relative separated space, crowded passengers, act as crucial resource to assist emergency rescue. (4) and complicated equipment increase possibility of Deployment of rescue resource, like Aid teams or rescue accident in URT caused by potential factors such as tools, is a basic factor for accident treatment, as well as equipment failure, natural disasters, especially in their coordination with other traffic tools for passenger randomness, dissemination, and diffusivity within urban evacuation, and corresponding information is necessary. rail net. In order to avoid or reduce the loss of accident, Thus, concerning complexity and characteristics of measures and technologies are employed. Safety greatly URT operation, hierarchical structure theory of system depends on disaster preparedness and emergency engineering is introduced to establish respective measures. Pertinently supervising danger zone and coordination between rail lines. Depending on location facilities by installing sensor devices can improve details and route programming of GIS technologies, reliability [1]. Emergency management visualization with location of every train on rails could be calculated to hypermedia digital application [2, 3] and available match on signal sections, which acts as basis to simulate equipments position by employing GIS technologies [4] trains motivation in real-time, then pre-alarm for can assist emergency decision [5, 6]. The safety analysis interruption and delay could be further attained through for dependable system is also researched to improve definitely algorithms. The information sharing strategy reliability of these tools [7]. To date, integrated serves as crucial method to build efficiently utility of supervision and control system (ISCS) has been installed supervision data analysis. Also, passenger quantities and in stations to monitor running situation of equipment and distribution orientation are chosen as basic factors to alarm while malfunction occurred [8]. ISCS of each rail figure out rational result of composite coordination in the line can deal with simple faults or running with them, background of evacuation among different types of traffic relying on its self-fitness performance, whereas it is tool. difficult to organize rescue activities like information The research outcome has been converted into an collection, making decision, and resource deployment. information platform for safety and emergency Emergency plan is urgently required in context of disaster

© 2013 ACADEMY PUBLISHER doi:10.4304/jsw.8.4.878-884 JOURNAL OF SOFTWARE, VOL. 8, NO. 4, APRIL 2013 879 preparedness in URT system. It indicates the needs of then low or high level can be counted based on percentage operation and emergency concern more information of fluctuating range. It defines utilization, including supervision, dynamic passenger + − statistics, and allocation of rescue resource on the map. u = u0 (1+ p%), u = u0 ()1− p% Pre-alarm for part of URT network contributes more on p uses as default value 10%, of course it also can be reducing damage and establish evacuation scheme by collecting and analyzing supervision information of modified by system user. While pre-alarm appears, equipments and trains. corresponding switches and their attributes will be searched, e.g. name, number, transformer substation II. SUPERVISION INFORMATION OF URT NETWORK which belongs to, etc. Alarm in FAS abides by divided area principle that fire A. Derivative of Alarm Information protected subarea sets out alarm message whenever any Supervision system collects all sorts of status from fire sensor detects fire signal, which lies in the zone of device monitors installed in stations or trains, and gives out subarea. Especially, the partition for these subareas in alarm above definite grade, such as overrun alarm of FAS system is consistent to fire management area at electric voltage and current of main devices in core station, more often, equipment room or independent space substation, breakdown of key equipment like drainage are treated as certain subarea. Thus, station is composed of system or exhaust fan in tunnel, fire alarm, etc. Concerning many fire protected subareas, which secures accurate fire complexity of monitor system caused by its large amount alarm for stations in FAS system is achieved. The set of of types, the strategy of hierarchical classification is subareas is described as f ()i = {f1 , f 2 ,L, f i ,L, f n }, introduced to identify extent of alarms. The classification means the subarea includes n fire sensors. Also, set of is defined as equipment “type, catalog and group”. “The station can be defined as types” is corresponding to public area of station, and every type is related to multiple “catalog”, which matches with s( j) = {s1 ( f ), s2 ( f )()(),L, s j f ,L, sm f } (1) location like station hall. “The grades” associates with equipment status. As system and equipment have It means there are m subareas in the station. If f i = 0 , respective alarm threshold, we set index of pre-alarm then sensor i sent out fire alarm. So, if f ∧ s = 0 , then which links to judge rules for emergency response. Defines fire alarm is true and a alarm message is activated on index Δi as alarm grade, so system screen. ⎧ 0 normal To be more important, the multi-level matched with ⎪ , alarm for low or high level like Power Δ i = ⎨ 1 low − level multi-category is pointed out in alarm supervision analysis. ⎪ In detail, all sorts of alarms exists in sensors show different ⎩2 high − level characters and extent. Concerning variety of alarms, the Supervision Control and Data Acquisition System (PSCADA). four-grade alarm is defined to identify how importance and urgency is and which kind of response should be started. ⎧0 normal Δi = ⎨ ,only for two values like Fire Alarm First-grade: the most emergent alarm, which is often ⎩ 1 alarm related to lose lives or destroy of URT, e.g. abnormal in the System (FAS). circumstances of train running, fire in station or train, etc. ⎧0 alarm Second-grade: momentous accident that leading to Δ = Alarm grade can be divided into five i ⎨ ，，，， interrupt normal transportation for definite time, such as ⎩1 Δ ij ∈[]1 2 3 4 5 types. electricity fault of crucial equipment. Third-grade: serious accident or great equipment ⎧0 normal Δi = ⎨ Alarm points to manual settings. malfunction, which require dispose in time to avoid ⎩1 alarm interfering trains running. Every cycle of system scanning will be execute full Fourth-grade: common malfunction that may bring record scanning until the index points to manual settings about accident, for example, super-high beyond water line and scanning end, with marking one cycle is over. If in pump. index ∃Δ = 1 , set mark as 1, which implies the equipment i B. Train Supervision and Real-time Positioning alarm appear. According to index grade, The status of train is basic information of URT network ∃Δ > 0 and Δ ≥ η , η means predefined advanced alarm, i i operation, as well as it includes arrival time, departure time so the alarm reaches danger level and emergency response and stuck alarm in tunnel. Applying real-time positioning is started. techniques, motivation of trains can be shown on map with The level of alarm is identified as different value in timetable of current train. Also, visualized collocation of these supervision systems, especially in PSCADA and rescue resource adopts devices location with their features FAS.Pre-alarm for PSCADA includes voltage and current to attain emergency and coordination with other traffic of electric switches and their protected operation, while tools, based on geography information. Hierarchical they are beyond of normal range. So rating values for structure is applied to describe motivating trains and their voltage and current are defined as u0 and I 0 in respective, relations.

