1. Energy Conservation in Sensor Networks
Group 11: Kevin Malby, Phu Nguyen
Reference: Giuseppe Anastasi, Marco Conti, Mario Di Francesco, Andrea Passarella, “Energy Conservation in Wireless Sensor Networks: a Survey”

2. Motivation For Energy Conservation
Sensor nodes are typically small and must run on battery power or self-generated power
They collect sensor information, process and store this information, and relay it to a central server for further processing. All of these processes and components require energy
They must also generally remain on for extended periods of time, often without human maintenance
Without energy conservation techniques, sensor nodes may not be able to perform their tasks effectively
Kevin

3. Architecture of Sensor networking Systems
Andy

4. Architecture of a typical sensor node
Andy
1. Communication Subsystem: consume the most power, compared to other subsystems of a sensor node.
2. Research shows that the energy needed to transmit 1 bit of data equal the energy need to
handle 1000 computation. So computation is cheap while transmission is expensive.

5. General Approaches to Energy Conservation
Three Primary Techniques
Duty-Cycling
Focused on networking subsystem
Switch the radio transceiver between on/off states according to whether or not data needs to be transmitted
Data-Driven Conservation
Due to temporal/spatial correlation between data samples, often information being transmitted is redundant and need not be sent
If sensor itself is power hungry, intelligently adjust sampling rate while retaining accuracy
Mobility Approach
Reduces the hop distance for transmission and balances the load within the network, ensuring specific nodes are not overworked
Kevin

6. Energy conservation schemas
Kevin

7. Duty Cycling Energy Conservation Schemes
Topology Control
Location driven
Connection driven
Power management
Sleep/awake protocols
MAC protocol with low duty-cycle
Andy
Topology control: increase network lifetime by a factor of 2-3 time

8. Topology Control: Location driven
Geographical Adaptive Fidelity GAF
Sensing area is divided into small virtual grids.
Nodes in each two adjacent grids can communicate with each other.
Only 1 node within the same grid need to stay active
Geographic Random Forwarding GeRaF
Nodes switch back and forth between sleep and awake periodically
Each node has its own priority basing on its distance to the destination
Active nodes in the highest priority regions are selected to relay packages. Backoff if collision occurs
GPS is needed
Andy
The main idea of Topology control basing on location is virtual grid
1. Sensing area where nodes are deployed is divided into smaller virtual grids
2. Each virtual grid is defined such that for any 2 adjacent grids A and B, nodes in
grid A can communicate with nodes in B and vice-versa.
3. All node within the same grids are treated the same (equivalent) for routing.
4. Only 1 node needs to stay active at time therefore, they have to coordinate
with each other to decide who can sleep and for how long.
5. Initially, a node starts in the discovery state where it exchanges discovery messages with other nodes.
6. After broadcasting the message, the node enters active state and periodically rebroadcast its discovery message.
7. When it detects that some other node within its grid is handling or will handle the routing, it goes into sleeping state for T seconds (Sleeping time)
8. Nodes in the sleeping state wake up after T seconds and go back to the discovery state.
9. The note that stays active manage routing for the virtual grid it belongs to.
Geographic random forwarding is also classified as location-driven duty-cycled topology
1. In this protocol, all nodes switch back and forth periodically between sleep and awake state.
2. When they are active, they can participate in routing traffic it is needed.
3. When a node has a package to send, it becomes active and start to forward the package including
its location and the intended receiver location.
4. It then selects one active neighbor to relay the package toward the destination
5. Each node has its own priority basing on how close it is to the destination. The package is forwarded to
active nodes in the region with high priority.
6. Backoff is applied to resolve collision if multiple nodes transmit simultaneously
1. GPS: GPS is needed to identify location of each sensor node. GPS units are expensive
and consume a large amount of energy and . However, we don’t need to install it to all sensor nodes.
A limited subset of nodes can be equipped with a GPS and the location of these nodes can be derived
from the GPS.
2. The topologies can increase the network lifetime by a factor of 2-3 with respect to a network with nodes
alway one.

