Transcript An Integrated approach to developing sensor network solutions

An Integrated approach to
developing sensor network
solutions
Presented by
Richie John Thomas
08/27/04
Introduction
• Paper on the development work on sensor
networks at Computer and Network
Architecture Lab. Of the Swedish Institute
of Computer Science
• System core
– ESB Sensor Hardware running the Contiki OS
– Contiki emulation/simulation enviornment for
development
• Communication Stack
– Adaptive energy efficient MAC
– TCP/IP layer optimized for resource
constrained devices – allows system to be
connected to internet system
Hardware Platform
• ESB (Embedded sensor board)
– Texas Instruments MSP 430 low power micro
controller
– RF monolithics TR 1001 single chip RF transceiver
– Collection of sensors
•
•
•
•
•
•
Light- visible light
Passive infra red-movement
Temperature
Vibration-movement of sensor board
Microphone-ambient noise level
Infra red sender and receiver
– MSP 430 has 60 kb flash ROM and 2kb RAM
– 32 kb EEPROM provides addl. Persistent sec.
storage
– RF transceiver operates at 868 MHz and
supports rates upto 115.2 kbps
– Board has two external Connectors
• RS 232 port – for communication with PC
• JTAG interface – code downloading and
debugging
– MSP 430 for low power appln.
– Provides sleep modes awakened by interrupts
from internal timers or sensors
– Supports selective rewriting of internal flash
ROM
– TR 1001 RF transceiver
• Baseband transmission with either amplitude shift
keying or on-off keying
• Provides half duplex bit level access to physical
radio medium
– Higher level mechanisms (MAC protocol
processing, data encoding, time multiplexing)
should be done in s/w
– Transceiver connected to one of MSP 430
UART-Bit shifting in h/w rather than s/w
– UART causes interrupt only after full 8 bit
received as against MICA motes where
interrupt for each incoming bit
The embedded sensor board
The Contiki OS
• Flexible- allows individual programs and
services to be dynamically loaded and
unloaded in a running system.
• Based on event based concurrency model
• But also provides preemptive
multithreading
• Event based systems have lower resource
requirements and well suited for sensor
networks
• Allows cryptographic computations as it
can be run on a separate thread
• Allows dynamic reprogramming of n/w
behavior – due to service layer
• Conceptual layer providing service discovery and
run-time dynamic service replacement
• Portability makes it trivial to run Contiki as
a user level process under different PC
OS
• Appln. pgms developed in simulator can
be directly run and compiled on the sensor
h/w
MAC Layer
• Plays key role in energy efficiency and
quality of service
• MAC layer under development
– Energy efficient TDMA-like structure overlaid
on CSMA based collision avoidance protocol
– Asynchronous – Meet requirements on size,
complexity and cost and deployment in
extreme environment with variable h/w
stability
– Lightweight
– No traffic overhead- foregoing synchronization
– Scalable for multihop sensor n/w-no centralized
coordination used
– Provide good best effort QoS
– Energy efficiency
• Asynchronous power save protocol
• Based on the observation if node awake for just over half of
the time is awake interval will overlap with that of each of its
neighbors
• Nodes can determine available transmission window of
neighbors
• Node sleeps when no transmission
– Flow adaptation
• Phase adjustment used to increase effective
capacity of a region and reduce latency
• Node adjust its phase to avoid sending data when
there are high levels of contention or interference
• Sequence of nodes forming a path can adjust their
phase to minimize intra path interference
TCP/IP for Sensor Networks
• This requirement for network
management, calibration, diagnostics,
debugging
• Possible to connect network directly to
Internet
• Sensor data is transmitted using UDP/IP
but for administrative tasks reliable unicast
connections required
• TCP/ IP used
• Individual nodes can be addressed and
necessary reprogramming of sensors performed
• Also for debugging and diagnostic tasks
requiring reliable connectivity to a specific
sensor
• uIP has been developed with size of few kb and
few hundred bytes of RAM – not only on ESB
but variety of 8 and 16 bit processors
• Spatial IP addressing
– Each node uses its spatial location to
construct its IP address
– The spatial IP address only denotes the
location and not single identifiable node
– If node replaced new node given same IP
address as replaced node
– Nodes aware of their spatial location neither
require central server or communication
between nodes for address assignment
• Distributed TCP Caching
– Packet loss result in heavy overhead due to TCP end
to end ack. and retransmission scheme
– Poor performance in energy consumption and
throughput
– DTC cache TCP segments in network and perform
local retransmissions
– Nodes are allowed to cache only one segment
– Nodes attempt to identify and cache segments not
received by next hop
• The segment lost i.e. for which no ack. has been
received is locked in cache
• DTC has to respond to lost packets more quickly
to avoid end-to-end transmissions
• DTC uses ordinary TCP mechanisms to detect
packet loss
• Analytical and simulation results indicate that
DTC increases TCP performance
• DTC currently being implemented in ESB nodes
using Contiki simulator
Applications
• Building security
– Unwarranted motion in the secured building notified
via GSM and security personnel logs into the building
network to obtain status
– Two functions for sensor nodes- motion detectors and
backbone nodes
– Motion detectors in rooms and backbone nodes along
corridor
– Motion detectors has direct comm. path with at least
one backbone node and each backbone node had
contact with one other backbone node
• One backbone node equipped with external
interface device
• Alarm from motion detector to its backbone node
and from there to its back bone node
• Eventually all backbone nodes have info. abt.
entire state of network
• Security team with mobile backbone node to
scan the information
• Uses spatial IP addressing but mobile backbone
node has fixed IP address from another n/w to
differentiate it from other backbone nodes
• Marine monitoring
– Used to study water temp. and salinity
– Sensors attached to a buoy takes
measurements at known depths
– These connected as fixed network as
communication expensive
– Above waterline on the buoy is a full
function ESB
– These collect data from fixed n/w below
and transfer over wireless interface to
gateway node
– From here by GPRS to marine sciences
center
– This gateway can also be used to transport
data to sensors for reprogramming,
debugging and monitoring
– This exemplifies usefulness of being able to
manage nodes directly via TCP/IP protocols
• HVAC Monitoring
– Explore feasibility of instrumenting a
residential complex to improve the efficiency
of its HVAC
– Temperature and vibration sensors of ESB
are used
– IP based sensor accommodated into the
Ethernet of the energy control room