Trains are defined as triple array Μ = {}Π, E,Φ . Π is protecting system is employed to prevent crush between first step to present all trains on the line, containing sequent trains. In order to solve this problem, real-time individual attributes like train number, order, train type, calculation and estimation are combined to find location of direction etc. E is second step to token stations followed trains actually to recognize their displacement through GIS by every train, which relate to station name, arrival time, technologies for pre-alarm of trains. The pre-alarm and departure time of corresponding train. Π and E estimation should be divided into two aspects: are bond to hierarchical structure. Φ means attribute of • Alarm for train stop in the range of sequent stations. train, e.g. loading rate, train type, capacity. Define lines In normal, train stops in section without demand from set L to certain transfer station as L = {}l , ,l , ,l . transport managers will be regarded as abnormal. The 1 L i L m interval of collect data of trains from Automatic Train Therefore, all of trains in line li can be expressed Supervision system (ATS) is defined as tc , the default as ′ , means transfer Π ()li = {Π1,Π 2 ,L,Π j ,L,Π n } ε h value is set as 0.5 seconds. The location of train related to station. All trains are totaled as section of track can be converted as GIS T T m n z coordinate (xO , yO ), and the updated location is (x , y ) ∑∑⎛ ∑ ⎞ Ω = L⎜ Π()E()ε k ⎟ (2) i==11⎝ j=1 k ⎠ in the next interval of collecting data after the train running for the same time. If y = yT , vt ≠ 0 , then τ =1 ,else Principle of train positioning on map is called “double O c corresponding law”, which means real-time train location τ = 0 . τ is used to expressed as alarm for train, and v is relates to rail section position, and further corresponds to average speed of train which running in this section of rail coordinate of longitude and latitude on the map [9]. track. Especially, some of sections like turnout, in and out of In specific, it means once a train has not any storage, and area of multiple platforms, have appropriate displacement in intervals, the system thinks the train is relation that one location relates to several rail tracks. Pick non-normal status and corresponding alarm is put out. The up the milestones at center of station as the base point, thus module of alarm for train stop in sections is shown in we can calculate trains location on the map. In order to Figure2. separate train location from each other in rail layout, we divide sections into types to build mapping sections and milestones as Figure1.

Figure2. Alarm estimation of train movement in sections

Figure1. Principle of train positioning • Pre-alarm for timeout of train stop in station.

