A Process Oriented Development Flow for Wireless Sensor Networks Bernard Pottier*, Guillaume Kremer & Jimmy Osmont Université de Brest * LabSTICC, UMR CNRS 3192

WSN and applications W ireless S e nsor N etworks (WSN) are an expanding application field Green technology : environment, process control, power saving Human friendly : assists decisions in number of situations Driven by progresses in communications, integration technologies Application examples Parking in San Francisco: 6000 sensors connected to GPS/GSM service reduce car traffic and pollution, Forest fire detection from mesh connected sensors powered from imbalance in pH between tree and soil (MIT), Services for disabled persons/ bus transportation. Farming : cattle, green house, fields monitoring.

W ireless S e nsor N etworks (WSN) are an expanding application field

Green technology : environment, process control, power saving

Human friendly : assists decisions in number of situations

Driven by progresses in communications, integration technologies

Application examples

Parking in San Francisco: 6000 sensors connected to GPS/GSM service reduce car traffic and pollution,

Forest fire detection from mesh connected sensors powered from imbalance in pH between tree and soil (MIT),

Services for disabled persons/ bus transportation.

Farming : cattle, green house, fields monitoring.

Our objectives To support WSN design as a top-down flow: Sensor physical distribution and connectivity Communication layer management Application level algorithm design Code generation for sensors 2 stages flow Abstract design ,simulation Code synthesis Developped in Smalltalk-80

To support WSN design as a top-down flow:

Sensor physical distribution and connectivity

Communication layer management

Application level algorithm design

Code generation for sensors

2 stages flow

Abstract design ,simulation

Code synthesis

Developped in Smalltalk-80

WSN components Sensors Execution is a loop including Communications Local interaction with control and acquisition devices Low power, when possible local power source, low cost Practical example PsoC (Mixed analogue/logic reconfigurable system on chip) Spread spectrum wireless transceiver (10s to 100s meters) Temperature, pressure, presence, light, buzzer, .. Power

Sensors

Execution is a loop including

Communications

Local interaction with control and acquisition devices

Low power, when possible local power source, low cost

Practical example

PsoC (Mixed analogue/logic reconfigurable system on chip)

Spread spectrum wireless transceiver (10s to 100s meters)

Temperature, pressure, presence, light, buzzer, ..

Power

WSN organizations Static Star network, hub to internet (green house) Mesh connected with routing and gateways (SF parking, forest) Dynamic Mobile fleet (roaming cattle) Mobile in static or fleet (shepherd collecting information on sheep, bus exchanging information together / with bus stop

Static

Star network, hub to internet (green house)

Mesh connected with routing and gateways (SF parking, forest)

Dynamic

Mobile fleet (roaming cattle)

Mobile in static or fleet (shepherd collecting information on sheep, bus exchanging information together / with bus stop

An evident challenge is distributed algorithms Static deployment problems Network connectivity, fault tolerance, adaptive routing, control flow, memory/buffer use, Control, data collection, dynamic code distribution Communication scheduling ( idle to bandwidth saturation) Dynamic behaviour problems Routing algorithms, managing identities, Data storing and collection

Static deployment problems

Network connectivity, fault tolerance, adaptive routing, control flow, memory/buffer use,

Control, data collection, dynamic code distribution

Communication scheduling ( idle to bandwidth saturation)

Dynamic behaviour problems

Routing algorithms, managing identities,

Data storing and collection

A Framework for WSN Modelling and Synthesis Uses a two stage flow: Behaviour level (design): Modelling network topologies as communicating processes. Local interactions (sensing, controlling) Local contribution to distributed behaviour Tuning and optimizing design and algorithms from simulations Synthesis level (implementation) Dimensioning node characteristics from stage 1 parameters Target isolation Native code production, or virtual machine for stage 1 intermediate code.

Uses a two stage flow:

Behaviour level (design):

Modelling network topologies as communicating processes.

Local interactions (sensing, controlling)

Local contribution to distributed behaviour

Tuning and optimizing design and algorithms from simulations

Synthesis level (implementation)

Dimensioning node characteristics from stage 1 parameters

Target isolation

Native code production, or virtual machine for stage 1 intermediate code.

WSN abstract model NetworkGraph Collection of nodes NetworkNode NodeName (process name) Input Link collection Output Link collection Program/Procedure Name Link Source node Destination node Algorithms, builders, generators

NetworkGraph

Collection of nodes

NetworkNode

NodeName (process name)

Input Link collection

Output Link collection

Program/Procedure Name

Link

Source node

Destination node

Algorithms, builders, generators

Syntax based builder Smalltalk 'methods' described using VW ParserCompiler methodName messages alphabet … . NodeName1 { destNodes1 … } ProgramNode NodeName2 { destNodes2 … } ProgramNode Network is developped from a browser

Smalltalk 'methods' described using VW ParserCompiler

Network is developped from a browser

Random topologies builder Generate a sensor distribution on a surface Surface known as a rectangle Sensor wireless neighborhood known as a reachable distance Arbitrary number of sensors Compute the connectivity between nodes Build a corresponding network, drawing, text.

