Quality of Monitoring and Optimization of Threat-based Mobile Coverage

Quality of Monitoring and Optimization of Threat-based Mobile Coverage

David K Y YauDepartment of Computer Science

Purdue University

David K Y YauDepartment of Computer Science

Purdue University

IntroductionIntroduction

• Part of federal SensorNet initiative Oak Ridge National Lab and university partners (including Purdue and UIUC)

• Initial deployment of a detection, identification, and tracking sensor-cyber network (DITSCN) in the Washington D.C. and Memphis Port areas; against radiological, biological, and chemical threats.

• DITSCN combining various modalities of sensors and cyber networks

– Sensor network provides information about the physical space– Cyber network provides storage and computational resources to predict plume propagation based on realistic dispersion models– Decisions regarding future sensing and communications are made in cyber network and carried out in the physical space

DITSCN ArchitectureDITSCN Architecture

Multi-hop communication

Control Center

Physical Space

Sensors…

SensorNet Node

Actuator

Cyber Space

1. Convergence between physical and cyber spaces Effectively gather information about the physical space Communicate most useful data to the cyber space given bandwidth,

delay and signal attenuation constraints Enable the cyber space to task and activate sensors to collect high-

quality data

2. Acknowledgment of the existence of uncertainty; enable decision making processes to deal with the uncertainty in a robust fashion Incorporate knowledge of physical environment: people, terrain, land

cover, and meteorological information Model physical phenomena adequately (e.g., plumes with respect to

the absorption, propagation, and dispersion coefficients)

3. Support for deeply embedded operations Integrate system components in an open, plug-and-play manner,

through the use of open data, control, and communication interfaces

Research TasksResearch Tasks

Cyber-space Analytical Results

RFTrax RAD Sensor to detect the presence and intensity of the plume source

WMS Wind Sensor to monitor background wind speed and direction

Physical Space Sensing

Sensor data communicated through RS-485 or 802.11x interfaces to the SensorNet Node Multihop wireless mesh network for robustness and flexibility

current implementation uses Linksys routers running AODV

IEEE 1451 interface to configure sensors at runtime

Wide-area Wireless Network Communication

Router

Mo

de

m

Serial Interface

Power Supplies

Processor

Antenna

Fan

SensorNet Node Hardware

TEDS

STIM

1451.2 Stubs

(Web) Server

Control

Auth

Data Services

Legacy Codes

Data Management

andStorage

Configuration

Sensors Interface

Comm. Mode Control

E.g., Sprint

Session

Other

Services for RDCand external users

TEDSTEDS

STIM

Legacy Codes

US

B

Mux

Serial

Ethernet

Link options: Dialup/PCS/ 802.11 Wired etc.

1451

.1 N

CA

P

Sen

sors

SensorNet Node Software Architecture

• ER-1 robots supporting autonomous and programmable movement are guided by the cyber center, through commands sent over 802.11x wireless network

• Tasking enables sensor mobility to increase the coverage of high-threat locations

ER-1 Robots

Physical Space Tasking

Detection of Radiation ThreatsDetection of Radiation Threats

Stealthy bombs Small explosions (can be dismissed as low harm), but Exposure of population to dangerous radiation Need detection by suitable sensors

Commercial sensors RFTrax RAD-CZT (limited range of tens of feet) Yankee Environmental System Inc. RAD 7001

(somewhat longer range but more expensive) High procurement and operation costs (may not have

sufficient sensors for full area coverage all the time)

Stealthy bombs Small explosions (can be dismissed as low harm), but Exposure of population to dangerous radiation Need detection by suitable sensors

Commercial sensors RFTrax RAD-CZT (limited range of tens of feet) Yankee Environmental System Inc. RAD 7001

(somewhat longer range but more expensive) High procurement and operation costs (may not have

sufficient sensors for full area coverage all the time)

Prior SensorNet DeploymentsPrior SensorNet Deployments

Washington DC deployment Gamma radiation detection by RFTrax in urban areas

Memphis Port deployment Chemical detection of fresh water supply to area

residents by Smith APD 2000

Lessons learned Management of resource constraints (mobile coverage) Importance of people protection (resource allocation) Uncertainty management (temporal dimension)

