Infrastructure Development for a London-based Disaster Test Scenario

Ongoing project...

Objective

Map source: http://www.cityoflondon.gov.uk

Detailed map

Satellite images are a powerful resource to our modelling process because we can have a better notion about the region and its features. For example we can retrieval, with a reasonable precision, information about, for example, ground area of buildings, shape of buildings and roads, length and width of roads and general relative positions of such objects. For that it is important to consider the scale used by the image, in this case about 1cm to 100m in the real world. Note that, during the real disaster operation, the command and control center must consider a wider area thah that directly affected by the disaster. For example, essential resources such as hospitals will be away from this area. Map source: http://maps.google.co.uk See also the interactive satellite image for this region.

Scale: |_____________| = about 200 meters.

• Geographic Information Modelling Process
  
  
  • Step 1.: Scenario definition with acquisition of maps (with scale) that give a clear idea about the objects of this scenario (roads, buildings, river, open areas and so on);
    
    
  • Step 2.: Information retrieval of roads' properties. Such properties are: name, length, width, width for walkers, road kind (elevation road, bridge or tunnel) and number of lanes;
    
    
  • Step 3.: Information retrieval of buildings' properties. Such properties are: name/number, entrance position(s), number of floors, kind of building (wooden, steel frame or reinforced concrete), area of the ground floor and total area of all floors;
    
    
  • Step 4.: Use of a modelling tool that convert all this information in data to be used by the RoboCup Rescue simulator (node.bin, road.bin and building.bin). In our case, this tool is JGISEdit. Download and details about this tool are available here;
    
    
• Simulator/I-X Integration
  
  
  • Step 5.: Definition of a semantic layer that maps simulation information about the scenario objects to a common sense representation. an examples in this direction is the use of buildings and roads names rather than object simulation codes;
    
    
  • Step 6.: Implementation of skeleton agents that contains functions associated with the representational translation of data between the two platforms. For example, consider the geographic information format. While the simulator uses the orthogonal distance (x,y) from a pre-defined point to a specific object, in millimetres, to indicate the position of such a object; I-X agents use latitude/longitude values for that;
    
    
  • Step 7.: Organisation of the stuff related to the simulation execution on the I-X platform: map, object images and any other resource.
    
    

• The City of London official page;
  
  
• Streetmap for maps of UK cities;
  
  
• Mapquest also for maps of UK cities;
  
  
• The Google map resource for maps and satellite images;
  
  

The architecture that is being used for this initial version is illustrated bellow. Agents are divided into two groups: RCR agents and I-X agents. RCR agents are provided in the simulator package (yapapi) and represent very simple agents that can be used as a basis for more complex implementations. I-X agents are the components that we have implemented and that are being evaluated inside the RCR environment. The main elements of this architecture are introduced in follow:

Police forces (PF) - one of their function is to cover the environment and send problems to the Police Office, which forwards such problems to the appropriate agent. For example, if any fire is discovered by a PF, a message is sent to the fire station via the police office; Fire Station - accounts for receive the fire messages, creating new activities to be allocated to its fire brigades in an optimal way. This process can be manually carried out by users, or autonomously by the fire station agent. Handlers associated with this process can be provided as plug-ins; Fire Brigades - if they do not have activities to be performed, they tend to stay in refugee places, where they can refill their water tanks. When they receive activities from the fire station, in the form Extinguish ?building, they try to find the best route to the specific building place to extinguish the fire in such a building; Gray communication channel - uses the two basic RCR methods to perform communication between two or more agents. These two methods are: tell(message) and head(sender,message) Red communication channel - uses the two basic I-X method to transmit and receive messages: sendObject(destination,contents) and handleInput(message), where message can bean issue, an activity, a constraint, a report or a chat message. Blue communication channel - IX agents are, in this application, hybrid agents because they implement inner classes that represent RCR agents. In this way, the communication between IX and RCR agents are done via the RCR methods, however special functions are used to translate information between them because the internal formats used by RCR and IX agents are different.

We need to understand two forms of information translation performed during the blue communication. The base of facts of RCR agents (we are considering their implementation via the YAPAPI) is represented as a unique object (world) that contains several other objects, each of them modelling a specific object of this world (building, road, civilian, etc.). So the first method of translation consists in mapping the objects attributes to the I-X constraint format. An example is given below to a building:



Considering this example, we can see that the information translation is a direct and easy process. However, there is a problem associated with the position due to the different semantic used for each system (RCR and I-X). While the RCR simulator uses the orthogonal distance (x,y) from a pre-defined point to a specific object, in millimetres, to indicate the position of such an object; I-X agents use latitude/longitude values for that. Thus, we need to use a additional method of translation. For this case in particular we are using the regression method, a numerical technique of fitting a simple equation to real data points. This is possible because we know the set of position RCR values and their correspondent I-X values. Then, we just need to find a function that map a set to the other set. An example of result of this process is shown in follow.

public Double convertLongitude(int x) {      double m = 65.532;      double b = -48572.5553;      double longitude = ((m*(x/1000))+b)/100000;      return new Double(longitude); }

This is the function, codified in Java, that is being used to convert x values to longitude values. The mathematical tool used to obtain that function was Data Lab - Version 2.1, which is available in a limited version (evaluation copy) for free. The same tool can be used to convert y values to latitude values.

This method of converting positions works well for small areas, like City of London. The inconvenient detail is that it is not a general method so that different areas will need different equations. A more general method is to use the UTM (Universal Transverse Mercator) projection. This approach is likely to be explored in future works.

Step 6.: Semantic Layer

The Semantic Layer accounts for augmenting the information about the disaster environment. In this way, there are two specific ideas associated with this layer: (1) to provide human-directed information rather than internal code or reference of objects, and (2) to produce new information that can improve the simulation purpose. Functions in this direction are:

• Buildings' address - buildings are referred in the RCR as a unique integer value, such as 256. Thus, the attributes representation of this building using I-X constraints is ((attribute 256) = value). Note that this is not an appropriate representation for human users;
  
  
• Roads' names - as buildings, roads are also referred as a unique integer value. Furthermore, due to internal representation used by RCR simulator, the same road can be represented for a set of segments. For example, the Billter Street is represented for the set (412,411,486,485);
  
  
• Fire floor - the simulator just indicates the reference of the building in fire. An interesing information could be de levels that are in fire. This information could be used during the reasoning of fire brigade allocation so that the most appropriate agent could be sent to each fire site.
  
  

A brief explanation about the implementation of such functions are given in the table below.

Buildings' address functions

Couple of functions where the first receives a RCR building reference number and returns the address of this building. In a similar way, the second function receives the address and returns the RCR reference. Note that here the relation between the two sets is 1:1.

Roads' name functions

Also involves two functions: the first function receives a RCR road reference number and returns the name of this road. The second receives an array of road references (route), returning the corresponded road names for this route. Differently of buildings, here we have a 1:n relation between the sets. In other words, each RCR integer value corresponds to a unique road name, while a road name corresponds to a subset of integer values. The package documentation brings more details about this feature.

Fires' floors function

As this information does not exist in the simulator, it is generated in an aleatory way by the semantic level. This can be performed by a Random number generator function, which must observe the total number of levels of the building

Step 7.: I-X Application Package

I-London is an I-X application that evaluates planning and coordination ideas using the RoboCup Rescue Simulator and the London scenario discussed here. The I-London manual in follow, available in pdf and doc versions, gives details about the features and configuration of this application. The QuickStart document is a practical way of setting and running the application. For that, users must have installed the current version of the I-X package. Then the I-London Zip file (available soon) must be extracted in the ix-/apps directory.

Documents:

I-London Manual (PDF version)
I-London Manual (DOC version)
I-London Quick Start