Generate a sensor distribution on a surface

Surface known as a rectangle

Sensor wireless neighborhood known as a reachable distance

Arbitrary number of sensors

Compute the connectivity between nodes

Build a corresponding network, drawing, text.

Physical deployment builder Flow Read a map, a photo Define scales for this background Define sensor transmit distance Place sensors and build a network. Computes network on the fly. Sandbox for the deployment of a sensor network along bus lines and bus stop

Flow

Read a map, a photo

Define scales for this background

Define sensor transmit distance

Place sensors and build a network.

Computes network on the fly.

Generators Graphviz (.dot file) Occam program: Concurrent organization : process skeleton, channel declaration, Process skeleton for a synchronous cycle

Graphviz (.dot file)

Occam program:

Concurrent organization : process skeleton, channel declaration,

Process skeleton for a synchronous cycle

Occam Generator Hundred of processes, thousands of channels Fixed behaviour for the synchronous model : Setup an initial state, prepare initial message output Loop forever Send messages on output links Receive messages on input links Change the node state according to current state and input messages Prepare next message output (c and d) place holders for programmer.

Hundred of processes, thousands of channels

Fixed behaviour for the synchronous model :

Setup an initial state, prepare initial message output

Loop forever

Send messages on output links

Receive messages on input links

Change the node state according to current state and input messages

Prepare next message output

(c and d) place holders for programmer.

Synchronous model : Why? Sensor networks are synchronous: Sharing a reference clock Sensors schedule listening and emitting interleaved with silences Time division, frequency or channel division Large number of known algorithms: Retrieving network characteristics – diameter – spanning trees Distributed reduction operations, broadcasts Routing with table exchanges, dynamic routing for mobiles Can be implemented on top of Occam: Concurrent send and receive operations avoid deadlocks Simulation of very large networks in native code for multi-cores

Sensor networks are synchronous:

Sharing a reference clock

Sensors schedule listening and emitting interleaved with silences

Time division, frequency or channel division

Large number of known algorithms:

Retrieving network characteristics – diameter – spanning trees

Distributed reduction operations, broadcasts

Routing with table exchanges, dynamic routing for mobiles

Can be implemented on top of Occam:

Concurrent send and receive operations avoid deadlocks

Simulation of very large networks in native code for multi-cores

Generator Status Very large process networks generate, compile and execute on multiprocessor Significant set of algorithms implemented: routing, naming, flow control.

Very large process networks generate, compile and execute on multiprocessor

Significant set of algorithms implemented: routing, naming, flow control.

Generator evolutions: VM Use of Occam intermediate bytecodes (TIS) to program sensors Virtual machine to execute TIS subset for synchronous model Channels operations are replaced by scheduled wireless exchanges (TDMA) Several bytecodes for parallel constructs, loops, operations implemented Stage 2 Stage 1

Use of Occam intermediate bytecodes (TIS) to program sensors

Virtual machine to execute TIS subset for synchronous model

Channels operations are replaced by scheduled wireless exchanges (TDMA)

Several bytecodes for parallel constructs, loops, operations implemented

Flow alternatives Use of Smalltalk to describe synchronous behaviour Methods export blocks for setup and transition Blocks operate on method temporary variables A process system is generated with communications operated using SharedQueues Translations similar to TIS Stage 1 (ST80 subset) Stage 2

Use of Smalltalk to describe synchronous behaviour

Methods export blocks for setup and transition

Blocks operate on method temporary variables

A process system is generated with communications operated using SharedQueues

Translations similar to TIS

Project Status Emerging project Appears as a contribution to PuceCom regional project in Brittany Cooperation developing with NUST in Pakistan Contributions of UBO students Hardware targets : Cypress PsoC and RF circuits, ARM7, GPS, … Contacts with industry : urban transportation, GIS, … VisualWorks NC, KroC (Occam), store 'students' @ as.univ-brest.fr Wish to contribute to sustainable development ( Oyster-catchers at round-island, bay of Brest )

Emerging project

Appears as a contribution to PuceCom regional project in Brittany

Cooperation developing with NUST in Pakistan

Contributions of UBO students

Hardware targets : Cypress PsoC and RF circuits, ARM7, GPS, …

Contacts with industry : urban transportation, GIS, …

VisualWorks NC, KroC (Occam), store 'students' @ as.univ-brest.fr

Wish to contribute to sustainable development