This is an extract from the first draft of the thesis I am working on.
Please read it if you are sincerely working in Wireless Sensor Networks.
Oxford English dictionary defines sensor as ”a device which detects or measures a physical property and records, indicates, or otherwise responds to it”. A better definition from Wikipedia is ”a device that measures a physical quantity and converts it into a signal which can be read by an observer or by an instrument”. From the thermometers, barometers to microphone and cameras works with this principle making it integral part of our daily lives. A taxonomy of sensors can be drawn in the way in which it works such as (i) Active/Passive, (ii) Electrical/Mechanical/Electromechanical (iii)Proximity (How close the sensor has to be from the phenomenon).
The sensors in WSN research considers those which can produce electrical/digital information for a phenomenon of interest. Monitoring applications developed so far works in very close proximity of the phenomenon which will be a few centimetres (A structure health monitoring system or a patient monitoring system). On the other hand, a tracking system based on WSN prefers a sensor which has a better range (a few meters at least). Thus they sense energy emitted by the target (which is an energy propagation) unless it is a gas sensor or a chemical sensors which detects existence of a particle using chemical/biological properties. The problem of target tracking for surveillance mainly focuses on energy propagated from a target (source) which can be detected by using a sensors.
A distributed version of this concept employing inexpensive sensors with a low power wireless network for data collection created the field of target tracking using Wireless Sensor Networks. Despite the talks for years with hundreds of publications in this topic, hardly a few implementations came out. No reports exists on systems in large scale, which existed outside the research labs.
A closer look in to the main assumptions on the sensor part in target tracking reveals that they are not well formed. An analysis over the current assumptions has been done. The papers listed for analysis are mainly from journal articles with high impact factor or articles with a good number of citations.
R is the deterministic sensing range and R’ is the probabilistic sensing range. The area between R and R’ is the probabilistic range area where detection probability is assumed to be inversely proportional to distance
from R towards R’. Beyond R’ the sensing probability is assumed to be zero
Assumptions from literature
• It is assumed that all the sensor nodes have the same sensing range r and uncertainty sensing range r’
(r’≤1)[1]
• For simplicity, we assume that the sensing ranges of the sensors completely cover the region of interest with no overlap. In other words, the region can be divided into cells with each cell corresponding to the sensing range of a particular sensor [2]
• We assume the sensing range of a sensor is a disk with the sensor at the center. In this paper, we use homogeneous sensors, and hence all the sensors have the same sensing radius. Realizing that in real applications sensors may generate faulty readings due to measurement errors, we employ a practical sensing model .Sensors are assumed to correctly detect the presence and absence of targets within the inner disk Ac of radius Rc ; we call Rc as the reliable sensing radius. Otherwise, sensors can correctly detect the presence or absence of targets with only some nonzero probability when targets are present within disk Au of radius Ru but outside of Ac ,and Ru ≤ Rc [3].
• We assume binary sensing with a sensing range of Ri for a sensor Si , i.e., the footprint of sensor Si is a circle of radius Ri centered at Si inside which it can sense and outside of which it cannot sense[4].
• The sensing area of every node is assumed to be circular. Every node has the same sensing range (Rs) and communication range (Rc). The communication range is greater than two times that of the same sensing range. This is a sufficient condition for coverage to imply connectivity [5]
• For simplicity, we assume that each sensor has a circular sensing region of radius R: a sensor outputs a 1 if a target falls within the sensing disk of radius R centered at its loca- tion. The parameter R is termed the sensing range. However,our framework also applies to sensing regions of more complex shapes that could vary across sensors. We assume noise- less sensing for the time being: the sensor output is always 1 if a target is within its sensing range, and always 0 if there is no target within its sensing range, with 100% accuracy [6]
• With the model, a sensor with a nominal sensing range R can always detect a target’s presence if it is within R Re range from the sensor. No signal from beyond distance R is ever detected. And, the detection probability drops off continuously as the distance increases between R Re and R[7].
The fundamentals of sensing
Passive sensors watches the environment and generates an electric signal corresponding to the received energy. Examples of such sensors are passive infra-red and acoustic. Without much signal processing, these passive sensors will produce only intensity levels. The assumption of ”deterministic” circular range R will not be valid in any of these sensors simply because of the fact that it fully depends on the emitting energy at the source. Things become worse when there are obstacles in between which can absorb or change the energy. R thus becomes variable depending on emissivity of the source.
The active sensors generates the energy and measures the change in the energy reflected by the target. Using active sensors creates deterministic ranges, but it either has to be deployed in non-overlapped covering range or has to be enabled and disabled using a TDMA schedule. The disadvantage of using active sensors is that the presence of such a sensor can be detected as well as jammed easily. Another factor to be considered is that it consumes more energy than a passive sensor. This invalidate the theory that sensing range is a constant and a sensor reports correct values inside the sensing range.
Another assumption on boolean sensors, which produce 1 bit,thus reduce the overhead also does not hold especially in near real time target tracking scenario. The simple micro-controllers in motes uses 10-16 bit ADC which produces 2 bytes data per measurement. 1 Bit information has to be sent over wireless with a minimum protocol overhead of 19 bytes which hardly makes any difference in sending more accurate 2 byte data.
[1] X. Wang, J. Ma, S. Wang, and D. Bi, “Distributed energy optimization for target tracking in wireless sensor networks,” IEEE Transactions on Mobile Computing, vol. 9, pp. 73–86, 2010.
[2] J. Fuemmeler and V. Veeravalli, “Energy efficient multi-object tracking in sensor networks,” Signal Processing, IEEE Transactions on, vol. 58, no. 7, pp. 3742 –3750, 2010.
[3] D. Cao, B. Jin, S. K. Das, and J. Cao, “On collaborative tracking of a target group using binary proximity sensors,” Journal of Parallel and Distributed Computing, vol. 70, no. 8, pp. 825 – 838, 2010.
[4] P. Manohar and D. Manjunath, “On the coverage process of a moving point target in a non-uniform dynamic sensor field,” Selected Areas in Communications, IEEE Journal on, vol. 27, no. 7, pp. 1245 –1255, 2009.
[5] J.-P. Sheu and H.-F. Lin, “Probabilistic coverage preserving protocol with energy efficiency in wireless sensor networks,” in Wireless Communications and Networking Conference, 2007.WCNC 2007. IEEE, pp. 2631 –2636, 2007.
[6] N. Shrivastava, R. M. U. Madhow, and S. Suri, “Target tracking with binary proximity sensors: fundamental limits, minimal descriptions, and algorithms,” in Proceedings of the 4th international conference on Embedded networked sensor systems, SenSys ’06, (New York, NY, USA), pp. 251–264, ACM, 2006.
[7] W. Kim, K. Mechitov, J.-Y. Choi, and S. Ham, “On target tracking with binary proximity sensors,” in Proceedings of the 4th international symposium on Information processing in sensor networks, IPSN ’05, (Piscataway, NJ, USA), IEEE Press, 2005.
