|Class description:||This class's topics will be wireless sensor networks, a multi-agent perspective. It contains information regarding the theory and practical implementation of wireless sensor networks.|
|Instructor:||Dr. Lotzi Bölöni|
|Office:||ENGR 3 - 319|
|Phone:||407-823-2320 (use only in last resort!)|
|E-mail:||email@example.com preferred communication method|
The assignments and the other announcements will be posted on the course web site
|Classroom:||ENG 2 - 103|
|Class Hours:||Monday, Wednesday 7:30 - 9:00.|
|Office Hours:||Monday, Wednesday 2:30 - 5:00|
|Pre-requisites:||Basic networking and programming knowledge.|
|FEEDS/Tegrity video stream:||TBA|
|Projects:||The class requires the students to prepare a major research project in one of the topics explored in the class.|
Project: 50 %
Standard 90/80/70/60 scale will be used for final grades (curved if necessary)
Why do I recommend the use of YAES as a basis for your project:
If you think that you rather write this from scratch in C or C++, feel free to use it.
In addition, I wrote YAES, so it is easier for me to help you with the project if you get stuck.
A version of YAES for the class (last updated Feb 4, 2008):
|Name:||Project type, topic|
Neural network-based transmission scheduling in a sensor network with mobile sink
Using the YAES simulator, you need to implement a strategy for transmission scheduling towards a mobile sync based on sensor networks. The strategy needs to optimize the energy consumption while reducing lost data.
|Marlon J. Fuentes and Bennie Lewis
Multi-attribute, energy optimal sensor fusion
Observations are reported which have multiple components: temperature, humidity, light, sound, presence. Observations are timestamped. The value of the observations decreases with their age, and increase with their spatial and temporal resolution.
Sensors can delay the sending of the observation to save energy by buffering. They can also fuse observations together in the temporal and spatial domain.
Objective: implement a sensor fusion and buffering algorithm which implements optimizes the value of transmitted observations for a fixed energy budget.
Implement it using the YAES simulator and perform comparative studies.
|Salman Saeed Khan and Omar Oreifej
Patch-based mobile sink movement
A sensor network is divided into patches. The patches contain sensors which collect their information, but do not forward them. Several mobile sinks are moving around the field, visiting patches and collecting information. We assume that we have a way to determine how much information was generated in a given patch.
Develop and compare several methods to intelligently visit the patches.
|Ramya Kavuri and Kudeja Khan
Mobile sinks on the roads
Assume that there is a sensor network which is traversed by a set of roads, which can be represented by a "road graph", with edges and undirected, straight edges. The mobile sink(s) travel with a constant speed on the edges. The sink can not turn back while on the edges, but when reaching a vertex, it can continue to move on any of the edges, including going back on the one where it came from. Assume that the nodes collect information while moving about the neighboring sensor nodes which are closer than transmission distance d. Assume that the sink knows about the information available to be picked up at every node.
Develop and compare several intelligent methods for the sink movement, which maximize the amount of information picked up by the sink(s). Note that this is simply the choice to decide which edge of the road graph to take at every node.
Complexity comparable to the patch based mobile sink project
|Project 1: Adriana Ogasawara and Joshua Mahaz
Project 2: Peter Matthews
Artificial immune system-based mobile node movement
A sensor network contains a set of mobile nodes. The idea is that the nodes are moving towards the areas where there is interesting phenomena (but, also, they are trying to load-balance themselves, that is, not gather all of them to the hottest part and ignore the rest). Design a mobility pattern based on an artificial immune system.
|David Benjamin and Phuoc Nguyen
Potential field-based mobile node movement
A sensor network contains a set of mobile nodes. The idea is that the nodes are moving towards the areas where there is interesting phenomena (but, also, they are trying to load-balance themselves, that is, not gather all of them to the hottest part and ignore the rest). Design a mobility pattern based on an artificial potential fields. The assumption is to assume that there are some artificial forces acting in the field. Areas of interest are attracting the nodes, while the nodes are attracting each other from longer distance, but are having a repulsion on small distance to prevent them clustering together.
|Jesse Goerz, Feras Batarseh
Survey: barrier coverage with wireless sensor networks
Some papers for start:
|Tim St. John
||Survey: target tracking in wireless sensor networks|
|Danish Riaz, Omar Amarin
Survey: the Guiness Book of Sensor Networks
Explore the extremes of proposed and implemented sensor networks
Survey: in-network data storage and sensor network databases
This project needs to cover at least these two projects:
|Ilhan Akbas and Volodymyr Prymma
Using the YAES simulator, simulate the Autonomous Sensor Network presented by Dr. Olariu in the following paper: link
The sensor nodes are very small devices with limited power and functions. However, they can harvest and store some energy from the environment, i.e. rechargeable sensors.
The network consists of a large number of sensor nodes and a set of aggregation nodes that organize and manage the sensor nodes in their vicinity. The aggregation nodes have special equipment for long range communications and may be stationary or mobile.
A patrol vehicle moves around in the area of the network and communicates with the aggregation nodes. Based on the information obtained from the aggregation nodes, the patrol vehicle may change its course in order to avoid a threat, or to investigate something of interest that was reported by the sensor nodes. Things to measure:
||Lecture Notes, Readings, Homeworks
-the science fiction aspect of sensor networks
-why the agent viewpoint?
||Wireless sensor networks
[slides] Wireless networks overview
[reading] A survey on sensor networks by I.F. Akyildiz, W. Su, Y. Sankarasubramanian and E. Cayirci
||Application example 1: Habitat monitoring
||[slides] Sensor network applications
Wireless Sensor Networks for Habitat Monitoring by A. Mainwaring, J. Polastre, R. Szewczyk, D. Culler, and J. Anderson
||Application example 2: Zebranet
Energy-Efficient Computing for Wildlife Tracking: Design Tradeoffs and Early Experiences with ZebraNet
by P. Juang, H. Oki, Y. Wang, M. Martonosi, L.-S. Peh, and D. Rubenstein
||Martin Luther King's day
Application example 3: Intruder tracking
|Jan. 28||Routing in sensor networks
[slides] Sensor network applications
[reading] ARRIVE: Algorithm for Robust Routing in Volatile Environments by Chris Karlof, Yaping Li and Joseph Polastre
||[reading] Rumor Routing Algorithm for Sensor Networks by
David Braginsky and Deborah Estrin
||Dissemination in sensor networks
||[reading] Directed Diffusion: A Scalable and Robust Communication
Paradigm for Sensor Networks by Chalermek Intanagonwiwat, Ramesh Govindan and Deborah Estrin
||The agent perspective.
Intelligence and rationality.
||[slides] Multiagent interactions. Games
||[slides] Reaching agreements. Auction models.
||Task oriented negotiation. Pareto efficiency.
The Contract Net
[slides] Agents working together.
[reading] The Contract Net Protocol: High-Level Communication and Control in a Distributed Problem Solver by Reid G. Smith
||Decision theory. Markov decision processes.
-partially observable Markov Decision Processes
-decentralized partially observable Markov Decision Processes
[slides] Partially observable MDPs.
[slides] Decentralized POMDPs.
||[slides] TinyDB and sensor network databases.
||Project presentations - 1
||Project presentations - 2
||Project presentations - 3
||Project presentations - 4
||Project presentations - 5