Underground Structure Monitoring with Wireless Sensor Networks

Underground Structure Monitoring with Wireless Sensor Networks

Date: 06th Dec. 2007 Presenter: KM Chen

Mo Li, Yunhao LiuHong-Kong University of Science and Technology{limo,liu}@cse.ust.hk

Outline

Motivation Overview of Structure-Aware Self-Adaptive sensor system

(SASA) Detecting and locating the collapse hole Accident reporting Displaced node detection and reconfiguration

Hardware Application Scenario Experiment and Performance Simulation Related Work Conclusion

Motivation

Over the last decade, collapses account for more than 50% of fatalities in U.S. in coal mine

The unstable nature of geological construction in coal mines makes underground tunnels prone to structure changes.

Environment monitoring in underground tunnels had been a crucial task to ensure safe working conditions in coal mines.

There is a need to develop a wireless sensor network system to quickly detect the collapse hole regions and accurately provide location references to evacuate workers from the dangerous zone.

Outline




Design of SASA (1/3)

a) Stationary sensor nodes are deployed on the walls and ceiling of tunnels to form a mesh network.


b) - Unfold 2 walls of the tunnel and builds a 2-D representation of the 3-D deployment on the inner surface of the tunnel.

- Nodes are configured with 2-D coordinates on the unfolded 2-D surface then transformed into the 3-D corresponding locations with the knowledge of longitudinal section.


c) - The distance between 2 nodes in the 3-D real environment is less than or equal to the distance between the pair in the unfolded 2-D view.

- The real connectivity of the sensor network is no less than shown in the 2-D representation.

Definitions and Theorem Edge node: A node defines itself as an edge node if the 2

adjacent neighbor nodes are detected lost during a time period.

Hole Polygon: The largest polygon outlined by the collapsed sensor nodes with every edge ending at 2 adjacent nodes.

Theorem: The convex hull of edge nodes in SASA encloses the hole polygon.

Edge node (2 adjacent neighbors are detected lost)

Convex Hull Hole Polygon

Detecting and locating the collapse hole

Collapse hole

Goal: Let sensor nodes to maintain a set of their neighbors. When nodes detect a loss of neighbors, a hole is detected.


Question: How to maintain a neighboring set?

Node Beaconing MechanismNode Beaconing Mechanism

Each node maintains a neighboring nodes list in Each node maintains a neighboring nodes list in memory.memory.Each node periodically broadcasts beacon Each node periodically broadcasts beacon messages that include its location.messages that include its location.Upon receiving a beacon message, the node Upon receiving a beacon message, the node updates the corresponding entry.updates the corresponding entry.If a node fails to update an entry in a fixed time If a node fails to update an entry in a fixed time interval, then it represents the loss of the neighborinterval, then it represents the loss of the neighbor


Problem:Problem: It was observed that the neighbor set of a node is highly It was observed that the neighbor set of a node is highly unstable, even if all the nodes work normally.unstable, even if all the nodes work normally.

Solution:Solution: SASA deploys sensor nodes in a cellular hexagonal SASA deploys sensor nodes in a cellular hexagonal placement such that the node distribution is uniform.placement such that the node distribution is uniform.Every pair of adjacent nodes are separated by the same Every pair of adjacent nodes are separated by the same interval .interval .Each node is limited to maintain a neighbor set to the 6 Each node is limited to maintain a neighbor set to the 6 adjacent nodes.adjacent nodes.

How to define a hole detection?How to define a hole detection? Only failure of at least 2 adjacent nodes are Only failure of at least 2 adjacent nodes are necessary to define an edge node. Nevertheless, if necessary to define an edge node. Nevertheless, if 2 adjacent nodes fail simultaneously, a hole is 2 adjacent nodes fail simultaneously, a hole is detecteddetected

What about single node failure?What about single node failure?

Unfortunately a small hole affecting only 1 sensor Unfortunately a small hole affecting only 1 sensor node can not be detected.node can not be detected.


Accident reporting

Goal:Goal: When edge nodes detect a hole, they report to the sink When edge nodes detect a hole, they report to the sink with the locations so that the hole can be illustrated by with the locations so that the hole can be illustrated by calculating the convex hull.calculating the convex hull.

Problem:Problem:Create traffic peak and increase collision domain.Create traffic peak and increase collision domain.