Washington DC deployment Gamma radiation detection by RFTrax in urban areas

Memphis Port deployment Chemical detection of fresh water supply to area

residents by Smith APD 2000

Lessons learned Management of resource constraints (mobile coverage) Importance of people protection (resource allocation) Uncertainty management (temporal dimension)

Temporal Dimension of Sensing (radiation detection)

Temporal Dimension of Sensing (radiation detection)

Other Utility FunctionsOther Utility Functions

People-centric Resource AllocationPeople-centric Resource Allocation

• Allocating goal of coverage time by mobile sensor• higher threats (people impact) higher coverage• proportional to numbers of residents in subregions

• Proportional sharing is well known (CPU, network, …)• but impact on sensor QoM not well understood

Problem FormulationProblem Formulation

One sensor moving among n points of interest (PoI) under some maximum speed vmax

Each PoI has given threat level (no. of residents) Road of distance dij connects PoIs i and j

Dynamic events appear at each PoI Stochastic event arrival, staying, and absent times (given

probabilistic distributions) Sensing quality increases with sensing time (according to

utility function) Sensing occurs when event falls within sensing range

R of sensor

One sensor moving among n points of interest (PoI) under some maximum speed vmax

Each PoI has given threat level (no. of residents) Road of distance dij connects PoIs i and j

Dynamic events appear at each PoI Stochastic event arrival, staying, and absent times (given

probabilistic distributions) Sensing quality increases with sensing time (according to

utility function) Sensing occurs when event falls within sensing range

R of sensor

Goals and QuestionsGoals and Questions

We seek to achieve proportional sharing of sensor coverage time among PoIs according to threat profile What does it mean in terms of QoM? Does r times coverage implies r times performance?

Questions: how should the sensor move among the PoIs to maximize the aggregate information captured? Subject to physical constraints of movement and proportional

sharing goal What’s impact of sharing granularity? What’s scaling law of mobile coverage? (Do we capture more

information by being mobile?)

We seek to achieve proportional sharing of sensor coverage time among PoIs according to threat profile What does it mean in terms of QoM? Does r times coverage implies r times performance?

Questions: how should the sensor move among the PoIs to maximize the aggregate information captured? Subject to physical constraints of movement and proportional

sharing goal What’s impact of sharing granularity? What’s scaling law of mobile coverage? (Do we capture more

information by being mobile?)

Periodic PoI SchedulePeriodic PoI Schedule

Analyze periodic presence/absence of sensor at given PoI Induced by mobile coverage algorithm (feasibility and

realization later) Sensor is present for q time units every p time units (min

present time is =2R/vmax) Same q/p proportional share can be achieved at different

fairness granularity P A A A vs. P P A A A A A A (25% share)

How much information captured as a function of event dynamics and type of event?

Analyze periodic presence/absence of sensor at given PoI Induced by mobile coverage algorithm (feasibility and

realization later) Sensor is present for q time units every p time units (min

present time is =2R/vmax) Same q/p proportional share can be achieved at different

fairness granularity P A A A vs. P P A A A A A A (25% share)

How much information captured as a function of event dynamics and type of event?

Periodic PoI Coverage: Blip EventsPeriodic PoI Coverage: Blip Events

Theorem: For independent arrivals of events that have the step utility function and do not stay, i.e. “blip events”, the QoM at any PoI is directly proportional to its share of coverage time

Corollary: For these events, the achieved QoM at a PoI is linear in the proportional share and does not depend on the fairness granularity p r times coverage r times QoM

Theorem: For independent arrivals of events that have the step utility function and do not stay, i.e. “blip events”, the QoM at any PoI is directly proportional to its share of coverage time

Corollary: For these events, the achieved QoM at a PoI is linear in the proportional share and does not depend on the fairness granularity p r times coverage r times QoM

Periodic PoI Coverage: Step UtilityPeriodic PoI Coverage: Step Utility

Theorem : For independent arrivals of events that stay and have the step utility function, the QoM at a PoI is given by

Theorem : For independent arrivals of events that stay and have the step utility function, the QoM at a PoI is given by

Corolloraries (Step Utility)Corolloraries (Step Utility)

Corollary: With the fairness granularity p kept constant, we have:

QoM is a monotonically decreasing function of the fairness granularity, i.e., Q decreases as p increases. Furthermore,

Corollary: With the fairness granularity p kept constant, we have:

QoM is a monotonically decreasing function of the fairness granularity, i.e., Q decreases as p increases. Furthermore,

QoM Justification of MobilityQoM Justification of Mobility

Theorem: For sensor moving among k PoIs, the expected fraction of events captured is an increasing function of k.

