You are on page 1of 4

2012 Eighth International Conference on Intelligent Environments

First analysis of custom WSN architecture as


hardware for pervasive computing
V.D. Velasco-Martínez, M.J. Flores-Medina and F.G. Flores-García
Departamento de Estudios de Posgrado e Investigación
Instituto Tecnológico de la Laguna
Torreón, México
e-mail: yoalieh@gmail.com

Abstract—This paper introduces some concepts of WSN and


Pervasive Computing, then presents AesSidhe, a WSN architec-
ture currently under design in Instituto Tecnológico de la Laguna.
The features and constraints of the hardware architecture are
described, and the current work in the software architecture
is described and commented. An analysis of usage of the
architecture in pervasive computer systems for three generations
is done after. Finally, the conclusions on the results of the analysis
of the architecture are presented.

Index Terms—Wireless Sensor Networks, Pervasive Computing

I. I NTRODUCTION
A. Wireless Sensor Networks
Wireless Sensor Networks (WSN) are a cluster of tiny Figure 1. Mandragora V1 mote picture, and Mandragora V2 3D View
devices able to measure, process and transmit real world
variables wirelessly to a data sink [1]. These devices ought to
be autonomous, thus they need a power supply, a measuring unnoticeable devices, wirelessly connected. These systems can
and processing unit and a wireless communication transceiver. detect events by using sensors, and actuate according to user
Each device is called a mote, in reference to one of the first requests. It focuses in each device being able to communicate
WSN proyects: SmartDust [2]. with others when they need. WSN and Wireless Sensor and
There are many constraints while implementing WSN in Actor Networks (WSAN) are examples of systems of this first
real world applications. The most important are enhancing generation.
power supply duration and the price of each mote. Actually, The second generation, context awareness, refers to the
the most implemented power supply are batteries, and having ability of using first generation systems to infer the context
to change them too frequently or in a large quantity, is (as defined by [9]) of all entitities involved in the system,
not practical. Energy harvesting has been used to recharge including user, system components and expected use. Ac-
batteries, or even to replace the power supply [3]. cording to the current context of entities, called a situation,
Reduction in the distance between motes permits lower information and services are presented to users, some activities
power transmission while using RF, as energy consumption are autoexecuted, and current context of entities can be tagged
rates raise exponentially for longer distances [4]. Wireless Per- for a later use.
sonal Area Networks (WPAN), as defined by IEEE 802.15.4 The third generation, called smartness, consists in the use
standard [5], are good for WSN with a high density mote of awareness and connectedness of entities to give meaning to
count in a range of 40m, or in longer distances by the use of the context of all entities in the system by the use of semantic
multi-hop mesh network protocols, as Zigbee [6]. There are ontologies. As there would be many nodes in the area, each
other protocols in development and early use right now. An one is an entity, and the whole system works as the interaction
example is 6loWPAN, that seeks the presence of the motes of needs and decisions taken by every entity. This system
in the Internet by wrapping Internet Protocol v6 packets [7] requires a larger deployment to be useful, and there is where
inside of 802.15.4 ones. projects of smart cities, as Oulu Urban Interactions research
program fit [10].
B. Generations in pervasive computing research
According to Ferscha[8], there are three generations (or C. Mandragora and AesSidhe projects
stages) in pervasive computing research. The first , con- Both projects started development in 2008, at Instituto
nectedness, refers to networked systems made of tiny and Tecnológico de la Laguna, as the need to create a cheap WSN

978-0-7695-4741-1/12 $26.00 © 2012 IEEE 343


DOI 10.1109/IE.2012.55
 "

  

 "

'(#


 !
 
 

   


 

  
  



(


 




