Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor...

13
Boleslaw Szymanski, RPI, Troy, Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu Center for Pervasive Computing and Networking Rensselaer Polytechnic Institute

Transcript of Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor...

Page 1: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 1

Wireless Sensor Networks

Boleslaw [email protected]

Center for Pervasive Computing and Networking

Rensselaer Polytechnic Institute

Page 2: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 2

networking distributed computing real time systems simulations multimedia

adaptive scalable run-time execution environments

mobile software engineering

scalable distributed real-time systems

distributed system security

next generation network management

crisis & catastrophe management

mobile asset management

educationassistive technologies for the disabled

secure access & protection of assets

integration, scalability and composability:

fundamental contributing fields of computer science and engineering:

applications:

RPI’s Center onPervasive Computing and Networking:

Hierarchy of Challenges

Page 3: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 3

Gap between Trends Drives System Design

Changes relative cost-structures… Larger implications also for impact of IT on society

Page 4: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 4

Pervasive Wireless Communications

Ever-growing communication capacity needs for broadband wireless and mobile accessing in multimedia-rich environment, everywhere and anytime

There is a worldwide recognition that traditional methods of radio resource usage reach their limit and are no longer optimal

New communications frontier need to be explored for future wireless and mobile environments

Page 5: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 5

Sense-and-respond systemsSense-and-respond systemsBackground

“Sentient” networks: Computer networks composed of embedded “nodes” with onboard sensing, computational, and communication capability used for autonomous environmental monitoring

Military acousticnetworks

Air-defense radars

DARPA-lead projects[SensIT]

Past (80s-90s)Past (80s-90s)

Present (last 5 yrs.)Present (last 5 yrs.)Multi-modal devices

Ad hoc comm.

Small form factor

Industrial apps.

Future (next decade)Future (next decade)

Smart “dust”

Low cost

Disposable

Consumer apps.

Heterogeneousnetworks

Pervasive Ubiquitous

Actuators: Responsive services/ devices offering sensor or environmental control

The “Embedded Internet!”The “Embedded Internet!”

•System size•Amount of decentralization•Projected revenues!!!

Page 6: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 6

SensorTemp., light, humidity,

chemicals, acoustics, vibration

Computer4 MHz Atmel ATmega 128L

(equiv. to original ’82 IBM PC)CC RRSS

Radio2.4 GHz IEEE 802.15.4,<100m TX range

Basestation

Sense-and-respond systemsSense-and-respond systemsWireless sensor networks and applications

Features Offers macroscopic observation for real-time

environmental/contextual interaction Self-organizing, self-regulating, and self-repairing systems Multi-hop or direct-connect configurations to base station(s) Current state – extremely application-oriented!!!

Page 7: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 7

SensorTemp., light, humidity,

chemicals, acoustics, vibration

Computer4 MHz Atmel ATmega 128L

(equiv. to original ’82 IBM PC)

• Enemy intrusion detection• Habitat monitoring• Structural monitoring• Home automation and safety• Traffic control• Supply chain management (RFID)

Practical applicationsPractical applications

CC RRSS

Radio2.4 GHz IEEE 802.15.4,<100m TX range

Basestation

Sense-and-respond systemsSense-and-respond systemsWireless sensor networks and applications

Page 8: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 8

Sense-and-respond systemsSense-and-respond systemsSalient Challenges

Constrained resources Limited CPU, battery, and storage Premium communication costs

Ad hoc routing Dynamic topology

Transient wireless links and devices Collaborative information processing

Faulty sensors produce erroneous data Tradeoffs between performance and

resource utilization Sensor and actuator interaction

Synchronization between independent and heterogeneous services

Many more (querying, tasking, security, pollution, etc.)

Crossbow® MICAZ mote

Page 9: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 9

ESCORT: MotivationESCORT: Motivation

Wireless communication is a premium cost

Transient wireless links threaten application integrity Experiments show that at least 20%

of nodes exhibit at least 10% packet loss, and at least 10% of nodes exhibit more than 30% packet loss

Assuming an ARQ protocol is used, transmission cost increases as link quality worsens

Many proposed routing protocols A protocol-independent method for

enhancing energy-efficiency must be adopted

WSNs are envisioned to be highly redundant

Function/component

Operating current (mA)

Transmission (full power)

25

Reception 8Radio(sleep)

<1µA

Sensor board (full

power)

5

Sensor board (sleep)

5µA

CPU(full power)

8

CPU(sleep)

8µA

MICA2DOT series specs. [Crossbow]

Page 10: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 10

ESCORT: OverviewESCORT: Overview

Blue and orange nodes form communities which act as “virtual” nodes to the network layer

Orange nodes help coordinate community operation Green nodes are shared neighbors of the community Signal quality assessment, a combination of two

separate metrics, is used to form clusters of redundant nodes

Source Sink

Communal node

Shared neighbor Communication borderCoordinator node

Page 11: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 11

Lecture Hall AlgorithmLecture Hall Algorithmand the local leader election problemand the local leader election problem

Local leader election describes the problem of finding a node (leader) with the most desired property among a local group of nodesExample desired properties may include

distance-from-destination, energy, computational load, etc.

Can readily be applied to routing (selecting the next hop neighbor)

The traditional approach would require at least n messages and log(n) time

We require at most 3 messages in constant time using the “self-selection” algorithm

Page 12: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 12

SSR: OverviewSSR: Overview

Forwarding tables are not used Packets are forwarded based on a gradient metric – “hop

count” Packets are freely broadcast to all neighbors and “self-

selection” is used to determine the forwarding node

X-dimY-dim

Hop

coun

t

-10-5

0 5

10 -10

-5

0

5

10

-15

-10

-5

0

5

10

15

Hop

coun

t

Basestation

Simulated WSN Simplified SSR example

Basestation

Page 13: Boleslaw Szymanski, RPI, Troy, NY Center for Pervasive Computing and Networking 1 Wireless Sensor Networks Boleslaw Szymanski szymab@rpi.edu@rpi.edu Center.

Boleslaw Szymanski, RPI, Troy, NYCenter for Pervasive Computing and Networking 13

The SSR algorithm: ConclusionThe SSR algorithm: Conclusion

Remarks Low overhead for route maintenance and repair Good performance with simulated device failures and

transient (and asynchronous) links Self-selection algorithm is well suited for application

tuningFuture work Evaluate SSR using real wireless sensor network under

various operational conditions Explicitly extend SSR to exhibit energy-efficiency (via

radio control) Interact radio behavior with link behavior or self-

selection results Reduce amount of required synchronization