In order to expose the conclusions of our work we are going to summarize all the important tasks that we have done. Firstly, an instantiation of the artificial cognitive architecture is designed and developed. In order to ease this task we have used inference models developed by our research group with the help of SWIG. To provide our instantiation with the ability of self-optimization and self-learning we have designed and developed an off-line optimization algorithm based on cross entropy algorithm and an on-line learning algorithm based on Q-learning algorithm.
In addition, we developed classes that provide this instantiation with the ability of running in a distributed manner. In order to ease the task of environment configuration we have use the IceGrid service to avoid the task of entering IP directions of the hosts when we use the instantiation.
Finally, we test the instantiation in a simulation environment and in a real manufacturing environment, obtaining in both very good results. It is important to note that in the experiments we are using low-cost platform hardware, Raspberry Pi, to run our architecture. This gives our work an important added value.
So, relying on the presented results we can say that an artificial cognitive architecture build on low-cost platform hardware with viability to implement control system has been developed.
All the realized work gives us an important set point from which we can do a lot of future work, centering in the implementation of control system built on low-cost hardware.
We are aware of the fact that much work remains to be done in order to achieve the ideal artificial cognitive architecture, so the next objectives in the near future are:
- To design and develop a practical goal management complement.
- To add more models to our repository to be able to do more complex tests.
- To implement an easy way to do a deployment of the architecture without an user intervention.
- To improve the way in which our instantiation execute the components to achieve a better performance using the Raspberry Pi.
- To realize more complex test in order to prove the viability of the developed architecture in more environment.
Here is a video that shows an experiment using the developed artificial cognitive architecture: