Common scenario
From SIGI IRTG Wiki
Disaster Management (Earthquake)
In the early morning hours of 6 April 2009 a severe earthquake hit the city of L'Aquila and its surroundings, in the Abruzzo region of Italy, causing the death of 307 people and more than 1'500 injured. Between 3'000 and 11'000 buildings were damaged solely in the city of L'Aquila.
Typically, urban and wild environments are represented in GIS application developed by national or local administrations, industries, military forces, etc. Each of them stores in a GIS system all the geographical informations useful for their kind of target: a local administration could have a good description of roads, buildings, natural zones belonging to the environment; an industry enterprise wants to manage, in most cases, information useful for their kind of business; military forces, indeed, would like to store and manage information related to strategic points or risk zones within an urban environment. Obviously the cited application are only a small part of the global GIS purposes and applications.
With full access to all the datasets stored by the different administrations or industries GIS, it is possible to say that one can solve almost all kind of problems and can develop all kind of geographical applications using the informations stored in existing GIS.
Data changes are often slow processes and changes affect only a minimal part of the whole data stored in a GIS; let us imagine the process of construction of a new road or building that requires a long time to be completed and that affects only the neighbor of the new entity.
In this case it is quite easy to update GIS information: the only mandatory operation is to make accurate measure of the new entity and add it to the system, and eventually update information related to objects that was affected by the changes.
But what happens when an unexpected natural event, like an earthquake or a tornado, changes in few seconds the typical “static” environment? This kind of event can cause a lot of changes in the environment: a bridge can fall down, a landslide can occlude a road, etc.
In this case, different from the typical situation described before, data changes are no more slow processes and can affect a big part of the whole data managed by GIS.
Obviously these kind of changes make the environment description stored in a GIS system unusable; at the same time, after a natural catastrophe GIS data is particularly important as it is needed to coordinate the different aid operations. In particular, considering the example of an earthquake hits a city, it is possible to find a lot of tasks that need to start in the minutes after the natural event and need to be completed in the shortest possible time. Let's consider, for example, crucial operations like victims rescue, aid dispatch to survivors, organization of gather points and of management points for people who have to manage the catastrophic situation.
The requirement of first aid makes an efficient infrastructure necessary. The firefighters and rescuers rely on a flow of latest information. The operational command has to update its information to avoid losing time by having emergency vehicles reaching deadlocks.
With a good description of the environment, it is possible to take care of these kind of tasks reducing the risk for rescue missions and for survivors; in contrast, using information stored in a non-updated GIS, e.g. to find the shortest way to reach a zone where there is necessity of aid, there are intrinsic risks to obtain a path that can not be used anymore, because of collapses or landslide; this eventuality increase hazard for rescue missions and also increase the time to take aid to people.
The challenge is then to try to update informations on GIS systems in a fast way, using when possible automatic agents instead of humans and take advantage from the power of modern automatic calculators.
Modern observation systems make a quick surveillance possible. Remote sensing, unmanned aerial vehicles, but also reports by helpers and victims provide valuable information. Different groups, like the military, local administrations, the Red Cross and other aid organizations, universities or single persons can contribute valuable information. Most of these contributions can be grouped as sensory information: observations and measurements. An efficient integration of various sensory information sources across communities and different data models is crucial in this situation. The red cross does not only need to know how many people are injured, but they need to decide on where to set up their camps and mobile hospitals to achieve the best accessibility. Blocked roads have to be removed from the route planning. Helicopter landing fields have to be scouted. Medical supply has to be delivered. In such case medical information, demographic information, military information and personal information have to be integrated and made available to the institutions that need them. In time critical applications it is not only important to make all information available to all parties, but to support the retrieval of precisely the required information.
The International Research Training Group "Semantic Integration of Geospatial Information" addresses problems arising when integrating geospatial information across different communities and data sources to reason and support decision about the human environment.
Personal research within the scenario
Giorgio De Felice: Map reconstruction
My proposed contribution has as target to help the operation of reconstruction of the changed environment map.
