Nov 1 - 2 2005: Review MeetingACCLIMATE Instrumenting Wireless Sensor Networks for Real-Time Surveillance Songhwai Oh Advisor: Shankar Sastry EECS UC Berkeley.

Nov 1-2 2005: Review Meeting ACCLIMATE

Instrumenting Wireless Sensor Networks for

Real-Time Surveillance

Songhwai Oh Advisor: Shankar Sastry

EECSUC Berkeley


Building Comfort,Smart Alarms

Great Duck Island

Elder Care

Fire Response

Factories

Wind ResponseOf Golden Gate Bridge

Vineyards

Redwoods

Instrumenting the world

Soil monitoring


Challenges = Research Opportunities

applications

service

network

system

architecture

data mgmt

Monitoring & Managing Spaces and Things

technology

MEMSsensing Power

Comm. uRobotsactuate

Miniature, low-power connections to the physical world

Proc

Store


• Limited capabilities of a sensor node– Limited supply of power– Short communication range– High transmission failure rates– High communication delay rates– Limited amount of memory and computational power

• Inaccuracy of sensors– Short sensing range– Low detection probabilities– High false detection probabilities

• Inaccuracy of sensor network localization

Challenges = Research Opportunities

mica2dot

mag ultrasound

acoustic


Tracking in Sensor Networks

• Representative application of sensor networks– Event detection– Communication– Sensor fusion and estimation– Sensor management– Decision making, etc.

• Applications– Surveillance and security– Search and rescue– Disaster and emergency response system– Pursuit evasion games – Inventory management – Spatio-temporal data collection – Visitor guidance and other location-based services


Multiple-Target Tracking (MTT) in Sensor Networks

• Model uncertainty– Unknown number of targets– Unknown target initiation and termination times

• Measurement noise and inconsistency– Noise, False alarms, Packet losses, Delays

• Data association problem• Real-time

– Timely outputs required for control applications (e.g., pursuit evasion games)


Yet Another Complication: Binary Sensors

• Why binary sensors?– Sensor output is too noisy to

correlate signal strength with range

– Simple detection code

– 1-bit to communicate

• Provides coarse measurements• Difficult to use them directly to

initiate, maintain and terminate tracks

• We use spatial correlation to fuse binary measurements into finer position measurements

• Needs an efficient fusion algorithm


Problem


Previous Work: Multiple-Target Tracking (MTT) in Sensor Networks

• Traditional – computationally intensive– [Chong et al. 1990] Distributed multitarget multisensor tracking

• Classification-based – multiple single-target tracking problems– [Li et al. 2002] Detection, classification and tracking of targets– [Shin et al. 2003] A distributed algorithm for managing multi-

target identities in wireless ad-hoc sensor networks– [Liu et al. 2004] Distributed state representation for tracking

problems in sensor networks

• Ad-hoc – not robust– [Liu et al. 2003] Distributed group management for track

initiation and maintenance in target localization applications

• No general algorithm suited to sensor networks


Outline

• Multiple-target tracking (MTT) algorithm– Multi-sensor fusion algorithm– Markov chain Monte Carlo data association

(MCMCDA)

• Results from the final experiment of the Network Embedded Systems Technology (NEST) project


MTT

Overall Architecture

Fusion MCMCDA

Controller


“Simple” Multi-Sensor Fusion


Multi-Sensor Fusion: Likelihood

Detections Likelihood


Multi-Sensor Fusion: Threshold

Likelihood after threshold

• But it requires detections from all sensors to account false alarms

• Instead we compute the likelihood if there are at least nd detections

Likelihood


Multi-Sensor Fusion: Position Estimation

Black circle: position estimate


MTT


Fusion MCMCDA

Controller


MTT Problem: General Setup


Solution Space of Data Association Problem

(a) Observations Y

(b) Example of a partition of Y


Two Possible Solutions to Data Association Problem


Markov Chain Monte Carlo (MCMC)

• A general method to generate samples from a complex distribution

• For some complex problems, MCMC is the only known general algorithm that finds a good approximate solution in polynomial time [Jerrum, Sinclair, 1996]

• Applications:– Complex probability distribution integration problems– Counting problems (#P-complete problems)– Combinatorial optimization problems

• Data association problem has a very complex probability distribution


MCMC Data Association (MCMCDA)*

• Start with some initial state 1 2

*[Oh, Russell, Sastry 2004]


MCMC Data Association (MCMCDA)

• Propose a new state ’ » q(n,’)

• q: £ 2 ! [0,1], proposal distribution q(n,’) = probability of proposing ’ when the chain is in n

proposen

’

• q(n,’) is determined by 8 moves:



• If accepted,

• If not accepted,

n+1=’

n+1=n

• Accept the proposal with probability

() = P(|Y), Y = observations



• Repeat it for N steps








Optimality in the Limit

But how fast does it converge?


Polynomial-Time Approximation to Joint Probabilistic Data Association*

*[Oh, Sastry 2005]



MTT

Fusion MCMCDA

Controller

Multi-agent coordination algorithm

• Minimize time to capture all evaders

• Robust Minimum Time Control (MTC)


Simulation: Multiple-Target Tracking & Pursuit Evasion Games in Sensor Networks


NEST Final Experiment: MTT Demo

• Goal– Track an unknown number of multiple targets

using a sensor network of binary sensors without classification information

– Coordinate multiple pursuers to chase and capture multiple evaders in minimum time using a sensor network • Done in simulation due to physical and time constraints


NEST Final Experiment: Summer 2005


NEST Final Experiment: Sensor Node

Telos B mote•8MHz TI MSP430 microcontroller •RAM: 10kB; Flash: 48kB •Chipcon CC2420 Radio: 250kbps, 2.4GHz, IEEE 802.15.4 standard compliant

•Radio range of up to 125 meters

Trio Sensor Board•Features a microphone, a piezoelectric buzzer, x-y axis magnetometers, and four passive infrared (PIR) motion sensors

•Solar-power charging circuitry

Trio Node


NEST Final Experiment: System

• Software– TinyOS– Deluge

• Network reprogramming

– Drip and Drain (Routing Layer)

• Drip: disseminate commands• Drain: collect data

– DetectionEvent • Multi-moded event generator

– Multi-sensor fusion and multiple-target tracking algorithms


NEST Final Experiment: Demo


Current and Future Work

• Sensor networks– Robust distributed tracking algorithm – Robust tracking against malicious attacks– Performance analysis and metrics for sensor

networks

• Camera networks • Distributed multiple-target tracking and identity

management


Distributed multiple-target tracking and identity management*: an application of MCMCDA

*[Oh, Hwang, Roy, Sastry, 2005]


Summary

• Sensor networks– Individual sensor nodes are incapable and inaccurate – But the aggregation of spatially spread sensors can provide

accurate estimates using spatio-temporal correlation

• System-level approach– Multi-sensor fusion may provide incorrect and inconsistent

position reports – The inconsistency in position reports are fixed by the

MCMCDA tracking algorithm using temporal correlation – Adaptive control system

Nov 1 - 2 2005: Review MeetingACCLIMATE Instrumenting Wireless Sensor Networks for Real-Time Surveillance Songhwai Oh Advisor: Shankar Sastry EECS UC Berkeley.

Documents

wireless sensor networks

sensor networks liu

sensor networks ad

sensor networks traditional

sensor output

store slide

tracking of targets

distributed algorithm