Transcript Energy consumption sources in WSNs

Smart Sensors and Sensor Networks
Lecture 12
Energy constraint
(overview)
1
Smart Sensors and Sensor Networks
Problem: lifetime of WSNs






The required lifetime of a WSN depends on the application type and
can reach several years;
Usually the sensors cannot be accessed by people for economic or
geographical reasons;
As a consequence the sensors have to be autonomous, they have
to be powered by batteries;
In some applications one may find a small number of sensors with
fixed power supplies but they have predetermined, specific roles,
such as gateway to Internet;
The majority of sensors must have mobile power supplies;
In most cases the replacement or replenishment of the batteries is
difficult or impossible, so the energy is the main constraint in
designing and maintaining WSNs.
2
Smart Sensors and Sensor Networks
Energy consumption sources in WSNs
 Sources at sensor level and sources at network level;
 Sources at sensor level are in accordance with its main blocks:
sensing, processing, communicating and power supply;
 The energy spent by the sensing block depends on the
phenomenon to be sensed, duty cycle and sampling rate;




The phenomenon: from 0.4 mW (STCN temperature sensor) – 1250 mW
(FCS-GL1/2A4-AP8X-H1141 flow control sensor);
The duty cycle is the ratio between the working time and the operational
period;
High sampling rate and high duty cycle give high accuracy but also high
consumption;
A trade off must be established between the sampling rate, duty cycle
and accuracy; It depends on the application type being different for
environment monitoring, tracking targets and event triggered
applications;
3
Smart Sensors and Sensor Networks

The energy consumption of the processing block:

The energy consumption of the processor;

The energy consumption of the extra circuits (external memories + extra
logic);

The energy consumption of the processor: the switching energy and the
leakage energy;




Both are highly dependent on the supply voltage;
The switching energy is determined also by the total capacitance switched by
the running software;
This can be reduced by shortening the length of the wires between the circuits
and by decreasing the number of circuit inputs connected to the same output;
The leakage energy is consumed when no operation takes place;
4
Smart Sensors and Sensor Networks


The energy consumed by the communicating block is the most
consistent from the overall energy consumption of a sensor;
Researches have shown high differences between the
communicating and processing energy; examples:



At the Berkeley Mote sending and receiving 1 b costs about 1 mJ,
respectively 0.5 mJ, while 1 mJ is enough for executing about 120
instructions;
For other processors the ratio is even higher, for example 1/220-2900 for
the MEDUSA II nodes and 1/1500-2700 for the WIN nodes;
The communicating energy depends on several parameters:




modulation scheme,
data rate,
duty cycle and
the state in which the radio module is;
5
Smart Sensors and Sensor Networks

The state can be:




transmitting,
receiving,
idle or
sleep;

Transmitting and receiving energies may have close values, e.g. 35 and
38 mW for CC240 radio module, 111 and 111 mW for JN-DSJN513x, or
different values but in the same range, e.g. 42 and 29 mW for CC1000
and 36 and 9 mW for TR1000;

The same for receiving and idling energies;

The least energy is consumed in sleeping mode, dropping to units or
tens of μW;
6
Smart Sensors and Sensor Networks

The energy consumption sources at network level are due to:





MAC protocols,
routing,
network topology and
node deployment;
Energy is wasted in MAC protocols because of:






idle listening,
collision and congestion,
overhearing,
overemitting and
overheads;
Idle listening appears when a node listens an idle channel waiting
possible packets, collision and congestion require retransmission and
are favored by the high density of the network, overemitting happens
when a sender sends a packet to the receiver but this one is not ready
and the packet must be sent again and overheads means extra control
packets, such as beacons, RTS-CTS packets and ACK packets;
7
Smart Sensors and Sensor Networks

Routing protocols can be divided in 3 categories:






Direct approach,
Location based and
Attribute based or Data centric;
An example for the first category is the flooding type protocol; it is
very simple but not energy efficient;
In Location based routing sensors communicate based on their
location identity; this requires that all the nodes are aware of their
location and this can be achieved by adding GPS receiver to all the
nodes or to part of them; in the second case new protocols must be
used for grouping the nodes without GPS receivers around a node
with GPS receiver for establishing their positions; the GPS receivers
increase the cost and energy consumption;
In Data centric routing decisions are taken based on the data held
by the sensors rather than their location; energy efficiency can be
obtained by reducing the number of packets to be routed and, for
that, data aggregation and in-network processing should be used;
8
Smart Sensors and Sensor Networks

WSNs can be organized in: flat, hierarchical (or tiered) and clusterbased topologies;




