430 likes | 702 Views
Application Case-Studies in Mobile Sensor Networks. Parixit Aghera Pejman Kalkhoran. Agenda. What is a Sensor Network? Why Mobile Sensor Networks? Benefits of Mobile Sensor Networks Example Applications Mobile Sensor Nodes: ZebraNet, SWIM, Mobile Robot: PlantCare Observations.
E N D
Application Case-Studies in Mobile Sensor Networks Parixit Aghera Pejman Kalkhoran
Agenda • What is a Sensor Network? • Why Mobile Sensor Networks? • Benefits of Mobile Sensor Networks • Example Applications • Mobile Sensor Nodes: ZebraNet, SWIM, • Mobile Robot: PlantCare • Observations
What is a Sensor Network? • Micro-sensors, on-board processing, wireless interfaces feasible at very small scale--can monitor phenomena “up close” • Enables spatially and temporally dense environmental monitoring • Networked Sensing will reveal previously unobservable phenomena Definition from CENS presentation
Why Not Traditional Solutions? • Long range communication more power • Problem in Precise Deployment Applications require ad-hoc deployment • Significant Human Intervention • Not scalable Thousands of sensor nodes
Mobile Sensor Networks • Various elements in sensor network can be mobile • Base Station – Better network coverage/power saving • Mobile Robot/s – Deployment/Maintenance/Data Collection • Sensor nodes – Because of underlying phenomena • Motility • Virtual Mobility • Software agents moving across the sensor nodes
Benefits of Mobile Sensor Networks • Reduces number of sensors scalability • Self-sustaining • Power Management • Recalibration • Improved data collection due to minimization of sensor uncertainty • Increased Power-Bandwidth Efficiency
Energy Efficient Computing for Wildlife Tracking: Design Tradeoffs and Early Experiences with ZebraNet Philo Juang, Hidekazu Oki, Yong Wang, Margaret Martonosi, LiShiuan Peh, and Daniel Rubenstein Princeton University
ZebraNet – Problem Definition • Gathering data and observations on a wide range of species • Understand species interactions and influences on each other • System must run autonomously for months • Large spatial scale (Thousands of km2) • Mobile nodes and mobile base station
Zebra Mobility Model • Mobility models determine: • Speed, Direction, Frequency of movement, Forces of attraction
Collar Design • 12-Channel GPS Receiver with 32-bit microprocessor and 640 KB of user-available memory • Linx Short-Range Radio with 100m range and low power consumption • Slow, High-Powered Data Radio and Packet Modem with 8 km range
Collar Assumptions • 30 position samples per hour, all day • 6 hours per day of searching for peer nodes and transferring data between them using short-range radio • 3 hours per day of searching for mobile base station • 640 Kb transferred to base in 5 day period
Protocol Requirements • All data gathered must reach base station • Not all nodes have access to base station • Data must “hop” to the base station from node to node • Base station is not always active
Protocol Design - Flooding • Flooding Protocol • Flood data to all neighbors upon discovery • Large bandwidth, energy, storage requirements
Protocol Design – History-based • History-Based Protocol • Sends data based on prior communication patterns • Hierarchy determined by contact with base station
Storage Maintenance • Node prioritizes own data over others • Data points with most recent timestamp have priority • Delete lists – data points transferred to base • Transferred between nodes • If data points are within delete list, they are removed from memory
Possible Improvements • Calibration of measurement sensors • Node failure analysis • Storage parameters not specified • Lack of base station initiated communication • No strong reason for selected communication pattern
Impala: A Middleware System for Managing Autonomic, Parallel Sensor Systems By Ting Liu and Margaret Martonosi Princeton University
Impala • Middleware Architecture of Collar nodes • Major Contributions • Modular application framework • Event based application programming model • Non-VM based middleware for infrequent code updates • Application adaptivity mechanism • Implementation on iPAQ 206MHz, Linux
Impala – System Architecture • Applications are composed of multiple modules • Runs one application at a time • Applications share global data structure
Impala Application Adaptation • Maintains application parameters and system parameters • Switches between different application based on shared parameters and rules. • Takes into consideration device failure • Doesn’t switch to an application if a device used by application has failed
Impala Application Update • Uses module based version system • Three stage protocols • Send software version info • Send module request if module version is old • Send module to requesting node
The Shared Wireless Infostation Model - A New Ad Hoc Networking Paradigm(or Where there is a Whale, there is a Way) By Tara Small and Zygmunt J. Haas Cornell University