Solution:Solution: Randomized Forward Latency and Data Randomized Forward Latency and Data AggregationAggregationInsert a flag into the beacon messages, which indicates Insert a flag into the beacon messages, which indicates whether the beaconing node is an edge node. whether the beaconing node is an edge node. Upon receiving other edge nodes’ beacon message, an edge Upon receiving other edge nodes’ beacon message, an edge node records them locally.node records them locally.When this edge node sends out its report message, it When this edge node sends out its report message, it aggregates all the recorded locations of its nearby edge aggregates all the recorded locations of its nearby edge nodes.nodes.If an edge node receives a report message containing its If an edge node receives a report message containing its own location, it simply forward this message instead of own location, it simply forward this message instead of creating a new one.creating a new one.The sink will send out reply to limit the number of The sink will send out reply to limit the number of retransmission.retransmission.

Displaced node detection and reconfiguration

Goal:Goal: Rapidly detect displaced nodes and reconfigure with Rapidly detect displaced nodes and reconfigure with correct locations in order to maintain system validity.correct locations in order to maintain system validity.

Centralized approachCentralized approachWhen the sink receives report messages with the edge When the sink receives report messages with the edge nodes locations and approximate the hole region, it nodes locations and approximate the hole region, it broadcasts the convex hull area.broadcasts the convex hull area.Every node within the convex hull will start detecting its Every node within the convex hull will start detecting its surroundings and check its location from beacon message.surroundings and check its location from beacon message.If the 2 locations differ beyond some threshold, then it If the 2 locations differ beyond some threshold, then it knows it’s being displaced.knows it’s being displaced.

Distributed approachDistributed approach There are 3 types of edge nodes:There are 3 types of edge nodes:

1)1) Edge nodes that lose neighbors but themselves do not moveEdge nodes that lose neighbors but themselves do not move Since their locations are correct, they don’t need to be Since their locations are correct, they don’t need to be

reconfiguredreconfigured

2)2) Edge nodes that fall into an area where no normal node existsEdge nodes that fall into an area where no normal node exists They have no impact on normal nodes, they do not need They have no impact on normal nodes, they do not need

to be reconfigured either.to be reconfigured either.

3)3) Edge nodes that fall into other normal node rangeEdge nodes that fall into other normal node range Stop beaconing . This operation will lead the neighboring Stop beaconing . This operation will lead the neighboring

displaced nodes to become edge nodes, if they are not displaced nodes to become edge nodes, if they are not yet.yet.


Centralized approachCentralized approach

Advantage:Advantage: Short latency when the hole is closer to the sink Short latency when the hole is closer to the sink

Disadvantage:Disadvantage: May suffer long latency and low accuracy due to May suffer long latency and low accuracy due to high link loss rate in coal mine, especially when a collapse area in a high link loss rate in coal mine, especially when a collapse area in a long tunnel is far from the sink.long tunnel is far from the sink.

Distributed approachDistributed approach

Advantage and disadvantage: Advantage and disadvantage: Independent of the distance Independent of the distance to the sink.to the sink.

In summary:In summary: Combining both algorithms provides efficient and reliable Combining both algorithms provides efficient and reliable for various situationsfor various situationsTurn them off or reconfigure their locations to conform to Turn them off or reconfigure their locations to conform to their new positions.their new positions.


Location Calculation:Location Calculation:Suppose A and B drop into a new area surrounded Suppose A and B drop into a new area surrounded by 3 resident nodes.by 3 resident nodes.When A first detect the surrounding 4 nodes, it When A first detect the surrounding 4 nodes, it calculates a new location as (32.5, 19.25) and calculates a new location as (32.5, 19.25) and replaces the original locations.replaces the original locations.When B detects its surroundings, it utilizes the When B detects its surroundings, it utilizes the new location of A and calculate a new location as new location of A and calculate a new location as (15.63, 11.56).(15.63, 11.56).When A iteratively calculates its new location, it When A iteratively calculates its new location, it will get a more accurate result of (11.41, 7.14)will get a more accurate result of (11.41, 7.14)


(0, 0)

(20, 0)

(10, 17)

A (100, 100)

B (100, 120)

A (32.5, 29.25)

B (15.63, 11.56)

A (11.41, 7.14)