 !"  &   %" 
$ '(# %   

#$% !"  %$ %

Figure 2. Hardware Top View Figure 3. Software organization

infrastructure showed up. Mandragora, the first project, had the • Power supply: Designed to work with a single CR232
goal of monitoring soil’s water content in a pecan tree field. battery. It has a voltage elevator and regulator, which can
As the budget was very low, commercial motes were out of be disabled by the microcontroller.
the scope. New electronic boards were designed as academical • Transceiver: It has a 802.15.4-2003 stack in hardware,
projects [11], [12]. and the electronic components needed to couple a 2.4GHz
During Mandragora’s development, the benefit to split the RF antenna. By design decision, the hardware only com-
electronic design, coding and application was seen. Front-end plies with 802.15.4 to test different network protocols
electronic design, operating system, and hardware libraries over it, as this is a heavily researched topic in WSN.
were grouped in a new project called Aes Sidhe. Its main goals • Data storage: The microcontroller is limited in EEPROM,
are being used as the base architecture of other application thus it needs an external flash memory, used to store
projects, and as a researching resource for programming and network routes, among other data.
testing WSN protocols. • Data acquisition: This module has an 8 channel, 12 bit
Mandragora became then the first application of Aes Sidhe ADC, with a programmable gain and voltage threshold.
architecture, targeted to agricultural projects. It includes the It has three banks of sensor power, and each bank can
back-end electronic design, assembly design, and specific be enabled or disabled to save energy. It is to be used in
programming for agricultural sensors. Fig. 1 shows one mote applications requiring good quality measurements.
of the first design, without wires. The next batch, using the III. A ES S IDHE S OFTWARE D ESCRIPTION
changes in Aes Sidhe, is expected to be ready on the first half
The firmware is coded in mikroC Pro for PIC (mikroElek-
of 2012.
tronika, Belgrade, Serbia). It shall consist of a wireless boot-
Aes Sidhe is still a work in progress, but the possibility of
loader, an operating system (OS), a hardware resource man-
being used in different pervasive computing tasks are analyzed
ager, an energy manager, a network protocol, and measuring
on this paper. This analysis will help with decision taking
and acting applications. Fig. 3 shows the organization of the
during the development of the architecture, to take into account
software blocks, and how they interact between them. The
what is needed for its use as part of a pervasive computing
bootloader is not shown.
system.
A. Operating System
II. A ES S IDHE H ARDWARE D ESCRIPTION
Aes Sidhe is an embedded system, thus, a monolithic ap-
Aes Sidhe is designed with modularity in mind. Version
proach was considered. The OS has all the hardware libraries
2.1, shown on Fig. 2, has 5 primary modules. Some are
needed to access hardware integrated circuits. It makes these
communicated by a SPI bus. For each application, modules
libraries available to the framework. Interrupt routines are
can be selected, used, mixed and modified as needed. More
defined in the OS as atomic actions.
optional modules are in design for foreseen applications, like
A priority aware scheduler is also part of the OS. At the
the DSP module.
moment, the response time between tasks to be executed is
• Microcontroller: It has a PIC18. It uses its internal oscil- 176μs when using a 4MHz oscillator. The individual tasks of
lator in low frequency for power saving mode, and has each process can request execution to the scheduler, though
the option to enable an external oscillator for fast speed the duration of each task is defined in each process.
processing and USB operation. It has a power saving
feature, reducing power consumption to a minimum. It B. Framework
works as the SPI bus master, and has three 10-bit analog The main hardware resource conflict happens when using
channels for low quality measurements. the SPI Bus because all circuits communicate through it.

344
 
 
  
 

Currently, the resource manager only takes care of the bus,
controlling which device has control over it, and informing  
other processes that the bus is busy and should wait for it.
   
The energy manager will take care of monitoring the use

of battery and, if available, of the harvested energy from the 

environment. It allows also to control the hardware resources


to save energy by slowing clocks and turning off unneeded
modules.
     
One of the main objectives of Aes Sidhe is testing different  
   
  



   
  