The parameter is defined as tstop , to determine whether So r()μ is expressed as mileage for section μ , and a train stop beyond normal time or not, which default time E is station set, while total mileages can be tailored as is 5minutes. The coordinates of start point and end point of li a station are meet for xi ∈[X s , X e ], yi ∈[]Ys ,Ye , and all z ⎛ ⎞ R = ∑∑∑ r μ′ θ + r μ′′ θ (3) track sections, described as set μ()S = {μ1,μ2 , μs }, are in li ⎜ () ()⎟ L εε=→1⎝ μ′∈[]εa →εb μ′′∈[]εa b ⎠ range of the station. If ntc > tstop , n =1,2,3,L , T T T ε a ,ε b ∈ Ez θ is variable of 0 or 1, if it is in region of x ∈[X s , X e ] , y ∈[Ys ,Ye ] , and μ ∈ μ()S , then the turnout then θ = 1 , else θ = 0 . As the analysis result, train has not leave the station in time of t . According to actual totaled mileages of a train must not be smaller than stop the difference value from the first station to the last estimation of train movement distance, we can achieve anyway R ≥ r l . Thus, calculated location μ andθ can outcome of alarm for stop timeout. li ()i be used to obtain spatial orientation on the map. III. DYNAMIC ANALYSIS FOR PASSENGER FLOW The Auto Transit Control system (ATC) is often used to inspect and direct trains in URT nowadays, but The statistic for passengers, based on Orientation and demander of transportation has no way to realize fault or Destination (called O-D) station, is counted for any station, malfunction of trains beforehand, in spite of auto transit

© 2013 ACADEMY PUBLISHER JOURNAL OF SOFTWARE, VOL. 8, NO. 4, APRIL 2013 881 according to units of time. In time of t , getting in called the transport coordination proportion to describe the O t O D t D ratio of transport capacity and passenger flow. passengers as p = ∑ p , and leaving as p = ∑ p . ek 0 ek 0 t0 t0 θ = Q P (7) The number of transfer passengers serves as important H H reference to identify passenger density, in order to adjust ∑ PH (tx ) = pH (tx ) (8) train plan like increasing or reducing train numbers on tx∈[]Ta ,Tb lines. Define route set Δ = Δ ,Δ , ,Δ , , O→D {}1 2 L i L Ifθ > 1then the capacity can meet for the requirement, weightW = {}ω1,ω2 ,L,ωi ,L , means route Δi has weight and while the larger the value that can transport more. ofωi . The weight is related to clearing algorithm between Thus, the model of transport coordination optimization lines, which considering factors of the shortest route, time, is built as below: and economic, applying fare clearing rate here. min f = min QH − PH (9) Define p HI as transfer passengers, from station ε k up min down min HO Constraint t ≥ t ， t ≥ t ； t ,t ∈[T ,T ]； array ε to ε in order, and p as those of transfer out liG G liG G 0 x a b 1 k εk t D ≥ T ， t F ≥ T ， t D ≤ T ， t F ≤ T of station ε k from station array ε h to ε z . So at the range of x a x a x b x b T toT , totaled numbers of transfer passenger as Parameters are predefined to evaluate effect that make a b it possible to analyzing and controlling vehicles t t HI t HO deployment, especially in passenger flow peak or special p = ∑∑ p + ∑∑ p (4) ε h εk εk t01εk∈[ε ,εh )(t0 εk∈ εh ,ε z ] period like city celebration. As far as passengers in hub are concerned, coordination B. Connection with Other Traffic Tools passenger flow can be calculated by AFC record. However, URT transfer hubs are often connected to the main road transfer passengers from other traffic tools, related to reach in city as well as passenger distribution center links to rate of specified tools, can be divided as three cases: passenger flow channel in town or suburb. Relatively, • Possessing cards. It is counted by loading those linked traffic tools include bus, taxi, plane, and ship. passengers of every reach of vehicles with planned The connection gives rise a problem for operation that how cycling timetables, fitting for direct transfer without to integrate different timetable for respective traffic tools pay again. Those of passengers cards which need without losing cost. The rule that connecting time and pay again will be counted for another time. density should be preplanned focusing on regular • Without cards. Reach of these passengers is treated passenger flow is proposed. as random variable, through simulating normal In the condition of emergency, temporal bus and more distribution. intensive vehicles will be added, while dynamic Therefore, total of passengers is expressed as information is used to organize connection plan, in order to achieve most effective route for evacuation. Define routes t t t P = pO + p HI + p HO (5) set as Δ = {Δ ,Δ , ,Δ , } in range of O-D. The H ∑ 0 ∑∑ εk ∑∑ εk O→D 1 2 L i L t t ε ∈[ε ,ε )(t ε ∈ ε ,ε ] 0 01k h 0 k h z strategy of connection plan contains two aspects: It contains three values: entering passengers, transfer • Parallel O-D route for evacuation passengers between rail lines, and transfer passengers from In direction of Δi (O, D) , there is any public transit other traffic tools. setV = {v1,L,vi ,Lvn } that relates to roads in the distance of λ with Δ ∩V ≠ φ , which means common or parallel IV. TRAFFIC COORDINATION route in the area of O-D station with weight ofωi . As more A. Coordination between URT Lines than one route appears, the weight Transfer hub is often called “the bottle neck” for set α = {α1,L,α i ,L,α n } is defined, which is order by transport operation to meet for requirement of passengers. more overlap and longer routes. Then evacuation In order to ease passenger flow in short time, enough V V ∑ capability conveys Q = qi ()vi ⋅α i , aiming to reduce transport capacity for transfer hub is needed [10, 11]. In the v∈V time set []Ta ,Tb , cycle of trains are set passenger load through the same or similar routes. If D ∃PH ≥ QH as Π′()li = {Π1,L,Π x ,L,Π n }, reach time is tx , and leave time t F . In direction of up and down, transport capacity V x max f ′ = maxQH ()PH − QH (10) within []T ,T of trains via the hub is expressed as a b • Vertical O-D route for nodal connection up down On purpose of passengers leaving transfer station in the ⎛ Πn up up up Πn down down down ⎞ ∑∑∑ ∑ QH = ⎜ ci si vi + ci si vi ⎟ (6) lT∈∈L ti[]T , i=Π =Π shortest time, coordination serves as primary approach to ib0 a ⎝ 1 1 ⎠ reduce accumulation effect of passengers. The field is

