Building Pervasive Computing Applications on Sensor Networks Rutgers

Building Pervasive Computing Applications on Sensor Networks Rutgers, The State University of New Jersey www. winlab. rutgers. edu 1

Introduction: Sensor Networks Wireless Sensor Nets Telecom Internet + Telecom Cell Phones Everywhere (~2000) The Virtual World Global Internet (~2000) virtualized via sensors & actuators The Physical World Information Tech Digital Media Convergence (2000 -2010) control data Global Internet for data & telecom Pervasive Computing (2015 -) IAB, May 13, 2004 2

Future Wireless: Pervasive Systems Compute & Storage Servers Pervasive Application Agents User interfaces for information & control Mobile Internet (IP-based) Overlay Pervasive Network Services 3 G/4 G BTS Sensor net/IP gateway GW Relay Node Ad-Hoc Sensor Net A Sensor/ Actuator Ad-Hoc Sensor Net B Virtualized Physical World Object or Event IAB, May 13, 2004 3

Future Wireless: Pervasive Applications n (Frictionless Capitalism)**2 Find goods and services on your PDA as you walk through town ¨ Walk into your dept store and pick up what you need (no cashier!) ¨ n “Smart” Transportation systems get routed around traffic jams in real-time ¨ receive collision avoidance feedback, augmented reality displays ¨ be guided to an open parking spot in a busy garage ¨ n Airport logistics and security Walk on to your plane (except for physical security check) ¨ Find your (lost) bags via RFID sensors ¨ Airport authorities can screen passenger flows and check for unusual patterns ¨ n Smart office or home Search for physical objects, documents, books ¨ Migrate your electronic media and documents between devices ¨ Maintain a “lifelog” that stores a history of events by location ¨ know where your co-workers and family members are ¨ IAB, May 13, 2004 4

Future Wireless: Key Technologies for Pervasive Systems n Sensors Tiny, low-power, integrated wireless sensors (hardware) ¨ Embedded OS and networking capabilities (software) ¨ n Ad-hoc wireless networks ¨ ¨ ¨ n Self-organizing sensor networks Scalable, capable of organic growth Interface to existing 3 G/4 G cellular and WLAN Power efficient operation Congestion control Pervasive computing software Dynamic binding of application agents and sensors ¨ Real-time orchestration of sensor net resources ¨ Robust, secure and failsafe systems ¨ Programming paradigm for sensor networks ¨ n emerging computer hardware category, optimized for size/power new type of wireless network without planning or central control fundamentally different software model - not TCP/IP Windows or Unix!! . . . beyond the scope of this talk Augmented reality, new displays, robotics, control, information processing. . . IAB, May 13, 2004 5

Enabling Technologies for Pervasive Systems 6

Sensor Technology: Hardware Integrated sensor/actuator + low -power microprocessor + radio Single chip or compact module Wireless networking Energy efficient design Applications of sensors include: Verticals: factory automation, security, military, logistics, transportation, . . Horizontal market: smart office, home pervasive computing Integrated wireless sensors are the “next microprocessor”. . . MIT DVS UC Berkeley MOTE Crossbow Sensor IAB, May 13, 2004 7

Sensor Technology: MUSE Prototype n “Multimodal” wireless sensor hardware being developed with NJCST funding. . . ¨ ¨ ¨ novel Zn. O materials for tunable sensors integration with low-power wireless transceiver designs focus on an integrated system-on-package or system-on-chip integrated ad-hoc networking software (as outlined earlier) sensor applications, including medical heart monitors, etc. Multimodal Zn. O device Sensor Device Sensor RF Modem, CPU, etc Zn. O SAW filter, MEMS, etc. RF Modem/CPU CMOS chip Embedded ad-hoc wireless net software Reduced functionality, optimized for low power consumption… 2002 -04 target: Multi-chip module for sub-802. 11 b Early medical applications at UMDNJ 2005 -06 target: Single chip prototype IAB, May 13, 2004 Pre-commercial applications w/ partners 8

