Main Page/SECURE

Semantics Empowered resCUe enviRonmEnt (SECURE)

This is a project in collaboration with the EE department. The goal of this project is to show how use of semantics can benefit unmanned or machine mediated rescue operations. We consider the problem scenario first and then propose possible solutions using a combination of semantic web technologies and a robot.

Error creating thumbnail: File with dimensions greater than 12.5 MP

What is the problem?

We consider a disaster scenario to exemplify the impact of using semantics in machine mediated rescue operations. Figure 1 depicts a disaster scenario at a chemical plant. Due to an attack on the plant, there is chemical spill and fire. In such a hazardous situation, directly mobilizing rescue personnel such as fire fighters, paramedics, etc, can be dangerous for two reasons: 1. Chemical spill may result in hazardous situation for rescue units. 2. Rescuers may not prepared to deal with specific scenarios due to lack of information (Situation Awareness). The overall goal in an disastrous scenario is to rescue lives, and minimise environmental impact.

What is our solution?

We propose a solution to this problem which addresses the two important points mentioned above.

Resources at our disposal

• Static sensor network: A chemical plant may have a variety of senors to detect fire, smoke, chemical, etc. Due to the damage caused by the fire, some of the static sensors might be damaged as shown in Figure 1. However, we can use the sensors that are still functional.

• Building plan: This provides information on critical resources in the place of attack, allowing the rescue team to focus on the most important tasks. e.g. plan a rescue path through the building.
• Inventory information: Provides details on type, quantity, and location of chemicals.

• Knowledge from curated sources: This is from domain experts like fire fighters, who possesses knowledge on type of extinguishers used for different types of inflammable materials.

A step by step solution:

• Gather as much information as possible using the static sensor infrastructure.
  • Damaged sensors can provide us an idea of the impact area.
  • Among the functional sensors,
    • the sensors that are not triggered indicate no fire.
    • sensors that are triggered, indicate fire.
  • Power of Abstraction [1]:
    • [Q without abstraction]: Are the sensor S1, S2, S3, S4 and S5 triggered? (let’s assume that these sensors are in a hallway between processing plant and pantry in Figure 1.)
    • [Q with abstraction]: Is there fire in hallway connecting processing plant and pantry?
• Send a robot inside the chemical plant instead of a human for initial investigation.
  • Robot would have sensors like camera, temperature sensor, sonar sensor, etc.
  • Has access to the static sensors, inventory, building plan, and knowledge from curated sources.
  • Possible steps that the robot can take:
    • Find all the rooms that might have many people (from domain knowledge)
    • Query static sensor infrastructure for finding known fires if any.
      • This might result in paths that may have fire if the static sensor infrastructure if unable to detect fire.
      • Since the robot has on-board sensors, it can detect fire.
      • For instance, consider a query for hallway between Room 4 and 5. There may be two results: H1 (hallway between Room 4 and 5) and H2 (hallway on the right side of Room 4). The static sensor might mislead the robot since the fire could not be detected by static sensors. Robot might start taking the path (H1), but since it has on-board temperature sensors, it can detect fire and take an alternative route (H2).

How are rescue units better informed?

• Once the robot makes its way in, it can send this path that is secure to the rescue units so that the rescue units can focus their attention on saving lives without investing any efforts on finding secure paths.
• The robot can check each room for presence of fire using sensors on board and use the background knowledge to indicate proper type of extinguisher to be carried by the fire fighters.

The robot’s knowledge base can be modelled using affordance [2], providing us a systematic way to represent the domain knowledge.

Demo

• Rescue robot detecting abstractions like "No Fire", "Fire", "Room Heater" from raw sensor data.

References

1. Harshal Patni, Cory Henson, Amit Sheth, Pramod Anantharam, “From Real Time Sensor Streams to Real Time Event Streams”, Kno.e.sis Technical report.
2. J. Ortmann, W. Kuhn, “Affordances as Qualities,” In Formal Ontology in Information Systems Proceedings of the Sixth International Conference (FOIS 2010), Vol. 209 (2010), pp. 117-130.