L is transfer station set, ci is basic capacity of vehicle, attained through transfer hub ε h , treated as center and the

si is group number, vi is loading ratio from ATS. So θ is

radiusγ that includes routes setY = {}y1,L, yi ,L yn . In order to accelerate passenger flow, objective function is

Y t D ⎛ ∑∑ ⎞ max f ′′ = max⎜Q PH ⎟ (11) ⎝ tL0 li∈ ⎠ The algorithm takes transfer stations as node layer of URT network, multiple connection banding areas are formed on the basis of distance from near to far.

V. DSS OF SUPERVISION AND OPERATION Considering requirement of information sharing and derivative information for emergency decision making, a hierarchical Supervision and Operation Decision Support Platform (SODSP) based on network operation is built, and supervision of trains or equipment, resource allocation optimization are included.

A. Structure of Multiple Hierarchical Platform Figure3. Structure of SODSP SODSP is designed for three layers of users, including stations, rail lines, and network center. Firstly, entire • Network and hardware. supervision information comes from this layer, which is SODSP adopts double net to attain reliability of collected into data exchange center, then disposed transmission. Excellent servers of application and GIS in information is converted into network center to produce high available mode can insure superior flexibility and operation decision schema, by applying algorithms before dependability. mentioned. • Data resource and processing. Business of URT network operation is set up as Crucial data includes real-time data, history data, and corresponding function modules with actual workflow. In transfer data from different systems, which is collected and specific, information collection module contains monitor disposed into united format in United Information alarms, trains status, and statistics, and network operation Processing system (UIP) to meet for respective application module includes evaluation, passenger analysis, and in SODSP. pre-alarm evacuation. The core module covers accident • Application service. treatment process that involves duty tasks, rescue resource Common applications offer GIS map query, remote collocation, emergency command, treatment evaluation, video meeting, CCTV, and information release. CCTV and drill training. Moreover, fundamental modules of GIS system screens video images of station hall or entrance to application and message dissemination are used to collect the scene in real time. Information release system common components for map and notification tools. provides notification and report while accident happened, All of data derived from Integrated Supervisory and and also acts as application service of daily forecast. As for Control System (ISCS) can be collected and saved in special emergency application performs on normal database, which supplies resource for Information Sharing procedures, relative independent node is extracted to sign platform logically [12]. The platform provides all sorts of different role, which allows cite or modify certain data disposal tools, including message queues, data access, information and the nodes posses previous or succinct XML data, custom APIs, data synchronization, GIS rest, relations, to secure exactness of procedure continuity for etc. The frame of SODSP is as Figure3. respective users. These processing data will be separated into two parts, Static data will be collected from supervisory systems one for online database, and the other for history database for decision making. It also supervises potential dangers that links to virtual tape library. Some of business data is and important protected facilities, e.g. flammable material, abstracted to show the current state of URT network, just gasoline nearby stations. In addition to modules of as trains or equipment operating mode. In order to supervision and operation management, there are special accelerate query of application records, the strategies of subsystems for information analysis are designed. retrieval like writing stored procedures and building index have been widely adopted. B. Information Sharing and Data Exchange The crucial techniques of large information sharing in SODSP are employed to solve problems of real-time and standardization. Collection cycle of supervision information is set about 0.5 second based on data exchange protocol. Another approach of transmitting data, like applications, data files, object, depends on message queue technologies that acting as container to finish collecting, translating, and filter in order to shield difference between