In flat topologies, packets travel form node to node till the destination;
In hierarchical topology, the nodes are organized in groups controlled by
one of them, in several tiers; a node communicates only with its group
head;
In cluster-based topology, the nodes are organized in clusters, with a
cluster head and on the top of the cluster heads there is the so called
sink or central node;
The transmitting energy is determined by the formula Rα where R is
the transmitting range and α is an exponent with values between 2
and 4 depending on the transmission environment (free space or
space with multiple-path interferences or local noise); this shows
that energy can be saved by using multi hop communication instead
of single hop communication; it is of high importance to establish the
optimum number of hops;
9
Smart Sensors and Sensor Networks

There are 3 main approaches in node deployment: deterministic,
random and pattern-based or grid;



In deterministic deployment, the sensors are placed in predetermined
positions; this solution requires a good knowledge of the area to be
monitored and is suitable only for small-scale applications;
Random deployment is suitable for hostile environments and is scalable
to large-scale applications; the sensor nodes are thrown randomly in one
or several steps; because the environment is not known, generally many
sensors are necessary this increasing the cost of the network;
Grid-based deployment is simple and scalable; the nodes are placed in a
regular form, row-by-row, using a moving carrier; in reality, the placement
may be less regular due to terrain accidents and placement errors.
10
Smart Sensors and Sensor Networks
Classification of solutions:
 Solutions for:

Energy conservation:



Energy replenishment



Node level;
Network level;
Node level;
Network level;
The solutions for energy conservation at node level are oriented on:
sensing, processing and communicating;



The sensing energy can be minimized by: adaptive sampling,
hierarchical sensing, model-based active sampling and triggered
sensing;
The processing energy can be reduced by: dynamic voltage scaling,
dynamic frequency scaling and low power modes;
The communicating energy can be reduced by best modulation strategy,
reducing the number of bits and intelligent radio hardware;
11
Smart Sensors and Sensor Networks

The solutions for energy conservation at network level solutions are
grouped in:






MAC protocols,
Routing protocols,
In-network processing,
Data aggregation and
System partitioning;
The solutions for energy replenishment at node level are divided in:


Energy recovery and
Energy harvesting; energy harvesting solutions are based on converting:





solar energy,
mechanical energy,
wind energy,
thermal energy.
At network level, only the network replenishment solution was
identified.
12
Smart Sensors and Sensor Networks
13
Smart Sensors and Sensor Networks

Energy conservation solutions

Adaptive sampling:




Adaptive sampling means to adapt dynamically the sampling rate depending
on the application requirements and the available energy; it reduces the
number of samples by exploiting spatio-temporal correlations between data;
Temporal correlation may appear when the monitored phenomenon varies
slowly; spatial correlation may appear when the monitored phenomenon does
not change sharply between areas covered by neighboring nodes;
The number of samples can be reduced while maintaining a certain level of
accuracy;
Hierarchical sensing


Hierarchical sensing assumes that the same phenomenon is sensed by
sensors with different parameters: some sensors provide high accuracy but
they are energy hungry while other sensors are energy efficient but provide a
limited accuracy;
The final measurement is obtained by inferring the readings of all sensors; a
trade off must be established between the accuracy and the energy
consumption;
14
Smart Sensors and Sensor Networks

Model-based active sampling:



The idea is to predict the data that should be acquired instead of acquiring it,
hence saving the energy needed for sensing;
This can be done by using a model of the phenomenon which is build on top
of an initial set of sampled data; if the model is no more satisfactory, it must
be updated by using a new set of acquired data;
Triggered sensing:


It consists in using two types of sensors for the same phenomenon: sensors
with high accuracy and high energy consumption and sensors with low
accuracy and low energy consumption;
Unlike the hierarchical sensing, where all the sensors work at the same time,
here, the low energy sensors are firstly activated and when they sense some
activity in the monitored field the sensors with high accuracy are activated for
providing fine grained data;
15
Smart Sensors and Sensor Networks

Dynamic Voltage Scaling (DVS) and Dynamic Frequency Scaling (DFS):
2

The mathematical support consists in:
com
c
dd
P






 C fV
The formula gives the commutation power which is the main component of
the power consumption in CMOS circuits;C c
By reducing the level of Vdd the power decreases drastically since the impact
of the supply voltage is high;
As it is shown in literature, there is an optimal point for Vdd, where the energy
is minimum, but which is less than the threshold voltage (Vt) for most circuits;
the problem is that some circuits, e.g. SRAMs, do not operate reliably below
Vt;
The best idea is to use a multiple DVS: to maintain high values for the supply
voltage on critical paths and for critical tasks and to reduce the values when
the requirements are more relaxed;
The commutation power decreases linearly with the clock frequency; the first
disadvantage is the decrease of the throughput; the second disadvantage is
the increase of the latency which can lead to fail some time requirements;
DVS and DFS are solutions for decreasing the power consumption not
necessarily the energy consumption;
16
Smart Sensors and Sensor Networks