network protocols, thus a network protocol is not fixed. At this
moment, a flooding algorithm has been used to request data Figure 4. Rerouting and flagging on failure.
from specific motes. When other motes in the way receive
a request, they retransmit it back until it gets to its destiny.
The response is retransmitted the same way. Data already mesh topology, but can be programmed in the network stack
retransmitted is ignored to avoid bouncing. by using the promiscuous mode of the transceivers.
C. Applications V. U SES IN P ERVASIVE C OMPUTING
Acting and sensing applications are processes that use the
First, the current architecture will be analyzed from the
framework libraries to achieve the sensing or acting functions
perspective of each generation of pervasive systems. Finally,
of the motes. The framework will not report capabilities to the
that can be done with the additional modules currently in
applications, as they can be defined at compile time.
design.
D. Wireless Bootloader
A. Connectedness
Each mote can have its firmware reprogrammed by a master
With the right routing protocols, Aes Sidhe can be used to
node. It saves time when updating the system without moving
fulfill many tasks of the first generation of devices. Each mote
motes from their place, and allows fast protocol and appli-
would keep information of other motes which can be reached
cation testing in a safe way. It uses the Advanced Encrytion
through itself. When receiving data for one of such motes, it
Standard (AES) for security, and memory zones protection to
will retransmit it back so the neighbors get the message. This
avoid overwriting the bootloader.
would be repeated for many hops, until the message gets to
When the mote restarts, the bootloader checks if reprogram-
its place. This way data can be sent between all devices. In
ming was requested. If it is, it will start listening for the new
case one mote fails, the neighbors would notify that it doesn’t
firmware data packets. If it is not, it will start normal operation.
respond, and erase the neighbor from the list. An algorithm
IV. A RCHITECTURE C ONSTRAINTS would trigger so that new routes to motes affected by the
A. Processor Speed, Calculation Power, and Memory failure are calculated. After this, if there are motes no longer
reachable by new routes, an alert to the user would be issued.
PIC18F4550 [13] has an 8 bit, RISC, Harvard architec- This way, the network tries to stay always available. This
ture microprocessor. Oscillator speed can go from 31kHz to mechanism is illustrated in Fig. 4.
48MHz, and each instruction takes four oscillator cycles to be Data aggregation is achievable, though RAM memory needs
executed, meaning 7.75KIPS to 12 MIPS. to be cared for, because it is not plentiful. At the current
An 8x8 hardware multiplier is present in the ALU, allowing stage of development, the operating system is using half the
8x8 unsigned integer multiplication in one cycle. Division still available RAM by itself.
needs to be made in software, and takes a significant time to The network can get environmental information through
be done, as any kind of floating or fixed point operation. sensors. Some simple actors can be switched to a value (on/off,
Multiprocessing is not available, and for multithreading analog). Requests from the user can be done through spe-
there are three independent threads, the main program, and cialized 802.15.4 devices. Numerical data processing would
two more for low and high priority interrupts. There are no be slow because the lack of computing power, but feasible.
software interrupts available. Multimedia data could only be captured by specialized sensors
Data memory is 2kB, and program flash is 32kB. There is resuming the data, but cannot be processed or transmitted.
not a direct way of executing code in external memories in a
bus, or in data memory, limiting whole code size. B. Context Awareness
B. Wireless communication From the context awareness focus, Aes Sidhe motes can
sense environmental and devices information, like power us-
802.15.4-2003 maximum data transfer rate is 250 kb/s,
age, power supply, data traffic and the like. Motes can be used
meaning that the whole wireless traffic in the channel, in a
as context aggregators. Context can be inferred by application
40m zone is restricted to that. It is able to communicate in
of single rules with the information about environment and
point-to-point and star topologies. It does not has integrated
devices. Information and inferred context can be requested

345
   ! #$ 
 !  power, each mote could be able to analyze its own data and
" 
  
    


 
create simpler abstractions of data and complex signals, which
%&
   
 !!   

could be transmitted in less time. more complex processing
   !
  
  
  could be achieved using WDC. More complex rules could be
  ' 
  

    

created with this extra processing. By using a common middle-
ware, both WDC and distributed context aggregation should be
 easier to achieve, though still limited by the RAM memory. For

 ( 
decision taking, technologies like Artificial Neural Networks

)
and small Expert Systems could be implemented.

 
         


  

   

    ! VII. C ONCLUSIONS

 
  
 
 
  "