Obviously a natural event, like an earthquake, can change a lot of entities that are part of an urban environment, but there are still some entities that may not have been affected by the catastrophe (e.g. a particular building or a particular historical monument). The main idea is to send some survey agents (humans or automatic) that start to explore the changed environment, and they describe it to a central system in a simple and natural “language” from their perspective. For example an agent can say that he “sees” a collapse “in front of” the “main square” of the city, or that there are survivors “between” two collapsed buildings. Of course each agent will use a different way to describe the environment depending on his background and his task. The presence of fire will be probably described in different ways by firemen and red cross operators.
The central system will then process the received informations and try to derive new information to help rescue and aid operations. One useful task is to reconstruct the new map of the environment as good as possible, making approximation when needed and trying to extrapolate new informations from the information given by the agents. This new map will be used for all the rescue operations cited before. Also a support to decision making tasks (e.g. path finding, settlements of rescue camps, etc.) can be furnished. Presentation of data can differ depending on the task the final user has to achieve and of course from his cultural background. Obviously the new information can not be precise, but still reduce risks for rescue missions and reduce time to send aid where it is needed, without scattering rescue forces in useless ways.
Problems involved in the process of description's reconstruction of the changed environment are various and of different nature. Below these problems are briefly described.
First of all the process of “recognition of known places” has to be considered or, in other words, the process that determine all the entities that did not undergo changes after the natural catastrophe.
The second task that can be identified regards how human and automatic agents communicate with the central system, furnishing information regarding the description of the environment from their point of view. Humans use typically a natural language to communicate and automatic agents use a “mathematical” language. It has to be noted that different human agents can refer to the same thing using different expression depending on their culture and tasks they are involved in. A fireman will describe a fire in a different way it is described by a red cross operator, due to his experience and knowledge in the field.
Humans expressions give typically qualitative descriptions instead of quantitative ones and also they give only a partial description of the entities. The system that reconstructs the map has to take these aspects into account.
Finally the last task is related to the final reconstruction of map. At this point it is possible to suppose to have a set of “known entities” (quantitatively described), a set of “unknown entities” and a set of relations that relate entities belonging to the two previous sets that give a qualitative description of the “unknown entities”. The target is to infer new information from the given one. One operation that can be addressed is to create a quantitative description of “unknown entities” as close as possible to the real characteristics of the entity. It is also possible to find out new qualitative relations among known and unknown entities. Problems to take into account concern the time performance requirements (the process of reconstruction has to be fast to help rescue missions) and also problems of consistence (descriptions given by different agents can contradict each other).
Within the proposed work only the last task will be addressed. The suppositions are that the description of “known and unknown entities” is available and also problems regarding interpretation of natural language and communications among humans and automatic agents are solved. Characteristics arising from this steps are still to be considered in the phase of map reconstruction, like the qualitative and incomplete description of environment; these characteristics will bias how the map reconstruction task will be developed.
Anusuriya Devaraju: Towards Process-based Ontological Approach: Breaching Semantic Barriers within Sensor Data
Environmental Sensors Networks or ESNs are networks of sensors that monitor qualities of phenomena in geographic space at different points in time. Recent technological advances have facilitated Environmental Sensors Networks or ESNs as a new way to study geo-phenomena dynamics. For example, sensor data takes the form of sequences of quantitative values of time-varying observables. Ongoing changes in any of these quantities created by processes. This indicates that the facts about processes in physical environment can be inferred from the low-level sensor measurements. However, this is not easy to achieve due to the nature of sensors which produce huge amount of data that comes with different format and structures. Despite the fact that ESNs are capable of providing users a more-or-less continuous record of measured values with respect to geo-phenomena, these data streams by themselves do not necessarily provide the semantics of data (Hornsby and King, 2008); These semantic barriers pose a challenge to the sensor data consumers who want to compile the data into a form that can be fed into decision support modelling and simulation tools.
