Ben Miller. A distributed algorithm is a type of parallel algorithm They are designed to run on...

22
Distributed Algorithms for Mobile Sensor Networks Ben Miller

Transcript of Ben Miller. A distributed algorithm is a type of parallel algorithm They are designed to run on...

Page 1: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Distributed Algorithms for Mobile Sensor Networks

Ben Miller

Page 2: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

A distributed algorithm is a type of parallel algorithm

They are designed to run on multiple interconnected processors

Separate parts of the algorithm are run simultaneously on independent processors

What is a Distributed Algorithm?

Page 3: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

What is a Distributed Algorithm? Each of the processors has its own memory

The processors send messages to each other

Page 4: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Wireless Sensor Networks are made up of a series of distributed sensors

The sensors collect data and pass it through the network

The sensors are designed to consume very little power

Wireless Sensor Networks

Page 5: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Wireless Sensor Networks Each sensor node typically consists of a

microcontroller, radio, and battery

Page 6: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Golden Gate Bridge◦ Accelerometers are used to collect data on the

bridge’s movement to monitor its structural health

◦ Researchers at the University of California at Berkeley used 64 sensor nodes to monitor vibrations

Applications

Page 7: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Volcano Monitoring◦ Sensors can be used to observe seismic events

◦ Nodes compare when they observe events to determine their location

◦ Allows geologists to collect data from large areas that are dangerous and difficult to access

Applications

Page 8: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

◦ Researchers from Harvard used a network of 16 sensors to monitor a volcano in Ecuador.

◦ In a period of 19 days, they recorded 229 earthquakes, eruptions, and other seismic events

Applications

Page 9: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Datacenter Provisioning◦ Datacenters use a lot of energy and the

computers produce a lot of heat

◦ Wireless Sensor Networks allow datacenters to determine which machines need cooling

◦ Making these dynamic adjustments can greatly reduce power consumption

Applications

Page 10: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Other applications include:◦ Environment and habitat monitoring◦ Wildfire detection◦ Battlefield surveillance

Wireless Sensor Networks have become a popular research area in computer science due to the wide range of applications

Applications

Page 11: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

One of the challenges in the field of sensor networks is the deployment of the sensors

The sensors must be properly spaced

If the sensors are too close to each other they won’t cover the entire area of interest

If the sensors are too far apart the coverage regions won’t overlap and the network could become partitioned

A Distributed Algorithm for Mobile Sensor Networks

Page 12: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Inspired by the equilibrium of molecules, which minimizes electronic energy

Each node tries to find its lowest energy point

When each node has found its lowest energy point, the nodes should be uniformly spread across the area of interest

Distributed Self-Spreading Algorithm

Page 13: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

A node’s movement is based on the combined forces acting on it due to its neighboring nodes.

Distributed Self-Spreading Algorithm

Page 14: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

The algorithm has 4 parts: 1: Initialization 2: Partial Force Calculation 3: Oscillation Check 4: Stability Check

Parts 2,3, and 4 repeat until the node stops moving

Distributed Self-Spreading Algorithm

Page 15: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Step 1: Initialization Initial locations and sensing

ranges are initialized

Local density is the number of nodes within its cR

Expected density is the number of nodes required to cover the entire area when the nodes are deployed uniformly

Page 16: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Step 2: Calculate Force The force depends on the

distance between the nodes and the node’s current local density

High local density will cause the force to be high

Page 17: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Oscillation is when a node is moving back and forth in the same spot

Counts the number of oscillations, if it is over the limit the node stops

Step 3: Calculate Oscillation

Page 18: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Step 4: Stability If a node moves less

than the threshold during the time duration, stability_limit,

the node is considered to be in its final position

Page 19: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Each node goes through the four steps until it stops moving. This results in the sensors being uniformly distributed.

The movement of each node is affected by the position of its neighboring nodes. If one node moves, the surrounding nodes will also move to maximize coverage and uniformity

Distributed Self-Spreading Algorithm

Page 20: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Mobile Data Gathering

Other Distributed Algorithms

Page 21: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Mobile Sensor Networks are a rapidly growing field with a wide range of applications

As distributed systems, like sensor networks, continue to become more popular, the demand for distributed algorithms will increase

Conclusion

Page 22: Ben Miller.  A distributed algorithm is a type of parallel algorithm  They are designed to run on multiple interconnected processors  Separate parts.

Golden Gate Bridge Study: S. Kim, S. Pakzad, D. Culler, J. Demmel, G. Fenves, S. Glaser, and M. Turon. Health monitoring of

civil infrastructures using wireless sensor networks. In IPSN ’07: Proceedings of the 6th international conference on Information processing in sensor networks, 2007

Volcano Study: G. Werner-Allen, K. Lorincz, J. Johnson, J. Leess, and M. Welsh. Fidelity and yield in a volcano monitoring sensor network. In 7th USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2006.

Lenzen, Christoph. "Distributed Algorithms for Sensor Networks." Hebrew University of Jerusalem . Zhao, Miao and Yang, Yuanyuan. “Optimization-Based Distributed Algorithms for Mobile Data

Gathering in Wireless Sensor Networks.”

Heo, and Varshney. “Energy-Efficient Deployment of Intelligent Mobile Sensor Networks” http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.307.4273&rep=rep1&type=pdf

Sources