Information Technology Reference
In-Depth Information
system have been applied with clear analogies between a robotic and a human im-
mune system. These are the notion of TLRs, inflammation and localisation. The sys-
tem has been developed with the help of a SOM as an adaptive state representation
of the robot, which enables local as well as global failure prevention and ratification.
A model has been implemented to support the above given principles. Results from
performed experiments show that the activity of TLRs causes an incremental in-
flammatory response over time, in case the robot is not returned to a stable state in
a reasonable period of time. This inflammatory response can be used along with the
SOM to locate the affected area of the robot in order to deal with it on a more global
level. The presented preliminary results support the described principles and en-
courage future development of a real robot implementation incorporating immune,
neural and endocrine control components.
Some of the potential advantages of this scheme are highlighted throughout
the earlier parts of the paper, but perhaps one of the most significant is that it of-
fers a relatively simple mechanism for integrating existing engineering knowledge
of how to deal with particular problems locally with the higher level and less well
defined parts of the control system. Some potential disadvantages include: that
the engineer must still manually assign fault conditions and remediation activi-
ties for local conditions which leaves room for oversight and error; the overhead of
maintaining a system-wide map of the robot's state may cause problems (whilst
maintaining the SOM is unlikely to be computationally expensive, the gathering
of its input data from all over the robot could be problematic); and last but by
no means least, it is not yet clear how such an innate system might fit into a full
multi-layer artificial immune system for a robot. Apart from the obvious next
step (implementing the system as described on a real robot), a pressing piece of
future work will be identification of how this might be achieved.
It is also interesting to consider the effect of the system on the combination
of task achieving behaviour and survival behaviour. Whilst the mechanism here
does not explicitly address this (potential) conflict, it does provide an interesting
possibility when combined with the neuro-endocrine control systems described
above and elsewhere. The “soft” switching, suppression and promotion of be-
haviours is precisely what this conflict requires in order to achieve the sorts of
complex trade-offs that are observed in nature. The addition of an inflammation
based driver for such behaviour mediation provides an additional homogeneous
driver
specifically for maintenance of homeostasis
.Thisisanimportantstep
forward as it provides a truly integrated mechanism for promotion of survival
behaviours within task achieving robot systems.
Acknowledgements
This paper arose as a direct result of the ARTIST Network
2
funded Student
Spring School held at Aberystwyth on April 8-13th 2006. Whilst the direct con-
tributors are listed as authors we would like to thank all of those who attended
2
ARTIST is an EPSRC (UK) funded network to support artificial immune systems
research.
