1. Agent Technologies for Sensor Networks Reference: Alex Rogers and Nicholas R. Jennings, University of Southampton Daniel D. Corkill, University of Massachusetts Amherst IEEE Intelligent Systems, March-April 2009 Presented By: Md. Merazul Islam 0507036 Dept. of CSE, KUET
2. Introduction Wireless Sensor Network Way of wide-area monitoring Work with environmental, security, and military scenarios Consist of small, battery-powered devices Connected through a wireless communication network Faces some challenges 2 Md. Merazul Islam, CSE, KUET
3. Challenges Wireless Sensor Network Collect data over extended periods of time Deployed in inhospitable environments Replacing batteries is impossible Goals not achieve Sensors don’t share their sensing actions Network don’t adapt responses in a dynamically changing environment 3 Md. Merazul Islam, CSE, KUET
4. Overcomes Multiagent Systems Need Extensive set of formalisms, algorithms, and methodologies Mapping from sensor to agent Use of more low power resources Reliable hardware and communication Rather than we need A New Synthesis 4 Md. Merazul Islam, CSE, KUET
5. New Synthesis Synthesis Has Succeeded Efficient decentralized coordination algorithms Sensor-agent platforms in the field Intelligent agents These three examples are Proved & Evaluated by the researchers in real, hostile environment 5 Md. Merazul Islam, CSE, KUET
6. Agent-Based Decentralized Coordination Coordination Might Include Routing data through the network Choosing appropriate sampling rates of sensors Coordination Should Performed No central point of failure exits Computation must shared over the distributed resources Number of devices in the network increases 6 Md. Merazul Islam, CSE, KUET
7. Agent-Based Decentralized Coordination Proposed Algorithms Agent update their state for its own not globally Max-sum algorithm used to solve it Requires less computational and communication resources Generates good solutions applied to cyclic graphs Researchers have implemented it in hardware 7 Md. Merazul Islam, CSE, KUET
8. Figure 1. Hardware implementation of the max-sum algorithm and the graph-coloring benchmark problem using the Texas Instruments CC2430 System-on-Chip. The seven-segment display indicates the number of neighbors that each sensor has located, and the three LEDs indicate their respective sensor’s chosen color. 8 Md. Merazul Islam, CSE, KUET
9.
10. The CNAS has created a agent-based sensor network
12. Sharing of information is better to inform high-level operational decision making9 Md. Merazul Islam, CSE, KUET
13. Figure 2. A CNAS sensor agent at the 2006 Patriot Exercise at Fort McCoy, Wisconsin, deployed to collect real-time weather data at a landing strip. (photo courtesy of the US Air Force) 10 Md. Merazul Islam, CSE, KUET
14. Information Agents for Pervasive Sensor Networks Agents Must be Able to Handle missing or delayed data Detect faulty sensors Fuse noisy measurements from several sensors Efficiently manage bandwidth Predict both the value of missing sensor A live implementation of this prototype agent is currently available 11 Md. Merazul Islam, CSE, KUET
15. Figure 3. The Bramble Bank weather Station, located in the Solent. Figure 4. Screenshot of an information agent. A live implementation is available at www.aladdinproject.org/situation 12 Md. Merazul Islam, CSE, KUET
16. Conclusion The examples described here illustrate that even experimental sensor agent technology has become sufficiently reliable. Doing so will no doubt introduce novel challenges 13 Md. Merazul Islam, CSE, KUET