Talk Critique by Deepa Colluru

Talk Critique by Deepa Colluru

Date: Sep 03, 2003

Talk: Sensor Network on Defense

Presented By: Deepthi Agarwal

Summary

The topic presented sheds light upon the development of a new application using Smart Dust Sensor in a network. The main use of the application is to monitor and detect changes in its vicinity and transmit the data to a remote base station which then compiles the data into useful information based on predetermined algorithms.

The Smart Dust Sensor also called a Sensor Mote is a tiny wireless microelectromechanical device packed into a cubic millimeter speck. A sensor network differs from an Adhoc network in a number of ways like the number of nodes in a network, the range of a node, topology of the network, the algorithms used, etc. A Sensor network is data-centric and does not have a global identity. Also a Sensor network is application specific where a network can be tailored to the specific task at hand. Intermediate nodes perform data aggregation and caching unlike routers. Data is named by attributes that is queried by applications.

The sensor network could be used in various fields like Defense, Environment Monitoring and Wireless Surveillance. The application of Sensor Network in Defense was presented in detail. The goal is to use a sensor network to monitor enemy troop’s movement in a hostile terrain. The sensor network should give an approximate count of the number of human beings in the area. This is achieved by using Passive Infrared Sensors (PIR sensors) that detect infrared heat energy emitted by humans and the change in levels caused by movement.

The network is established by random deployment of sensor motes. Once deployed, the sensors use clustering and routing algorithms to form a communication network. A cluster is formed during bootstrap and a cluster head is chosen. A cluster head uses a data aggregation algorithm to compress data before transmitting. A routing algorithm is used by the cluster head to transmit data to the base station. The routing algorithm uses a hierarchical structure between the cluster heads and base station. When a cluster head dies, automatic reconfiguration of the cluster occurs. Security features are also implemented in the network to avoid manipulation and tampering of data.

Due to the extremely small size of the sensor, there are limitations as to how long a sensor can survive. The power used in a sensor is directly proportional to the amount of computation done by a sensor. Also the sensor has very limited memory that can be used for buffering. The challenge is to develop algorithms that use intelligent and fast computations that will directly result in longevity of the sensors in the network.

COMMENTS

·  The idea to use Sensor motes in Defense to monitor enemy movement is novel and promising.

·  The choice to use a sensor mote is justified by its small size that makes it easy to deploy in rough terrain and undetectable.

·  The constraints of the sensor mote and how they translate to its performance was shown.

·  A lot of emphasis has been given on the limited power in a sensor and how it is affected by the computations done by the sensor.

·  The typical life of a sensor is told to be 7 to 8 days. Alternate rechargeable power sources like solar energy could be explored for the sensors. This could increase the life of the sensor which directly translates to a longer life for the network.

·  The clustering and routing algorithms presented gave a clear idea as to how a sensor network can be deployed.

·  The feature of automatic reconfiguration of a cluster when a cluster head dies is good.

·  It is not clear as to what triggers the re-clustering algorithm, whether the dying cluster head sends a message to the cluster to reconfigure when it reaches a certain power threshold or when the sensors in a cluster do not get any response back from the cluster head.

·  More details need to be provided about the concept going to be used in the data aggregation algorithm. Will the relative positions of the sensors in the cluster be used in data compression?

·  A lot of emphasis needs to be given on the security of the network since its use in Defense directly translates to National Security. The algorithm also needs to be power efficient and fool proof.

·  The reliability and robustness of the sensor network needs to be established.

·  Also the percentage of accuracy in the data generated needs to be determined by using computer simulations and/or actual field tests.