Theorem: For sensor moving among k PoIs, the expected fraction of events captured is an increasing function of k.

Periodic PoI Coverage: General UtilityPeriodic PoI Coverage: General Utility

Theorem: For events at a PoI that have the utility function U( ・ ) and whose event staying time pdf is given by f(x), the achieved QoM equals

Theorem: For events at a PoI that have the utility function U( ・ ) and whose event staying time pdf is given by f(x), the achieved QoM equals

Exponential Staying TimeExponential Staying Time

Exponentialutility function

Pareto Staying Time

Pareto Staying Time

Implications of Theorem (General Utility)Implications of Theorem (General Utility)

Step and Exponential Utilities: QoM decreases monotonically in p Concave function advantageous to move around and

look for new information

But for other utility functions (e.g., Delayed Step), optimal QoM may occur at intermediate p Competitive effects between observing existing event

long enough for significant information vs. looking for new information elsewhere

Step and Exponential Utilities: QoM decreases monotonically in p Concave function advantageous to move around and

look for new information

But for other utility functions (e.g., Delayed Step), optimal QoM may occur at intermediate p Competitive effects between observing existing event

long enough for significant information vs. looking for new information elsewhere

Periodic Global Sensor SchedulePeriodic Global Sensor Schedule

Smallest periodic sequence of PoIs visited and the visit times S={(L1,C1) … (Lm,Cm)} (PoI L1 visited for C1 time, etc)

Not all periodic global schedules produce simple periodic PoI schedules E.g., {(1,T) (2,3T) (1,T) (3,2T)}

When each PoI appears in S no more than once, S is called linear periodic schedule

Maximum feasible utilization of S:

Smallest periodic sequence of PoIs visited and the visit times S={(L1,C1) … (Lm,Cm)} (PoI L1 visited for C1 time, etc)

Not all periodic global schedules produce simple periodic PoI schedules E.g., {(1,T) (2,3T) (1,T) (3,2T)}

When each PoI appears in S no more than once, S is called linear periodic schedule

Maximum feasible utilization of S:

Maximum Feasible UtilizationMaximum Feasible Utilization

Theorem: The maximum feasible utilization of S is

Theorem: The maximum feasible utilization of S is

where

Optimization of Linear Periodic SchedulesOptimization of Linear Periodic Schedules

Find linear visit schedule that minimizes aj TSP, but good approximation algorithms exist Once visit schedule known, all aj’s are determined,

remains to determine Cj’s

Express each Cj as function of C1 (reduce problem to single dimension)

Choose C1 that optimizes Q* (one dimensional optimization depending on event utility function)

Find linear visit schedule that minimizes aj TSP, but good approximation algorithms exist Once visit schedule known, all aj’s are determined,

remains to determine Cj’s

Express each Cj as function of C1 (reduce problem to single dimension)

Choose C1 that optimizes Q* (one dimensional optimization depending on event utility function)

Illustration (Blip Events)Illustration (Blip Events)

If aj = 0, then any choice of C1 is optimal

Otherwise, there is no optimal choice but we can get arbitrarily close to the optimal by selecting a sufficiently large C1 (hence, a sufficiently small travel overhead)

If aj = 0, then any choice of C1 is optimal

Otherwise, there is no optimal choice but we can get arbitrarily close to the optimal by selecting a sufficiently large C1 (hence, a sufficiently small travel overhead)

Linear Periodic Schedules are Sub-optimalLinear Periodic Schedules are Sub-optimal

Consider three PoIs and Step utility events d12 = d13 = d23 = 2R Proportional sharing objective r12 = n/(n - 1) and r13 = n