© 2013 ACADEMY PUBLISHER JOURNAL OF SOFTWARE, VOL. 8, NO. 4, APRIL 2013 883 hardware, database, or communication format, etc. SODSP rail lines and other traffic tools. At first, the quantity of adopts combination of Client/Server and Browser/Server, passenger PH on unit of date in May, is accumulated in the just for respective subsystem and data exchange. range of week, the results can be shown as Figure 5. As large scale processing and higher throughput, UIP defines XML as standard format to exchange data to secure transmission reliability through introducing message queue that dispose information when moving data from orientation to destination. It supplies delivering, collecting, and routing for data in order to realize application integration. The prominent performance lies in obviating differences in hardware, database, message and communication protocol, which determines convenient and swift correspondence abilities between applications. Due to security level of network that running supervision control system is often higher than other system like OA, we apply independent server to serve as front processor, which finish duty of collect resource data, provided by special supervision systems running in producing network.

Figure5. Passenger accounted in SODSP

After analyzing passenger flow on the network, several transfer stations with gathering large amount of passengers will be found, including Civic Center station, which acts as main cooperated object in platform. Then linking time of trains for the transfer station is calculated to build coordination between the rail Line 2 and Line 4, and more often than not, part of trains could get to be coodinated. The transfer walking time of pedestrian is defined as average value, 5 minutes from Line2 to Line 4, and 4 minutes from Line 4 to Line 2. Focusing on passenger peak hours, interval of trains in Line 4 is 5′24〞, and 8′in Line 2. Then algorithm mentioned before is employed to get time of the first cycle of trains and the result is shown as Table 1. So we can find Line 4 is easier to build transfer linkage to Line 2, otherwise the opposite, because of longer interval of Line 2. Figure4. Process of data disposal in supervision systems TABLE I. In the process of data disposal in SODSP, there are COORDINATION TIMETABLE IN PEAK HOURS four crucial steps in Figure 4 as follows: (1) information Wait Reach Time in Civic Center Station collection with cycling read from supervision systems; (2) time Linkage Line 2 Line 4 Data disposal and analysis for saving in database server; (3) direction second Up Down Up Down retrieval data of operation and equipment is transfer to application server; (4) According to alarm information, 146 4→2 × 8:12:38 8:05:12 × business subsystems analyze and evaluate for users 207 4→2 × 8:12:38 × 8:05:11 correlated to respective rights. 271 4→2 8:14:39 × 8:05:12 8:10:35

VI. SYSTEM IMPLEMENTION 99 4→2 × 8:21:38 × 8:15:59 In our project, SODSP is applied to emergency center 22 2→4 × 8:12:38 8:17:00 × for certain city. The net for data transmission lies in private 235 2→4 × 8:04:38 8:10:36 8:10:35 network channel, using dual-net redundancy, higher available clusters, multi-levels storage, and nested backup 164 2→4 8:14:39 × × 8:21:23 to attain security. The system also integrated several independent systems, such as message dissemination system, video meeting system and CCTV monitoring SODSP collects supervision equipments from five rail system, to supply data for decision support. lines with 130 stations. It collects equipment status from A case of passenger full accident in Civic Center supervision system, e.g. PSCADA, BAS, FAS, ATS station is picked up to simulate operation coordination of system. The four-grade alarm is designed according to