Outline




HardwareMica2 platform developed at UC BerkeleyMica2 platform developed at UC BerkeleyThe MPR400 radio board employed has a 7.3 MHz miThe MPR400 radio board employed has a 7.3 MHz microprocessor.croprocessor.128K bytes of program flash memory.128K bytes of program flash memory.512K bytes of measurement flash memory.512K bytes of measurement flash memory.868/916 MHz tunable chipcon CC1000 multi-channel 868/916 MHz tunable chipcon CC1000 multi-channel transceiver with a 38.4 kbps transmitting rate is emplotransceiver with a 38.4 kbps transmitting rate is employed for wireless communication with a 500 foot outdoyed for wireless communication with a 500 foot outdoor rangeor range

Outline




Application ScenarioCooperate with S.H. Coal Corporation and Cooperate with S.H. Coal Corporation and selected the D.L. coal mine as the experimental selected the D.L. coal mine as the experimental environment.environment.D.L. coal mine is the is one of the mist automated D.L. coal mine is the is one of the mist automated coal mines, yielding the second largest production coal mines, yielding the second largest production of coal worldwide.of coal worldwide.Slightly sloped 14-kilometer long main tunnel Slightly sloped 14-kilometer long main tunnel from the entrance above the ground surface and from the entrance above the ground surface and goes 200 meter deep underground.goes 200 meter deep underground.

Application Scenario

Requirements for SASA implementation in Requirements for SASA implementation in D.L. coal mineD.L. coal mine

1.1. Remote management:Remote management: Remotely maintain and Remotely maintain and manage the entire monitoring system, efficient manage the entire monitoring system, efficient and robust communications and routing and robust communications and routing mechanisms are required under all conditions.mechanisms are required under all conditions.

2.2. In-Situ interactions:In-Situ interactions: Besides stationary sensors Besides stationary sensors deployed on the walls, poles and floors, miners deployed on the walls, poles and floors, miners carry mobiles sensors providing real-time carry mobiles sensors providing real-time geographical references.geographical references.

3.3. Awareness of structure variations:Awareness of structure variations: Using Using node collaborating mechanism for collapse node collaborating mechanism for collapse detection.detection.

4.4. Maintenance of system validity:Maintenance of system validity: Maintaining Maintaining the validity of the monitoring system in extreme the validity of the monitoring system in extreme situation.situation.

Outline




A prototype system with 27 Mica2 motes is A prototype system with 27 Mica2 motes is implemented in the D.L. coal mine. implemented in the D.L. coal mine. It is distributed on a tunnel wall about 12 meters It is distributed on a tunnel wall about 12 meters wide and 5 meters high. wide and 5 meters high. Nodes are pre-configured with their location Nodes are pre-configured with their location coordinates.coordinates.Nodes are placed in hexagonal mesh regulationNodes are placed in hexagonal mesh regulation

Experiment and Performance

Hole detection percentage:Hole detection percentage: A hole is counted as A hole is counted as undetected if less than 3 nodes reports are undetected if less than 3 nodes reports are received by the sink.received by the sink.Hole detection error:Hole detection error: The error in distance The error in distance between the real and detected position of the hole between the real and detected position of the hole region.region.Reconfiguration error (2-D or 3-D):Reconfiguration error (2-D or 3-D): Localization error in the reconfiguration process.Localization error in the reconfiguration process.



Over 80% of the detected Over 80% of the detected holes are located within 1 holes are located within 1 meter from its real position meter from its real position and 99+% are less than 2 and 99+% are less than 2 meters.meters.

All the 2-D All the 2-D reconfiguration errors reconfiguration errors and over 80% of the 3-D and over 80% of the 3-D reconfiguration errors reconfiguration errors are below 3 metersare below 3 meters


Detection latencyDetection latency: Time from when the hole emerges until it is : Time from when the hole emerges until it is detecteddetectedTurn-off latency:Turn-off latency: The latency when the displaced nodes were t The latency when the displaced nodes were turned off.urned off.Reconfig latency:Reconfig latency: Latency when reconfiguring the displaced no Latency when reconfiguring the displaced nodes according to the normal nodes surrounding them.des according to the normal nodes surrounding them.