Low power modes:









Processors and radio modules from sensors may enter low power modes.
Such a circuit may have several low power modes;
For example the 80C51 microcontroller has Active mode, Idle mode and
Power-down mode;
In Idle mode the CPU is halted but the internal periphery is active this allowing
to sense the environment and to act on it;
In Power-down or Sleep mode the entire processor is halted this leading to
important energy savings;
The microcontroller needs 16 mA in Active mode, 3.7 mA in Idle mode and 50
μA in Power-down mode;
However, the energy saving is affected by the startup energy and startup time
a circuit needs for going from the low power mode in the active mode;
A processor needs time for ramping up phase-locked loops or voltage
controlled oscillators and for restoring the processor context;
During this transition time no operation is possible;
For example, the μAMPS-1 transceiver needs a startup time of 466 μs and
startup energy of 58 mW, [8]
17
Smart Sensors and Sensor Networks

Modulation strategy:



Reducing the number of bits:


The communication can became efficient by reducing the transmission time;
this can be obtained by codifying more than one bit per symbol, that is M-ary
modulation should be used;
The disadvantages are: M-ary modulation requires more complex digital and
analog circuitry than 2-ary modulation, M-ary modulation schemes require an
increased radiated power, for increasing M, to achieve the same bit error
target and choosing M-ary modulation schemes may be irrelevant in WSNs
where it is expected that most packets are short and for such packets the
startup energy dominates overall energy consumption;
The solution is based on compression and aggregation of data;
Intelligent radio hardware:


Generally, a sensor node may have 2 roles: to sense or to route packets
received from other sensors; it is shown that in a typical sensor network,
around 65% of all the received packets require a node to act as a router; in
most nodes, protocol functionality is implemented in the CPU, meaning every
packet, regardless its destination, is processed by CPU; this is a energy
waste for the CPU;
Intelligent radio hardware should be able to detect the packets that must only
be redirected and avoid their processing by the CPU;
18
Smart Sensors and Sensor Networks

MAC protocols:



When designing energy efficient MAC protocols, the following parameters
should be considered: network topology, deployment strategy, antenna mode,
controlling mechanisms, delay, throughput, QoS requirements and number of
channels to be used in communication;
For example, directional antennae in WSNs became an interesting alternative
to omnidirectional antennae due to their potential for high throughput and
reduced delay, interference and power transmission; by rotating the
orientation of directional antennae, the signal level at the receiver can be
increased;
Routing protocols:

Research on energy-efficient routing has two main targets:



minimizing energy cost per packet and
balancing energy consumption in the network;
If only the first target will be considered, some nodes will be overloaded, they
will deplete their energy faster and the network will became disconnected or
not operational;
19
Smart Sensors and Sensor Networks

In-network processing:


Data aggregation:


In-network processing or cooperative computing means that several sensors
jointly work to take a decision; sometimes a single sensor can not decide if an
event occurred or not and it has to collaborate with other sensors to take a
decision and to send it to a remote location; in-network processing saves
energy by increasing the weight of the local computation to communication’s
detriment; although some communication still exists this is between
neighboring nodes so the communication energy is low;
In order to reduce the size of the sent data, thus saving energy, a solution is
to aggregate data and to send once the relevant data; examples are
operations such as: sum, average, max, min or data fusion which consists in
combination of unreliable data measurements to produce a more accurate
signal; data fusion may be achieved by reducing the uncorrelated noise and
enhancing the common signal; the data aggregation problem can be
approached as a bicriteria optimization problem: to minimize energy
consumption of a sensor and the latency cost of a message;
System partitioning:

System partinioning means to allocate the intensive computation tasks to
certain nodes which have more powerfull processing and energy ressources;
another approach is to spread complex computation tasks among more
sensors to avoid overloading certain nodes;
20
Smart Sensors and Sensor Networks

Energy replenishment solutions:

Battery recovery:





This solution is based on the battery recovery or battery relaxing effect which
means that an empty or almost empty battery self-recharges when no current
is drawn from it;
This is based on chemical diffusion process within the cell;
For maximising the benefits a scheme must be developped for chousing and
adjusting idle periods of batteries before reaching the saturation threshold and
avoiding too much idling time;
Such a scheme should take into account the duty cycle parameter and
buffering strategies.
Energy harvesting:


