Architectures and bio-inspired algorithms based on the principles of self-organization to control robots and machines

Research themes:

The research unit studies, through the construction of computational models of neural networks and robotic models, phenomena connected with the formation, operation and mutation of nervous cell networks, on the basis of the auto-organization principles typical of complex systems. The research aims will focus on the study the biological basis and will model the following phenomena:

• Neural network parameters and architecture evolution.

• Dynamic neuron network operation characterized by auto-organization.

• Learning based on auto-organization principles (ex.: Hebb rule, Hebb rule in time, competitive Hebb).

• Mechanism for autonomous continuous and cumulative motor-sensor skills learning based on intrinsic motivation.

• Mechanisms for trial and error learning.

• Operation of emotional and reason systems which adjust the learning (e.g.: classical conditioning, tool conditioning, devaluation, etc.).

Objectives:

• Understand auto-organization mechanisms which permit the brain's nervous networks to implement robust behavioural responses to noise and flexible responses to changes of the internal and external environment; transfer the knowledge acquired in algorithms for autonomous robot control.

• Understand the biological basis of brain's learning mechanisms based on auto-organization principles (genetic algorithms, Hebb rule, Hebb rule in time; competitive Hebb) and use these mechanisms to produce powerful computational mechanisms.

• Build neural architectures and autonomous, continuous and cumulative learning mechanisms of motor-sesnsor skills based on intrinsic motivations (e.g.: entropy decrease, news, rate of increase in the ability to anticipate events), learning by trial and error and hierarchical structures, using these skills to build robot with flexible behaviour.

• Understand the biological basis and model the emotional and motivational systems which regulate learning in organisms (ex.: classic conditioning, tool conditioning, devaluation, etc.).

Return to the Innovative computing models for Bioinformatics Research Area

Other Research Units in the Innovative computing models for Bioinformatics Research Area:

Integration of ontologies and programming techniques for the representation and management of workflows