Short beacon interval leads to short latency. However, frequent Short beacon interval leads to short latency. However, frequent beaconing brings large overhead, heavy collisions and increased beaconing brings large overhead, heavy collisions and increased packet losspacket loss


The packet loss rate rapidly drops as the beacon The packet loss rate rapidly drops as the beacon interval increases while under short beacon interval increases while under short beacon intervals (0.8s), then becomes stable around a fixed intervals (0.8s), then becomes stable around a fixed level.level.The loss rate is heightened as the exerted traffic The loss rate is heightened as the exerted traffic overhead increases.overhead increases.


Observations:Observations: We can carefully select a proper beacon We can carefully select a proper beacon

interval for a specific application workload to interval for a specific application workload to balance communication quality and the balance communication quality and the processing latencyprocessing latency

Shorter beacon interval to reduce the Shorter beacon interval to reduce the processing latency if the application workload processing latency if the application workload is light.is light.

Longer beacon interval to reduce the packet Longer beacon interval to reduce the packet loss rate if the application workload is heavy.loss rate if the application workload is heavy.

Outline




Simulation

2000 nodes were simulated on a 1000m x 2000 nodes were simulated on a 1000m x 20m plane.20m plane.Nodes were placed in a hexagonal mesh Nodes were placed in a hexagonal mesh regulation with 3 meter interval between regulation with 3 meter interval between each node.each node.A transmitting rate of 16 packet/s is used A transmitting rate of 16 packet/s is used in the simulation for the nodes’ in the simulation for the nodes’ communication channels.communication channels.

Simulation

When the hole size When the hole size increases, the outline of the increases, the outline of the edge nodes becomes tighter edge nodes becomes tighter therefore the precision is therefore the precision is dramatically increased.dramatically increased.

Detection error is stable as Detection error is stable as slightly decreases as the hole slightly decreases as the hole size increases.size increases.Larger hole includes more Larger hole includes more edge nodes, giving a more edge nodes, giving a more accurate outline of the hole accurate outline of the hole region.region.

Simulation

As the packet loss rate As the packet loss rate between any 2 between any 2 communicating nodes, and communicating nodes, and the random node failure rate the random node failure rate increase, the misreport ratio increase, the misreport ratio also increases.also increases.Need to decrease the beacon Need to decrease the beacon frequency in order to frequency in order to preserve a better preserve a better communication channel.communication channel.

When the hole is close to the When the hole is close to the sink, the centralized sink, the centralized algorithm benefits from rapid algorithm benefits from rapid information collection and information collection and reaction from the sink.reaction from the sink.When the hole is far away When the hole is far away from the sink, centralized from the sink, centralized algorithm suffers from the algorithm suffers from the round-trip time from the sink. round-trip time from the sink. The distributed algorithm is The distributed algorithm is not affected.not affected.

Outline




Related Work

Wireless sensor networks for habitat monitoring [Wireless sensor networks for habitat monitoring [A. Mainwaring, J. Polastre, R. Szewczyk, D. Culler and J. Anderson]

A Wireless Sensor Network for Structural Monitoring [N. Xu, S. Rangwala, K. K. Chintalapudi, D. Ganesan, A. Broad ]

Hole problem in wireless sensor network [N. Amed, Hole problem in wireless sensor network [N. Amed, S. S. Kanhere and S. Jha]S. S. Kanhere and S. Jha] Coverage holeCoverage hole Routing holeRouting hole Jamming holeJamming hole Sink/black/worm holesSink/black/worm holes

Outline




Summary and Future Work

By regulating the mesh sensor network By regulating the mesh sensor network deployment and formulating a collaborative deployment and formulating a collaborative mechanism based on the regular beacon mechanism based on the regular beacon strategy, SASA is able to rapidly detect strategy, SASA is able to rapidly detect structural variations caused by underground structural variations caused by underground collapsescollapses

The collapse holes can be located and outlined, The collapse holes can be located and outlined, and the detection accuracy is bounded. We also and the detection accuracy is bounded. We also provide a set of mechanisms to discover the provide a set of mechanisms to discover the relocated sensor nodes in the hole region.relocated sensor nodes in the hole region.

How to organize mobile nodes to form efficient How to organize mobile nodes to form efficient collaborative groups is a challenging issue.collaborative groups is a challenging issue.

Single hole detection.Single hole detection.

Underground Structure Monitoring with Wireless Sensor Networks

Documents

Transcript of Underground Structure Monitoring with Wireless Sensor Networks