In earthquake study, the data such as earthquake parameters (time, location and magnitude) and ground motion (acceleration, velocity, displacement and so on), paleoseismic data, fault database, and so forth, are provided by various seismological observatories, geophysical agencies and research institutes. Each data provider has their own interpretations. Although the data can be obtained, there is no guarantee that data carry identical, different or compatible meanings; their interpretation may depend on context and other factors. To enable earthquake forecasting, heterogeneous nature of earthquake data from various sources must be intelligently integrated. Ontology is a desirable solution for achieving semantic interoperability since it captures concepts of domain and provides foundation for discovering and resolving semantic conflicts in the underlying earthquake datasets.
My research aims at communicating and integrating data among multiple observatories using ontological approach. Although the use of ontology for data integration is not new, the study takes different approach by introducing a ‘process-based ontology’. This process-based ontology relates both to the semantics of the objects and to the semantics of the processes associated with earthquake phenomena. A formal, explicit representation of both processes as well as objects is essential (a) to hide differences amongst data and present unified view and (b) to explain, make prediction and to allow an effective mitigation of earthquake hazards. Thus, a process based ontological approach can breach semantic barriers as well as improve the understanding of earthquake processes.
Paolo Fogliaroni: Towards an Hybrid Quantitative - Qualitative Geographic Information System
"Broadly speaking, qualitative-reasoning research aims to develop representation and reasoning techniques that will enable a program to reason about the behavior of physical systems, without the kind of precise quantitative information needed by conventional analysis techniques such as numerical simulators. ... Observing pouring rain and a river's steadily rising water level is sufficient to make a prudent person take measures against possible flooding - without knowing the exact water level, the rate of change, or the time the river might flood." (Y. Iwasaki, Real-World Applications of Qualitative Reasoning)
Human-being, indeed, are used to reason with qualitative information rather than numbers. On the other hand, big part of actual artificial systems implements mathematic/numerical theories and calculi. This makes “tricky” for humans to interact with machines, as they cannot use them natural expressivity. Development of qualitative reasonings and them application to automatic systems, can facilitate the communication among humans and machines. This is one of the reasons Qualitative Reasonings have been quickly developing during the last decades. Nevertheless, although qualitative reasoning methods are rapidly emerging and developing in several areas, they are still relatively unknown in the Geographic Information field, indeed, up to now, just the topological aspect has been taken into consideration and integrated in the major part of GISs. Again, even if GISs comprise the topology, they are not optimized to answer to topological queries as processing the query will always go down to the geometric level. It is evident that, whether topological relationships would be directly stored within a GIS, actual topological query execution would be highly overcome in performances. Finally, other qualitative relations, direction and orientation, relative position, as well as what we can name "target-oriented" qualitative models, are completely unconsidered within nowadays GISs. This means for example that whether one would like to query a GIS on a qualitative requirement will have to translate the query at a geometric level.
If a GIS would explicitly include the storage of qualitative relations, it would provide a powerful instrument for operations and searches based on them. Indeed it would be possible to reduce efforts that today are needed to translate from qualitative representations (typically human) to mathematic/geometric representations (standard GIS). Furthermore, qualitative queries would also allow to use Geographic Information Systems in a completely new way. Indeed it would be possible to satisfy a big range of requirements that today cannot be accomplished. The users would be furnished with a high power system, able to satisfy more human-like requests. The system would become much more user-friendly leading to a range of novel kinds of spatial analyses that are impossible to imagine nowadays. The intrinsic properties of the system would also lead to the possibility to directly collect and store qualitative data avoiding a specific geometric description. Nevertheless it would be possible to reconstruct a geometric approximation directly from qualitative relations if sufficient qualitative information is available in the system.
If a hybrid GIS stores dataset regarding a specific location, a city, a forest and so forth, it would provide the basis for a fast dataset update when a catastrophic natural event occurs. Such a system would allow a rapid recontruction because it will be able to directly manipulate and store qualitative information as well quantitative ones. Furthermore, it will provide quicker answer to qualitative queries, being the qualitative nature of the spatial information directly stored, that means that when the system is queried on qualitative requirements, it will be possible to immediatly access and retrieve the right information. Finally, Qualitative Reasonings will provide further information by inferences that will be possible to do on already stored qualitative data.