SWIM • Proposes SWIM as combination of Infostation and ad-hoc network model for increased capacity and delay • Infostation - Geographically intermittent coverage at high speed. • Designs a biological information acquisition system for whales. • Several SWIM stations floating on water surface • Each whale is attached with a sensor node (tag) • Communication takes place only when whale is on water surface • Data collected by tag should be offloaded to any SWIM stations
SWIM • Network Model • Information exchanged between tags with some probability • Information is offloaded to SWIM station • Each packet carried by a whale has time-to-live associated with it. • Packets are erased from memory when time-to-live expires • Mobility Model is based on • Feeding areas • Grouping behavior
SWIM • Analytical Model • Models inter tag packet transfer as discrete Markov chain. Uses analogy of infectious disease • Defines relationship between time-to-live (T) and CDF F(T) of a packet being offloaded at one of the SWIM station • Defines relationship between various parameters and storage requirements of each tag • Parameters • β = contact rate γ = whale-station contact rate, S(t) = susceptible whale, R(t) = recovered whale, I(t) = infected whale, N = number of whales
Making Sensor Networks Practical with Robots Anthony LaMarca, Waylon Brunette, David Koizumi, Matthew Lease, Stefan B. Sigurdsson, Kevin Sikorski, Dieter Fox, and Gaetano Borriello Intel Research Laboratory @ Seattle University of Washington
PlantCare • PlantCare project objective: to build a zero-configuration and distraction-free system for automatic care of houseplants • Contributions of this paper: • Analysis of use of mobile robot in sensor networks • Demonstration of in-situ calibration • Implementation using pioneer robot and Berkeley motes
Mobile Robot Use in sensor network • Context aware deployment of sensors • Robot can take sensor measurement at different places in environment and determine best place to deploy sensor. E.g. Temperature sensor • Continuous Calibration of Sensors • In-situ calibration • Recalibration at certain time interval • Recharging • A mobile robot goes to deployed sensor nodes and recharges them
Hardware Implementation • Sensor Node • Berkeley Motes • Sensor – Photo resistor, thermistor, irrometer, charge monitor • Power source – Capacitor that can be recharged by inductive coil
Hardware Implementation • Robot Hardware • Pioneer 2-DX mobile robot • Following custom hardware components controlled and monitored by a micro-controller • Small water tank with dispensing spout & pump • Inductive charging coil for sensor node • Inductive charge coil for robot • Human calibrated sensor node • Laptop with 802.11b connectivity • Maintenance Bay for charging and water refilling
Software Implementation • Rain software infrastructure provides a framework in which applications are implemented as co-operating services that communicates via asynchronous events • PlantCare application is composed of 15 services running on laptop • Sensor Software • TinyOS – An event-triggered operating system
Navigation, Deployment and Maintenances[6] Navigation Field computed by sensor node. Sensor Node Deployment Pictures and Animation from Maxim Batalin [6]
Multi-Robot Task Allocation [7] Pictures and Animation from Maxim Batalin [6]
Observations • Requires good understanding of problem domain • Problems are interdisciplinary in nature • For mobile sensors nodes, a good mobility model is required • Definition of distributed protocol for information exchange • Sensor Nodes and Mobile Robots can benefit from each other
Summary • Introduced sensor networks and their characteristics • Introduced mobile sensor networks and their characteristics • Presented several examples of mobile sensor network applications • Mobile Sensor Nodes: ZebraNet, SWIM, • Mobile Robot: PlantCare • Presented our observations from these applications
References • “Energy Efficient Computing for Wildlife Tracking: Design Tradeoffs and Early Experiences with ZebraNet”, Philo Juang, Hidekazu Oki, Yong Wang, Margaret Martonosi, LiShiuan Peh, and Daniel Rubenstein ASPLOS-X conference, October 2002 • “Impala: A Middleware System for Managing Autonomic, Parallel Sensor Systems”, Ting Liu and Margaret Martonosi PPoPP2003 • “The Shared Wireless Infostation Model - A New Ad Hoc Networking Paradigm (or Where there is a Whale, there is a Way)”, Tara Small and Zygmunt J. Haas, MobiHoc 2003 • “Making Sensor Networks Practical with Robots” Anthony LaMarca, Waylon Brunette, David Koizumi, Matthew Lease, Stefan B. Sigurdsson, Kevin Sikorski, Dieter Fox, and Gaetano Borriello, International Conference on Pervasive Computing 2002. • “Networked Infomechanical Systems(NIMS) for Ambient Intelligence” William J. Kaiser, Gregory J. Pottie, Mani Srivastava, Gaurav S. Sukhatme, John Villasenor, and Deborah Estrin • Maxim Batalin, Gaurav S. Sukhatme, and Myron Hattig, "Mobile Robot Navigation using a Sensor Network," To appear in IEEE International Conference on Robotics and Automation, Apr 2004. [PDF] • Maxim Batalin and Gaurav S. Sukhatme, "Using a Sensor Network for Distributed Multi-Robot Task Allocation," To appear in IEEE International Conference on Robotics and Automation, Apr 2004. [PDF]