When Aes Sidhe first development stage is finished, the
Figure 5. An example of a second generation system using Aes Sidhe motes.
system is ready to be adapted for first generation systems,
though with poor user interaction. For second generation
systems, an application for context inferring, remote motes
from other motes, extending the context awareness of each data retrieval, and situations tagging can be made. The second
mote. Situations with this contextual information can be tagged stage of development shall include the DSP and gateways to
in each mote. For each situation, a mote can offer a service, WIFI and Bluetooth, for a good way of interacting with the
or execute an automatic action. user through personal devices.
The user data cannot be directly acquired. It still needs input
R EFERENCES
from the user through a computer attached to a mote, or a
specialized device sensing user information. [1] J. Zheng and A. Jamalipour, Eds., Wireless Sensor Networks: A Net-
working Persepective. United States of America: Wiley, 2009.
This actions can be done with the current motes, as long as [2] B. Warneke, K. S. J, and S. Dust, “Smart dust: Communicating with a
data is managed in binary form, because RAM restrictions. cubic-millimeter computer,” Classical Papers on Computational Logic,
Another possibility would be using a Wireless Distributed vol. 1, pp. 372–383, 2001.
[3] W. Seah, Z. A. Eu, and H.-P. Tan, “Wireless sensor networks powered
Computing (WDC) framework [14], enabling processing con- by ambient energy harvesting (wsn-heap) - survey and challenges,” in
text aggregation as data hops from mote to mote. An example Wireless Communication, Vehicular Technology, Information Theory and
is presented in Fig.5. Aerospace Electronic Systems Technology, 2009. Wireless VITAE 2009.
1st International Conference on, may 2009, pp. 1 –5.
C. Smartness [4] W. Dargie and C. Poellabauer, Fundamentals of Wireless Sensor
Networks: Theory and Practice, ser. Wireless Communications and
Smartness, being so complex, cannot be achieved by single Mobile Computing. John Wiley & Sons, 2010. [Online]. Available:
motes. Considering that this generation needs a big quantity http://books.google.com/books?id=8c6k0EVr6rMC
[5] “Part 15,4: Wireless medium access control (mac) and physical layer
of data, a lot of motes would be needed. As they are not (phy) specifications for low-rate wireless personal area networks (lr-
as expensive as other devices, this would be an opportunity wpans),” IEEE, Standard SH95127, Oct. 2003.
for Aes Sidhe motes. If nodes were using distributed context [6] “Zigbee specification,” Zigbee Alliance Inc., Specification 053474r17,
Jan. 2008.
aggregation, the data processing capability would be a lot [7] G. Montenegro, N. Kushalnagar, J. Hui, and D. Culler, “Transmission
bigger. Even with this processing power raise, Aes Sidhe of ipv6 packets over ieee 802.15.4 networks,” RFC 4944 (Proposed
motes don’t have enough RAM memory to keep all data, Standard), IETF, RFC 4944, Sep. 2007.
[8] A. Ferscha, “20 years past weiser: What’s next?” Pervasive Computing,
thus all semantic ontologies and data mining still needs heavy IEEE, vol. 11, no. 1, pp. 52 –61, jan. 2012.
processing in a data warehouse. [9] A. K. Dey, “Understanding and using context,” Personal Ubiquitous
Comput., vol. 5, pp. 4–7, January 2001. [Online]. Available:
VI. E XPECTED R ESULTS AND F UTURE WORK http://dx.doi.org/10.1007/s007790170019
[10] F. Gil-Castineira, E. Costa-Montenegro, F. Gonzalez-Castano, C. Lopez-
Aes Sidhe motes, can be well used in first generation Bravo, T. Ojala, and R. Bose, “Experiences inside the ubiquitous oulu
systems, with some delays in processing and the inability of smart city,” Computer, vol. 44, no. 6, pp. 48 –55, june 2011.
managing multimedia data. For being able to fetch informa- [11] V. Velasco-Martínez, “Diseño de nodos y de una red inalámbrica de
sensores para mediciones agropecuarias,” Master Thesis, Instituto Tec-
tion from the user, a gateway with protocols used by other nológico de la Laguna, Torreon, Mexico, Diciembre 2009, iSSN:0188-
commonly used devices needs to be made. 8060 RIITEC:12-TMCIE-2009.
By using clever algorithms and binary representations of [12] M. FloresMedina, “Implementación de una red inalámbrica de sen-
sores en aplicaciones agropecuarias,” Master Thesis, Instituto Tec-
data, tasks from the second generation can be done by the nológico de la Laguna, Torreoń, Coahuila, Febrero 2011, iSSN:0188-
motes. Extending it’s processing power to strengthen the con- 8060 RIITEC:01-TMCIE-2011.
text aggregation by using a small WDC framework is possible [13] “Pic18f2455/2550/4455/4550 datasheet - high-performance, enhanced
flash usb microcontrollers with nanowatt technology,” Microchip Tech-
too. For the third generation, they can be used in helping with nology Inc., Datasheet 39632E, 2009.
data and context extraction helping data warehouses in having [14] D. Datla, X. Chen, T. Tsou, S. Raghunandan, S. Hasan, J. Reed,
pre-proccessed data. C. Dietrich, T. Bose, B. Fette, and J.-H. Kim, “Wireless distributed
computing: a survey of research challenges,” Communications Magazine,
With the addition of the DSP module under design, math IEEE, vol. 50, no. 1, pp. 144 –152, january 2012.
operations should be faster, and multimedia data could be
analyzed (though not transmitted). With this extra processing

346

You might also like