Title: Wireless Ad Hoc Internets and Sensor Nets: Progress and Challenges
1Wireless Ad Hoc Internets and Sensor Nets
Progress and Challenges
- Anurag Kumar
- Professor, ECE Department
- Indian Institute of Science, Bangalore
- anurag_at_ece.iisc.ernet.in
2A Historical Perspective
- Ideas of the 1970s,
- packet networking
- packetised voice
- packet radio networks
- facilitated by advances in technology,
- digital transmission, digital signal processing,
microelectronics, high performance networking - are deployed in the 1990s
- ATM Networks and Internets
- VoIP (Internet Telephony)
- Wireless Ad Hoc Networks
3Structured Wireless Networks
- Point-to-point links
- Terrestrial or satellite
- Point-to-multipoint networks
- Satellite TDM/TDMA
- Cellular mobile networks
4Unstructured Wireless Networks
- Each node has a wireless transceiver
- Every node can forward packets
- Nodes associate in an Ad Hoc manner to form a
network - need to self organise to form a network
- multiple access wireless communication
- Certain periphery nodes may be linked to the
wired network
5Wireless LANs
- Usually extensions to an Ethernet LAN
- IEEE 802.11 MAC
- Several PHYs defined
- 2.4GHz ISM Band
- FHSS 1 Mbps and 2 Mbps
- DSSS 1, 2, 5.5, 11 Mbps
- OFDM up to 54 Mbps
- 5GHz Band
- OFDM 6 54 Mbps
- TCP/IP protocol stack
- Expected functionality is the same as from a LAN
access point
6Wireless Device Interconnection
- Bluetooth
- 2.4Ghz ISM Band
- Slow FH CDMA
- 79 hops, each 1Mhz band
- Bit rate 1Mbps
- Short range 10m
- Multiple access
- master slave polling
- self organising (ad hoc n/w)
- Voice and Data support
- IEEE 802.15 (WPAN)
piconet
piconet
piconet
Bluetooth scatternet
7Ad Hoc Sensor Networks
- Multifunction devices
- Sensing
- temperature, chemicals, light, body pulse rate
- Processing
- e.g., 8 bit, 4Mhz, 8KB flash, 512 B RAM
- Digital Radio
- Battery operated
- long idle life
The Berkeley Mote with a light temperature
sensor
8Sense, Compute and Communicate
- Sensor nodes pick up data
- e.g., temperature at (x,y), V(x,y)
- Process the data on the fly
- e.g., what is the maximum temperature?
- or, what are the locations of the maxima
- data compression and signal processing
- Communicate results to designated nodes
- Provides a rich class of problems in distributed
computing and communications
9Sensor Nets Forest Resource Management
- Sensors embedded in trees and attached to wild
animals - smart nails and smart collars
- Tree inventory management
- Felled timber tracking
- Forest fire detection and management
- Wildlife conservation
10Sensor Nets Disaster Management
- Emergency management networks for disaster areas
- Earthquakes, tornadoes, floods
- Search and rescue
- Sensors embedded in walls, environment, wearable
sensors (e.g., in watches) - Form an ad-hoc network after building collapse
- Biosensors for detecting the injured but living
- Locating such people for directing rescue teams
11Sensor Nets Ad Hoc Instrumentation
- Possible applications
- in hazardous situations monitoring radioactive
or chemical leaks - factories or large machines
- locating people in large offices, hospitals
(e.g., for forwarding calls) - Ease of deployment and replacement
- Minimal disruption of ongoing operations
12Common Technical Aspects
- Physical digital wireless communication
- a common spectrum is shared
- Multiple access
- how nodes coordinate their transmissions
- Self organisation, routing and forwarding,
scheduling - determining neighbours
- which packet to send, and to which node?
- Traffic flow in the network
- depends on application
13Capacity A High Level Model
arriving pkts
multihop pkts
complex server physical layer and multiple access
arriving pkts
departing pkts
node pkt queues
14Is the Network Connected?
- Consider a given area
- Randomly located n nodes
- Transmission range is r(n)
- power path loss receiver noise
- Small r(n) gt more spatial reuse
- but could lose connectedness
- r(n) should shrink no faster than
r(n)
15What is the Capacity?
- New packet arrival rate per node ?
- r(n) scaled to keep network connected
- Avg number of hops that a packet is relayed
- Rate of packets to be transmitted by the ad hoc
network n ? h(n) - Service capacity of the network S(n)
- spatial reuse
- physical layer and multiple access
- routing and forwarding approach
16Capacity Scaling with n
- Clearly it is necessary that
- This yields a bound on ? as a function of n
- If we consider only spatial reuse
- Hence ? scales slower than
17Capacity Remarks
- Thus the per node capacity of ad hoc networks
scales poorly with network size - scaling is much worse in practice
- multiple access TCP effects
- Multihop packet relaying is the key problem
- Need to develop approaches and applications that
minimise multihopping
18Performance of Multihop WLANS
- WLANS adopt the TCP/IP protocol suite
- TCP is well known to have poor performance over
lossy links - because of the inability to distinguish between
congestion related loss and wireless link
performance related loss - Multihop ad hoc situation causes additional
problems - TCP max window needs to be optimally set
- unfairness between connections over paths that
interfere
19Local Commn. Computation
- Sensor network needs to communicate
- Each node sends all values to the wired node
- no. of hops required
- Each node sends to neighbour that computes
cumulative maximum - no. of hops required n
20Routing in Ad Hoc Networks
- Multiple access and routing have been the most
active areas - Proactive routing
- periodic updates (as in the wired Internet)
- routes between all nodes maintained
- DSDV (1994)
- improvement to Bellman-Ford algorithm
- poor performance under high mobility
- since routes are often stale
- excessive routing overhead under low mobility
21(contd.) Routing
- On-demand routing protocols
- routes created only when desired by nodes
- such routes are cached, but timed out
- large overheads under high mobility
- under low mobility less overhead than DSDV
- DSR (1996)
- route entries contain entire sequence of nodes
- AODV (1999)
- route entries contain only next hop
22Routing in Sensor Networks
- On-demand routing preferred
- low mobility
- energy efficiency
- Among the protocols discussed AODV seems the most
appropriate - Data based routing
- instead of providing node address, provide a
value that we are seeking - all nodes with temperature reading above 30ºC?
23Locating Nodes
- What is the absolute position of a node?
- what is the location of the node with max (V) ?
- GPS may not be appropriate
- cost, size, energy needed, indoor applications
- Landmarks
- these could be fixed, with known locations or
could have GPS receivers - Relative location with respect to these landmarks
- inferring range with multipath propagation
- combining these range measurements to yield
absolute location estimates
24Major Initiatives
- Wireless LANs
- IEEE 802.11, Bluetooth, HomeRF
- Sensor Networks
- UCLA WINS 1993-1998
- DARPA funded projects 1999 onwards
- DSN Project (USC, UCLA, Virginia Tech)
- Energy efficient routing, sensor management
- SCADDS Project (USC)
- Distributed coordination and information
processing - Smart Dust Project (UC Berkeley)
- Hardware (including MEMS)and operating systems
- Distributed signal processing
25Final Remarks
- Sensor networks will be a rich source of
applications and problems - Challenges
- Limited communication and computing resources
- Energy conscious networking
- Decentralised coordination
- Distributed information processing
- Compact and efficient electronics, sensors,
operating systems and batteries - Standardisation of physical and MAC for sensors