Remote Deployment of Remote Deployment of Sensor NetworksSensor NetworksCase Study: Monitoring the Case Study: Monitoring the mountains of Afghanistanmountains of Afghanistan

ContextContext

Military monitoring of Oppositional Forces in remote, inaccessible areas like the mountains of Afghanistan and Pakistan using wireless sensor networks.

The deployment and uplink methods in this application are largely unexplored

Kyber Pass – One of the more accessible roadways in this part of the world, rarely used

by guerilla fighters

Main ProblemsMain Problems

Wireless sensor networks still require (in most applications):

• Manual (even if random) placement of nodes• Infrastructure at head end

These obstacles must be overcome to deploy a monitoring network readily in the target environment

Issues outlined todayIssues outlined today

• Distribution method (hands-off)

• Data aggregation, sensor configuration

• Uplink strategy (no infrastructure)

• Network management

• Energy optimization

• Security

• Final Goals

Distribution: MethodsDistribution: Methods

Air-scatter, Air-dropLikely conducted at night

to avoid detectionMultiple Launch Rocket System

GOOD: Takes ground personnel out of dangerous or hostile-controlled terrain

BAD: Result in very poorly-behaved, random distributions: wind currents or terrain features may cause many sensors to channel into a ravine, or river

Ballistics and delivery will not be a main focus of this research

Distribution: PackageDistribution: Package

Smart Dust or bigger motes could be dropped at altitude

as a bundle with parachute… then scattered at low altitude

to control distribution areaSmart dust is great for military applications because it is small (appx 1mm), difficult to detect, can be dropped at high altitude Motes could be disguised as rocks…

Data, Sensors: The TargetData, Sensors: The Target

Seismic, audio, heat, visual, chemical, electromagnetic signatures

Trucks, horses, pedestrians

Data, Sensors: Sensor DevicesData, Sensors: Sensor Devices

• Heterogeneous approach: multiple types of sensors, nodes

• Since life of network is dependent on functionality of many types of nodes, lifetime and failure calculations become more complex

Data, Sensors: Sensor DevicesData, Sensors: Sensor Devices

• Use passive or low power sensors as first-responders (piezoelectrics like audio, seismic)

• The first-responder sensors will wake up neighboring nodes with other sensors (thermal, visual, electromagnetic, chemical, radiological, etc)

• Power-hungry sensors should sleep as much as possible

Uplink: Heterogeneous NodesUplink: Heterogeneous Nodes

• Connect to military satellite constellation

• Uplink nodes will be different from sensor nodes: larger, more battery, more power

• Uplink nodes form some small % of all nodes

Uplink: InitializationUplink: Initialization

How to initialize the network when the head is seen as being in multiple places? How to avoid partitioning? This will be researched and protocol developed.

?

Network: Expendable UplinksNetwork: Expendable Uplinks

Active

Network: Expendable UplinksNetwork: Expendable Uplinks

ActivePOWER LOW

TAKE OVER

Network: Expendable UplinksNetwork: Expendable Uplinks

Active

One can see this rerouting may be computation intensive, and is in fact similar to MANET routing when nodes fail. The main difference here: the network can still be operational if it is partitioned.

Network: ProtocolsNetwork: Protocols

• A significant undertaking of the project is to develop and simulate the protocol which reassigns the network head in an energy efficient manner for various configurations

• Important to optimize whole network power consumption

Network: ProtocolsNetwork: Protocols

• Routing must transition to avoid periods of downtime or routing confusion

• Current head node must select its successor to eliminate routing confusion

Energy: LifetimeEnergy: Lifetime

• Network needs to be operational for at least a few months at a time: otherwise satellite monitoring becomes preferable in some circumstances

• The heterogeneous nodes will have to adapt to over-stimulus to keep network alive. i.e. if animals constantly pass by the network and trip seismic sensors, they should be able to be ignored in favor of reliable detectors like electromagnetic

SecuritySecurity

• Primary risk is detection of presence not data decoding

• Spread spectrum, noise-like signals should be used

• “Chatter” must be kept to minimum

• Alert uplinks are okay

SecuritySecurity

The network may be compromised by several means:

• Stimulate for false positives, perhaps rapidly depleting the network of energy

• Compromise the uplink signal/device

• Jamming

Security: CountermeasuresSecurity: Countermeasures

• Detect, isolate, ignore false-positive flood input

• Encode uplink transmission

• Spread spectrum jam resistant communications links

GoalsGoals

• Develop basic application proposal for DARPA SBIR

• Identify one narrow topic within the system which needs research and development, write a paper on that topic

• Current candidate for narrow research:protocol to identify, shift network head-ends

ConclusionsConclusions

• Next lecture will involve the analysis of key challenges, quantitative approaches, and outline of the draft proposal for DARPA