Optimal linear periodic schedule is

However, QoM increases with finer grained sharing; hence, optimal non-linear periodic schedule is

Consider three PoIs and Step utility events d12 = d13 = d23 = 2R Proportional sharing objective r12 = n/(n - 1) and r13 = n

Optimal linear periodic schedule is

However, QoM increases with finer grained sharing; hence, optimal non-linear periodic schedule is

Optimization of General Global CoverageOptimization of General Global Coverage

Start with some schedule of length n Could be optimal linear schedule if it exists

Search for optimal general schedule of the same length (while respecting physical constraints) Search space is huge: n! permutations Use simulated annealing to guide the search

and obtain global optimal with high probability

Start with some schedule of length n Could be optimal linear schedule if it exists

Search for optimal general schedule of the same length (while respecting physical constraints) Search space is huge: n! permutations Use simulated annealing to guide the search

and obtain global optimal with high probability

Optimization Algorithm

Simulation ResultsSimulation Results

QoM of Blip EventsQoM of Blip Events

QoM of Step Utility EventsQoM of Step Utility Events

QoM of Exponential Utility EventsQoM of Exponential Utility Events

QoM of Delayed Step Utility EventsQoM of Delayed Step Utility Events

Optimization of General Periodic SchedulesOptimization of General Periodic Schedules

Another Proportional ShareAnother Proportional ShareProportional share ratios53:29:17

ConclusionsConclusions

Extensive analysis and supporting simulations to understand QoM of proportional-share mobile sensor coverage

Higher share higher QoM (but not linear except for blip events)

When events stay, QoM can be much higher than proportional share due to ``extra’’ events captured Sensor gains by moving around to look for new information

Optimal coverage depends on event utility Step, Exponential utilities: finer granularity is better Linear utility: initially flat, then finer granularity is better Delayed Step and S-Shaped utilities: intermediate fairness granularity

is best

Extensive analysis and supporting simulations to understand QoM of proportional-share mobile sensor coverage

Higher share higher QoM (but not linear except for blip events)

When events stay, QoM can be much higher than proportional share due to ``extra’’ events captured Sensor gains by moving around to look for new information

Optimal coverage depends on event utility Step, Exponential utilities: finer granularity is better Linear utility: initially flat, then finer granularity is better Delayed Step and S-Shaped utilities: intermediate fairness granularity

is best

Conclusions (continued)Conclusions (continued)

Linear periodic schedules can be optimized as one dimensional optimization problem But optimal linear periodic schedules are generally sub-

optimal

General periodic schedules of given lengths can be optimized using simulated annealing Near-global optimal schedule with high probability Practical search time even for huge search spaces

Search terminates in seconds in our experiments

Linear periodic schedules can be optimized as one dimensional optimization problem But optimal linear periodic schedules are generally sub-

optimal

General periodic schedules of given lengths can be optimized using simulated annealing Near-global optimal schedule with high probability Practical search time even for huge search spaces

Search terminates in seconds in our experiments

DiscussionsDiscussions

Advantages of mobile coverage have been established in prior work Bisnik, Abouzeid, Isler, ACM MOBICOM 2006 Liu, Brass, Dousse, Nain, Towsley, ACM Mobihoc 2005 Increased mobility is always better (ignoring costs)

Our new angles/results Proportional sharing of coverage time, motivated by people

protection Temporal dimension of sensing, captured in event utility functions Mobility is useful, but not always the more the better when

temporal dimension is present (in terms of QoM) Linear periodic schedules can be significantly suboptimal; solved

optimization of general periodic schedules

Advantages of mobile coverage have been established in prior work Bisnik, Abouzeid, Isler, ACM MOBICOM 2006 Liu, Brass, Dousse, Nain, Towsley, ACM Mobihoc 2005 Increased mobility is always better (ignoring costs)

Our new angles/results Proportional sharing of coverage time, motivated by people

protection Temporal dimension of sensing, captured in event utility functions Mobility is useful, but not always the more the better when

temporal dimension is present (in terms of QoM) Linear periodic schedules can be significantly suboptimal; solved

optimization of general periodic schedules