9. Topology Control: Connection driven
Span
Elected Coordinators with the “Coordinator eligibility rule”
Backup delay: each node generate its own delay time
ASCENT (Adaptive Self-Configuring Sensor Networks Topologies)
Each node decides to stay active or inactive based on local information
Initially, only some nodes are active.
When the sink node or active nodes experience with unacceptable package loss rate, they will send out help messages
Naps
Decentralized topology management protocol.
Andy
Span: connectivity driven protocol
1. Some nodes within the network are elected as COORDINATORS
2. Coordinators stay awake continuously and perform multi-hop routing other nodes are in sleeping mode
3. Sleeping nodes periodically check to see if they need to wake up and become a coordinator.
4. To guarantee a sufficient number of coordinators, SPAN uses the coordinator rule
1. Basically, if two neighbors of a non-coordinator node can not reach each other, either directly or through
one or more coordinators, then one of them should become a coordinator
2. When multi nodes decide to become coordinations at the same time due to the lack of coordinators,
a random backoff delay mechanism is carried on. Each node uses a function that generates random delay time b basing on the number of neighbors that can be connected by a potential coordinator node and its residual energy.
ASCENT: Adaptive self configuring sensor network topologies
(scalability, reduce package loss, independent of the routing protocol)
1. Is also connectivity-driven protocol
2. Each node decides to stay in active or inactive mode basing on information about
connectivity and package loss that are measured locally by the node itself.
3. Passive nodes have their radio on and listen to all packages but not transmit them.
4. Active nodes forward data and routing messages until they run out of energy.
5. When the sink or active nodes experience with unacceptable package loss rate, they will
send out help messages to its inactive neighbors and ask them to join the network to help
6. As soon as a passive node receive a help message, it starts monitoring the network conditions. I also signals its presence as an active node through a neighbor announcement message.
7. The process continues until the number of active nodes is good enough to handle traffic within the network
so that the sink node message loss rate is below predefined threshold.
Naps: Decentralized topology management protocol based on a periodic sleep/wakeup scheme
1. Time is split into time periods with duration T.
2. Each node waits for a random amount of time s, uniformly distributed into the range from 0 to T
3. when the waiting time is expired, the node broadcasts a HELLO message to advertise its activation to neighbors.
4. It also listens HELLO messages from other neighbors.
5. The node can go to sleep until next time period if it receives c HELLO messages for neighbors. Otherwise, it remains
active for all the time period T.

10. Sleep & Wake Conservation
Energy conservation can be achieved without the need to manage network topologies
Controlling sleep/wake cycles of sensor nodes accomplish this by implementing protocols at either the network or application layer
There are three primary categories under which sleep/wake protocols fall:
On-Demand
Scheduled Rendezvous
Asynchronous
Kevin

11. Sleep & Wake Approaches
On-Demand Protocols
Nodes should only wakeup when needed by other nodes
Generally achieve this by utilizing multiple radios with different power requirements
Scheduled Rendezvous Protocols
Neighboring nodes all wakeup in accordance to a scheduling algorithm
Woken nodes will remain active for a short period of time to communicate with their neighbors
Asynchronous Protocols
a node wakes up and communicates when it wants
Kevin

12. On-Demand Scheme
Nodes are generally in monitoring state (low-energy) and only switch to transfer state (high-energy) when event detected
STEM (Sparse Topology and Energy Management)
Two radios: one for wakeup, one for data transmission
Wakeup radio controlled through duty cycle and sends a stream of periodic beacons when it has to communicate with neighboring (target) nodes
Once the target receives a beacon, it will activate its data radio and send a wakeup ACK to the initiator node
The initiator node will activate its data radio upon receiving this acknowledgement and transfer can begin
Two flavors of STEM: STEM-B and STEM-T
Kevin

13. On-Demand Scheme (Cont.)
STEM can be combined with topology control protocols for greater power conservation
GAF in combination with STEM has been shown to reduce energy consumption in a sensor network to ~1% of its original utilization
STEM suffers from large wakeup latencies due to its asynchronicity and general low-bandwidths of sensor networks
An alternative On-Demand protocol that tries to balance energy savings and wakeup latency is Pipelined Tone Wakeup (PTW)
PTW does this by delegating responsibility of tone detection to the sender node instead of the receiver.
A wakeup tone is only sent when an event is detected

14. Low Duty Cycle MAC Conservation
TDMA main idea
Time is divided into frames
Each frame has one or more time slots
Each node is assigned one or more slots per frame
Nodes use assigned slots time transmit/receive data
Nodes turn on their radio during their time slot.
TDMA Drawbacks
Limited flexibility and scalability
Tight synchronization
Sensitive to interference
Perform worse in low traffic conditions
Andy
TDMA: Time division Multiple Access
1. Time is divided into into frame. Each frame has a certain number of time slots.
2. Every node is assigned to one or more slots per frames and use these slots for
transmitting/receiving packages to/from other nodes.
TDMA drawbacks:
1. Limited flexibility and scalability because in a real sensor network, channel conditions are changed frequently.
Node failures also happen quite often. Therefore slots allocation may be problematic.
2. TDMA-based protocols need tight synchronization. They are also very sensitive to interference.
3. Furthermore, TDMA-based protocols perform worse in low traffic conditions.

15. Low Duty Cycle MAC Conservation
Contention-based MAC protocols
More robust and scalable
Low delay and easily adapt to traffic conditions
Weakness: energy expenditure is high because of contention and collisions
Hybrid
Combine the strength of TDMA-based and contention-based protocols while offsetting their weaknesses
Complex to implement with a high number of nodes
Andy
Content-based MAC protocols:
1. Content-based MAC protocols are proposed to solve problems in TDMA-based MAC protocols.
2. There are more robust and scalable.
3. In addition, they have lower delay than TDMA-based protocols and can easily adapt to traffic conditions

16. Conclusions
While topology control protocols increase the performance of a sensor network, they often require GPS units which may make the project not affordable
On-Demand power conservation protocols are most effective because nodes are only active when they need to communicate and there is little impact on latency
However, the other methods of energy conservation (rendezvous and asynchronous) are less expensive
Many methods to reduce energy consumption, duty cycle controlling being one of the most prominent.
Kevin