Sensor Technology: Multimodal Zn. O device n “Tunable” Zn. O sensor developed by Prof. Y. Lu’s group Can be “reset” to increase sensitivity, e. g. in liquids or gas ¨ Dual mode (acoustic and UV optic) ¨ Applicable to variety of sensing needs ¨ Courtesy of: Prof Y. Lu, Rutgers U IAB, May 13, 2004 9

Sensor Networking: Congestion Alleviation n Resource control schemes to alleviate transient congestions in sensor networks ¨ ¨ ¨ Transient congestions are common Throttling traffic is not always an option (e. g. , an heart emergency generates a large volume of data within a short time frame) Sensor networks have elastic path capacity (e. g. , variable transmission power, directional antenna, etc) We can also use multiple routing paths to guarantee reliable event delivery Timely and accurate congestion level monitoring is the key IAB, May 13, 2004 10

Pervasive Computing: Software Model n Ubiquitous or pervasive computing scenarios require a fundamentally new software model (…not TCP/IP or web!!): Large number of context-dependent sources/sensors with unknown IP address ¨ Content-driven networking (…not like TCP/IP client-server!) ¨ Distributed, collaborative computing between “sensor clusters” ¨ Varying wireless connectivity and resource levels ¨ Pervasive Computing Application Agent 2 Agent 3 Agent 1 Pervasive/Ubiquitous Computing Software Model Overlay Network for Dynamic Agent <-> Sensor Association Sensor Cluster B Sensor Cluster A Resource Discovery Ad-hoc Routing Run-time Environment (network OS) OS/Process Scheduling IAB, May 13, 2004 11

Pervasive Computing: System Model Affinity Groups Autonomous Agents <> <> <> • • • <> <> Hierarchical Ad-Hoc Data Network • • • Content Network • • • Sensors & Actuators Courtesy of Prof. Max Ott IAB, May 13, 2004 12

Software Model: Pervasive Computing Stack hent Content Network on icati Ontology, Taxonomy ec uri t y : A Discovery/Messaging (Content-DHT/ Associative Rendezvous) Programming Model Aut Meteor Middleware Stack Opportunistic Interactions uthori zation Coordinate Flows S Ne. TS Applications (Autonomic Living, Ad hoc Control) Self Organizing Overlay st Tru Orbit Testbed Ad Hoc Routing Self Configuration Wireless/Wired Infrastructure Prof. Manish Parashar: Programming Model & Resource Discovery IAB, May 13, 2004 13

Pervasive Computing: Content Routing “Content routing” method for association between sensor devices, end-users and application programs Use of XML content multicast to dynamically find consumer/producer match ¨ XML multicast can be implemented as an overlay network on IP tunnels Network Infrastructure XML data Interest Profile ¨ Application Programs Routing XML n Sensor data XML descriptor Radio Forwarding Node End-user devices content multicast Storage Concept of Using Content Multicast for Data-Centric Software Model IAB, May 13, 2004 14

Pervasive Computing: Process Orchestration n Programming ad hoc control systems – Coordinated Flows Dynamic binding of application with sensors & actuators ¨ Orchestration of computing and network resources in real-time ¨ Allocate closest available space Look for parking space subscribe (plate-num, car-type, student) Parking Center Data Center Monitor incoming car Check parking space availability Check registration, Deduct parking fee Look for parking space: subscribe (plate-num, car-type, IAB guest) Incoming Car ( check ID: Registered student/faculty/staff, guest reservation? Fee deduction) Monitor available space Campus Parking Service courtesy of Prof. Manish Parashar IAB, May 13, 2004 15

Pervasive Computing Platform: Scheduling & Network OS MN MN Micro-level Cluster interface MN Micro-level • which sensor nodes should Cluster interface be sending back their readings? • how to split a task between a group of sensor nodes? MN be active while others sleep? Micro-level FN • which sensor nodes should Cluster interface Micro-level scheduling issues: FN provides certain data or functionality that is necessary to perform the specified task. • A sensor group can participate in multiple clusters • Work must be dynamically assigned to each group based on everyone’s energy budget, load, etc. • Each sensor group should schedule its work for different clusters according to other members in these clusters. MN Cluster I: • A cluster is formed because each sensor group MN FN FN MN MN IAB, May 13, 2004 16