Information Technology Reference
In-Depth Information
So how could this be implemented on the Motes? The Mica Motes are pro-
grammed in nesC and use TinyOS [35] for their program execution model. The
first step is the perception of the temperature sensor. A request is sent to the
ADC to sample temperature. On return the ADC gives a value in the range
0 to 1023.
2
Using a predefined conversion formula, the temperature is given in
degrees Celsius. This in effect causes the adoption of the following belief:
BELIEF(temp(25))
This is then matched to the left hand side of rule 1 in Fig. 10. Now in our
deliberate cycle we call the actuator
temperatureExceeded
, which will cause
the agent to adopt a belief that the threshold has been breached if in fact this
is the case. This will happen at the end of this current cycle and the agent will
pause. On the next cycle the agent will have:
BELIEF(thresholdIsExceeded(25))
So the agent will commit to calling the
sendToBase(25)
actuator, which will
inform the base station of the temperature. The actual representation of the
beliefs of the agent is represented in bytecodes for which we have implemented
an interpreter. A bytecode interpreter already exists for the Motes [19] but we
felt it would be more ecient to implement our own version tailored to our
specific needs.
Using bytecodes has one great advantage that we are currently exploring,
namely agent mobility. Bytecodes could encapsulate an agents mental state and
be sent to a remote node for incarnation. An interesting corollary to this could
be the investigation of the possibility of migrating an agent off the Mote network
and into a Java environment.
Looking to the future we are keen to investigate the migration of a
ยต
Agent-
Factory Agent across the whole sensor network for example to perform simple
data aggregation. We are also currently experimenting with mechanisms to im-
plement fully deliberative agents.
6 Conclusions
This chapter has advocated the adoption of agile mobile intentional agents as
the instrument of choice for the effective delivery of ambient intelligence. Such
systems by virtue of their complexity, dynamics, distribution and fault tolerance
need to be empowered with intelligence and autonomic capabilities.
We have described the Agent Factory environment together with the nature
of the agents fabricated and the tool and methodological support for this pro-
cess. We have furthermore illustrated, through three case studies, how such a
development metaphor can be used to deliver ambient intelligence systems.
Ambient intelligence systems of the future will exhibit all of the functional
characteristics contained within these three case studies and more and will typi-
2
Itisa10bitADC.
Search WWH ::
Custom Search