Energy harvesting consists in extracting energy from the environment;
The final goal is the so called energy neutral operation, when the sensor node
operates without batteries having the energy harvested the only energy
source;
21
Smart Sensors and Sensor Networks

Solar energy:








Solar energy is converted to electrical energy by solar panels;
Generally, sensor nodes are small and light weight, so the size of the solar
panels must be limited;
Thus, an important parameter is the power density which can be obtained (in
W/m2 or W/kg); for example: Heliomote, a solar energy harvesting system
which uses a solar panel of area 3.75 inches x 2.5 inches; his panel outputs
60mA at a voltage of 3.3V; this power can recharge two AA-sized Ni-MH
battery of capacity 1800mAh each.
Another solution is based on the maximum power point tracking;
The energy converted from solar source depends also on the particular
location on the earth’s surface, time of day, latitude, atmospheric conditions
and incidence angle of the sun’s beams;
The average values for the solar energy received are from 300 W/m2 near the
ecuador to 100 W/m2 near the poles;
The solar energy varies also with the season being 5-25 times less in winter
than in summer for temperate regions, depending on atmospheric conditions
too;
Storage elements for the energy become necesary for ensuring the proper
functioning of the WSN during all seasons, day and night;
22
Smart Sensors and Sensor Networks

Mechanical energy:

Mechanical energy from pressure, vibrations or force can be converted in
electrical energy by using piezoelectric materials;



Walking can generate voltage if a piezo-electric element is mounted in a shoe;
Pushing buttons/keys can also generate electricity;
Vibrations are another source for electrical energy obtained by converting
mechanical energy:




The generated electrical energy depends on the amplitude of the vibration, on
its frequency and on the extend to which the presence of the harvesting device
affects the vibration;
This, in turn, is affected by the difference between the masses of the harvesting
device and the vibrating mass;
Values for the available energy ranges from 0.1 μW/cm 3 up to 10000 μW/cm3;
Vibrations are present in cars and in build environments; measurements for a
number of vibration sources have shown that the amplitude and frequency
varies from 12 m/s2 at 200 Hz in a car engine compartment to 0.2 m/s 2 at 100
Hz for the floor in an office building with the majority of measured sources
having the fundamental frequency in the range 60-200 Hz;
23
Smart Sensors and Sensor Networks

Wind energy:


Thermal energy:


Wind energy is converted through rotors and turbines that convert circular
motion into frequency and than in voltage; the principle of electromagnetic
induction is used; the main disadvantage consists in the size of the turbine
which is too big compared with the requirements for most sensor nodes;
Thermoelectric generators generates electrical potential from thermal
difference between two points; high efficiency requires high temperature
difference; the literature describes solutions which used the thermal difference
between the air and soil, the ambient temperature difference and solid-states
thermoelectric generators; a commercial device provides 100 μW from a 10 K
temperature difference in a 9.3 mm diameter device 1.4 mm thick;
Electrostatic energy:

It is a form of mechanical – electrical energy conversion; it is based on
changing capacity of vibration-dependent varaiable capacitors; vibrations
separates the plates of an initially charged variable capacitor; the main
advantages are the possibility to integrate with microelectronics and the fact
that they do nod need any smart material and the main disadvantage is the
need of initially charging the capacitor;
24
Smart Sensors and Sensor Networks

Radio frequency energy:




Harvesting radio frequency energy is done in RFID (Radio Frequency
Identification) systems;
RFID tags are used to identify, locate and track people, animals and assets;
There are active and passive RFID tags; passive RFID tags power themselfs
by using the radio frequency energy emited by active RFID tags; passive
RFID tags report to active RFID tags their ID and location specific data; for
harvesting the received energy, passive RFID tags must be tuned to the
frequency of the radio source and the distance between an active and a
passive RFID tag shoud be in the range of few meters; contrarly, the
harvesting process will have very low efficiency because the ambient level of
RF frequency is low and spread over a wide spectrum and converting it would
require large broadband antennas;
Network replenishment:


Network replenishment extends network lifetime by adding new nodes, on the
fly, after the initial deployment of nodes;
This ensures energy saving because a minimum number of nodes have to be
deployed initially; additionally, this increases network lifetime by replacing died
nodes with new ones;
25
Smart Sensors and Sensor Networks
Conclusions


High independency is obtained by harvesting the environment but
the disadvantages are low efficiency and the fluctuation in time of
the generated level of electrical energy;
The solutions consist in developing energy usage profiles which
match the energy generated or to use energy storage elements.
26