![](bg6.png)
scenario II. That means, the reservation way (AllTra-
jectoriesVector) has approximately a quadratic com-
plexity, whereas the reservation way (PhotoOfGrid)
has approximately a cubic complexity, because time
is additional to the (x,y) form this 3-D configura-
tion. Since the reservation way (AllTrajectoriesVec-
tor) outperforms significantly the other reservation
way (PhotoOfGrid) by computation time in several
situations, we have further measured only the first
way (AllTrajectoriesVector).
Figure 6 shows the system performance (average
response time) for the scenario I (Equal-Equal) after
3000 ticks using the reservation method (AllTrajec-
toriesVector). We have measured in this scenario the
Figure 6: The average response time of system in scenario I
(Equal-Equal) after 3000 ticks using the reservation method
(AllTrajectoriesVector).
average response time of the system in the case that
the traffic level of cars in south-north and west-east
directions is (5) cars/tick, whereas we have repeated
the measurement in the cases that the maximum num-
ber of cars in each direction is 20, 40, 80 and 100 cars.
Note here, on the x-axis is the total amount of cars in
the two directions together. The value of the average
response time of the system in this scenario is about
(0.1) ms when the total number of cars in the two di-
rection is (40) cars, whereas about (0.83) ms by (200)
cars. This means, that the average response time of
the system increases approximately quadratically re-
quiring less that (1) ms when the total number of cars
in the two direction is (200) cars.
5 CONCLUSIONS
The choice of the appropriate o/c realization is a de-
sign decision that has to be done by the developer in
the design phase of the technical system. In this pa-
per, we presented a new approach towards building a
robust hybrid central/self-organizing multi-agent sys-
tem. This approach solves the conflict between a
central planning algorithm and the autonomy of the
agents. Additionally, it aims to increase the auton-
omy of agents in the fully central architecture. This
means, our hybrid concept tolerates that some agents
behave in fully autonomous way in the central o/c ar-
chitecture. The scenario used in this paper is a traffic
intersection without traffic lights. We introduced con-
trol features of the system designed to deal with de-
viations from the plan which can occur in the system
behaviour. These control features intend to intervene
in due time when it is necessary so that the system
remains demonstrating robustness. Finally, as evalu-
ation metric to measure the system performance, we
used the response time in four different test scenarios
in various combinations of parameters.
6 FUTURE WORK
Since we implemented the generic o/c architecture
adapted to our traffic scenario and accomplished our
experiments assuming that no deviations occur in the
system, the next step is to continue with the imple-
mentation of the case when deviations occur in the
system to completely realize our vision. Then, we
will measure the system performance and compare
the two cases, the system performance with and with-
out deviations. This comparison will be used to deter-
mine whether the system performance remains effec-
tive as long as possible when a deviation occurs and
consequently to assure a robust system.
REFERENCES
Branke, J., Mnif, M., M
¨
uller-Schloer, C., Prothmann, H.,
Richter, U., Rochner, F., and Schmeck, H. (2006).
Organic computing - addressing complexity by con-
trolled self-organization. In ISoLA, pages 185–191.
Cakar, E., Mnif, M., M
¨
uller-Schloer, C., Richter, U., and
Schmeck, H. (2007). Towards a quantitative notion of
self-organisation. In IEEE Congress on Evolutionary
Computation.
Chaaban, Y., H
¨
ahner, J., and M
¨
uller-Schloer, C. (2009). To-
wards fault-tolerant robust self-organizing multi-agent
systems in intersections without traffic lights. In Cog-
nitive09: Proceedings of The First International Con-
ference on Advanced Cognitive Technologies and Ap-
plications, November 15-20, 2009 - Athens, Greece.
IEEE, (To appear).
Mnif, M., Richter, U., Branke, J., Schmeck, H., and M
¨
uller-
Schloer, C. (2007). Measurement and control of self-
organised behaviour in robot swarms. In ARCS 2007,
volume 4415 of LNCS, pages 209–223. Springer.
Richter, U., Mnif, M., Branke, J., M
¨
uller-Schloer, C.,
and Schmeck, H. (2006). Towards a generic ob-
server/controller architecture for organic computing.
In INFORMATIK 2006. Bonner K
¨
ollen Verlag.
ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence
346