Group of advanced Automation of Machines, Highly complex processes and Environments

Artificial Cognitive Control System for Micromanufacturing Processes. Method and Application (CONMICRO)
National Research Project.

Duration: 2012-2015

Nowadays there is a great interest, some work done, and on-going research on cognitive architectures for control, especially in the field of robotics. The main objective of CONMICRO is to design and implement a cognitive architecture for controlling micromanufacturing processes, specifically the micro-drilling and micro-milling processes. The basement to carry out CONMICRO is the project COGNETCON, prelude to this proposal. The scientific basis relies on the nexus between the paradigm of internal model control and brain-cerebellum connectivity as a basis of human intelligence, and a cognitive architecture based in imitation of own abilities and socio-cognitive experiences was then proposed.

CONMICRO aims to go beyond the current state of the art through research in four main pillars, namely: a stochastic optimization method based on hybridization with a view to self-optimization, a coordination mechanism by managing goals based on Artificial Intelligence techniques, a strategy of reinforcement learning to provide the architecture for learning capabilities and a computationally efficient strategy (based on AI techniques) of switching between modules that form the cognitive architecture. Thus, we intend to continue evolving artificial cognitive controllers with advanced neurofuzzy and fuzzy strategies that lead to appropriate local and global representation of micro-scale complex processes.

The main singularity of CONMICRO is the biological inspiration supported by the most recent results of Computer Science and Artificial Intelligence in conjunction with strategies and methods of Control Engineering and Automation. Therefore, the cognitive architecture is linked to control laws, strategies and performance indices of control engineering (e.g., internal model control) bringing an alternative scheme to perform efficiently control tasks, similar to how humans perform certain cognitive processes. While conventional control architectures are focused on solving problems with a hierarchy of goals and sub-goals, little has been done about how to achieve the same level of skills in cognitive-inspired architecture as intended by CONMICRO.

The expected scientific contributions are related to the technical contributions by means of a method that will support the practical feasibility of artificial cognitive control strategies for micromanufacturing processes. Mechanical micromanufacturing processes are manufacturing processes that are of great importance and complexity but the performance of current control strategies is inadequate and insufficient, or in many cases nonexistent. The developed method will be applied to micro scale processes (micro-milling and micro-drilling) in which control of micro-forces and micro-torque is essential. In short, the method of designing a cognitive control system will provide the what, how and why of goal-oriented behaviour in cooperating individual modules. CONMICRO is therefore focused on converging techniques in the field of COGNO-INFO-MICRO, according to the arguments above presented.