Mareike Kritzler: Localization and Tracking of Moving Objects via Sensor Fusion
In my PhD I am working with spatial and temporal data obtained from moving objects, animals or human beings. I am focusing on actively tracking these things via different sensor sources. Furthermore I will fuse these heterogeneous data to obtain different kind of information concerning movement, behavior and location. It will be possible to find often used pathways, make predictions about future movements or behavior. Obtained information are scenario depended.
After a natural disaster has happened a lot of rescue teams are on duty. First of all it is important to know where they have to go and where they are. And furthermore they need instructions depending on their location.
In the case of an earthquake it would be possible to track the firefighters via GSM (Global System for Mobile Communication) and GPS (Global Positioning System) when they begin with their supporting measures and start to rescue injured people for example. The leader of a fire engine uses a cell phone or another mobile device which often is equipped with a GPS receiver. Also if there is a loss of the GPS signal the position could be determined by using GSM data. And if both data sources are available it would be possible to combine these two to have a good location result. Is the position of the firefighters determined it would be possible to give them location depending information on the basis of actual map information obtained from different services. If they know that their actual position is in a residential area where a lot of people while the earthquake lived, they can be guided to the singles house. And a service can provide information how many people they probably have to rescue. Or other important information can be given: if there are mobile agents collecting data about air pollution or chemicals or other dangerous stuff. On the on hand the exact position of rescue teams could be obtained and on the other hand location depended information can be provided for guidance and support.
Jens Ortmann: Integration of human observation and technical sensor observations/Affordances
In decision-making processes an understanding of the observations and the possible actions and especially of their relations is crucial. The theory of affordances provides an approach how perception and potential action are linked in humans and animals. Affordances describe the properties of the agent-environment system. We argue that modeling the properties of humans and the properties of the environment individually is not sufficient, but it needs an understanding of the emergent properties that results from the relation between human and his environment.
The scenario of an earthquake poses a very interesting and challenging problem to the modeling of affordances. The earthquake changes the environment within one time dramatically, basically resetting the agent-environment system and its affordances. New observations by technical sensors and by humans have to be integrated to remodel the environment for efficient decision making. Especially after an earthquake this can be a factor that decides over life and death.
My research will first address the question of how we can account for affordances with ontologies. An ontology is a logical theory that allows reasoning, e.g. testing the correctness of hypothesis or inferring new facts within your theory. For affordances we envision that you can infer affordances from properties of agents and properties his environment. And if we know the affordances we can make statements about the properties of the agent and properties of the environment. The changed environment in the earthquake scenario is likely to realize different affordances to the humans. Knowledge of the affordances like, whether a certain way is passable or whether a certain building still provides shelter is crucial in decision making for disaster management.
Secondly, we approach the integration of human observations. Especially in extreme situation humans can provide quick information. The first rescuers can communicate their impressions to the operation command within minutes. Victims can send messages for help from which we can draw conclusions on the environment. Again affordances play an important role in evaluating the human observations. A person calling for a salvage team is likely to be in an environment where things are buried, and where roads are blocked. In general we try to integrate a field-based technical sensors observations and object-oriented human observations within one infrastructure.
Finally, interactions with the environment are always coupled with our perception of the environment. Perception allows actions, but our possible actions also determine our perception, i.e. we see what we can do. Observations of actions by others can also give a hint on what these people perceived. Action-perception cycles play an important role in many system-theoretic applications. An analysis of complex processes, as for example in the environmental sciences suggests an investigation how perception and interaction relate to each other, how perception can be enhanced and how interactions can be enhanced. To the earthquake scenario this would mean that we model the observations, the affordances and reason about consequences of interactions. In a simple case one could decide which roads should be cleared first, or whether collapsing a damaged building would be helpful (for example if a landing site is required).
To sum it up, our contribution to help in the earthquake case is a model that allows an analysis and integration of human observations and technical sensor observations with special respect of human capabilities and needs.