© 2013 ACADEMY PUBLISHER 884 JOURNAL OF SOFTWARE, VOL. 8, NO. 4, APRIL 2013 extent of malfunction, with alarm report screened if it [5] G.Z. Konstantinos, N.A. Konstantinos, George M. Vasilakis, accepts alarm data. The amount of passenger is “A Real-time Decision Support for Roadway Network proportional pictured on rail lines of map based on data Incident Response Logistics”, Transportation Research, Part received from AFC system. Analysis of passenger flow is C, No. 10, pp. 1-18, 2002. used to establish train plan and time tables according to [6] K.G. Zografos, K.N. Androutsopoulos, “A decision support coordination schema in transfer stations, and more system for integrated hazardous materials routing and emergency response decisions,” Transportation Research, important, passenger forecast can also be executed for Part C, 16(6), pp.684-703, 2008. correct directions of transport. On the condition of crowed station, the commander [7] MIGUEL A. de Miguel, Bernard Pauly, Thierry Person, et al. “Model-Based integration of safety analysis and reliable sets out evacuation directions. SODSP is started through software development,” Proc. 10th IEEE International definite rescue process, in which emergency scheme and Workshop on Object-Oriented Real-Time Dependable resource allocation are involved. Connection with other Systems (WORDS’05), IEEE Press, May. 2005. traffic tools with their routes relies on maps that can screen [8] ZHANG Ming, WANG Fuzhang, LI Ping. “The Platform of resource location and behavior records, which facilitates Network Operation Decision Making and Emergency for commander to master rescue resource layout and find out Urban Rail Transit,” China Railway Science, vol 33 No.1, rescue program quickly, as Figure 6. pp.113-119, 2012. [9] BRADLEY M. Estochen, Tin Strauss, Reginald R. Souleyrette, “An Assessment of Emergency Response Vehicle Pre-Deployment Using GIS Identification of High-Accident Density Locations,” Transportation Conference Proceedings, Ames, Iowa, pp. 221-226, 1998. [10] ZHANG Ming, DU Shimin. “Transfer Coordination Optimization for Net Operation of Urban Rail Transit Based on Hierarchical Preference,” Journal of the China Railway Society, vol 31 No.6, pp.9-14, 2009. [11] ZHANG Yushi, CHEN Xumei, YU Lei, et al. “Study on Model of Coordinated Operation between Urban Rail and Bus Systems at Transfer Stations,” Journal of The China Railway Society, vol 31, pp. 12-19, March, 2009. [12] NIKOS Zarboutis, NICOLAS Marmaras. “Searching Efficient Plans for Emergency Rescue through Simulation: • Figure6. Rescue resource collocation in SODSP the Case of a Metro fire,” Computer Science, vol 6 No.2, pp. 117-126, 2004. SODSP shows usefulness of supervision and emergency prevention and its superiority in establishing rescue scheme in actual accident treatment since installed Ming Zhang was born in 1979. She received her Ph.D degree in in URT emergency center. We have other problems like Tongji University in 2008. Now, she is the Associate Researcher how to intelligently control equipment based on analysis of Institute of Computing Technologies of China Academy of results interacted with users and optimizing speed for Railway Sciences in Beijing. She is author/co-author of more than further explore. 20 publications in transportation computing system and optimization theory. Principal research interests: intelligent transportation system and emergency planning in urban transit ACKNOWLEDGMENT system. This work was supported by a grant from the Key Projects in the National Science & Technology Pillar Fuzhang Wang was born in 1962. He is a Researcher and Ph.D supervisor in China Academy of Railway Sciences. His research Program of China (NO. 2011BAG05B02). interests include traffic information and control.

REFERENCES Ping Li was born in 1972. She is a Researcher in China Academy of Railway Sciences. Her recent research interests are intelligent [1] ANDRE Zerger, David Ingle Smith, “Impediments to using GIS for real-time disaster decision support,” Computers, emergency system and application. Environment and Urban Systems, No. 27 pp.123-141, 2003. Yingjie Wang was born in 1968. She is a Researcher in China [2] H.C. José, Custavo, A, Gustavo Alonso, “A multimedia Academy of Railway Sciences. Her recent research interests approach to the efficient implementation and use of focuses on GIS and spatial analysis computing. emergency plans”, IEEE Multimedia, vol. 11, No. 3, pp. 106-110, 2004. [3] M. Morin, Jenvald J., Thorstensson M, “Computer- supported visualization of rescue operations”, Safety Science, No. 35, pp. 3-27, 2000. [4] CHEN Xiu-wan, YANG Kai-xin, FANG Yu, “Study on urban emergency response system based on spatial information technology,” Geography and Geo-Information Science, vol. 19, pp. 49-52, 2003.