Members |
Project Members are:
| Dittrich Rudi: | rdittrich at cosy, |
| Findeisen Lutz: | lfindeis at cosy, |
| Gitschthaler Werner: | wgitscht at cosy. |
|
Overview |
Intentions at the begining
We try to implement a fog detection system, consisiting of
interconnected telos motes. The Fog detection seems to fit perfectly
for the different sensors (humidity,temperature and light) provided by the telos motes.
The motes are able to build up an ad hoc network to provide
highest flexibility and easy expansion.
The system should work as follows: one mote
detects fog in its area and transmits this informations to all motes in range.
The recieving motes evaluate the information, e.g. the hops the message took ( is the message
from the mote which has detected the fog or has it been passed through by motes between the
detecting and the recieving mote ) According to these informations the motes can act e.g. turn
on warning lights or try to send the message to a control center.
|
What we actually did
The actual implementation we finally did, does not differ too much from the
idea you can read above. The ad hoc network will only work if a master node ( root node )
sends out a resync or init message. The information is passed to a laptop which
has a mote connected that works as a kind of a network bridge.
|
Engineering the project |
Intentions at the begining
We want to implement the project using two different approaches:
To assure that the project comes to an positve end we first try to implement our Fog Detection System
using tinyos, the simpler approach, because an abstraction layer is added on top of the
MSP430f149 Microcontroller and the sensors. Furthermore TinyOS comes with a special programming language called nesc
which is a derivat of C and allows rapid application development due its modularity.
If there is enough time left we will try to implent the project directly in C
to avoid the overhead (see above) that comes with tinyos
|
What we actually did
Implementing the whole system diretly with C did not work out. We realized very soon that it is impossible
to write the needed drivers (e.g. radio) in the given time so we pretty soon switched to implement
everything in nesc the language in which tinyos is done. We defined
our own little multihop message system, which is not only able to detect the motes in the network but
also provides some, but not very accurate, information where the motes are located, if the motes
are distributed along a street. It would be very interesting to improve this network architecture
to be even more reliable and more accurate about the positions of the motes.
For demonstration purpose we used Java to implement an application for prove of
concept and simulating a street environment. During our tests we not only verified that the multi hop message
protocol works with 3 hops ( meaning to motes are routing a message ) but that it is quite reliable ( the network connection
was never interupted ) which is much more the merit of the really well done telos motes than ours.
|
Design issues |
Intentions at the begining
It is a fact that the Fog Detection System not necessarily follows any real time constraints nevertheless we want to
incoperate some design principles heard in the lectures. As we explained above Tinyos already comes with strong abstraction, therefore
only the implementation with c will take advance of e.g. an e-machine. |
What we actually did
As mentioned above the Fog detection System is event driven and does not fulfil any real time constraints
nor has it to. Therefore an implementation of the e-machine seemed quite senseless and would only
cause more complexity and a little bit larger code. Moreover tinyos already comes with high
abstraction ( which includes scheduling ) and as we allready pointed out, we did not implement
the pure c version.
|