The aim of the SPARK II Project is to develop, evaluate, optimize and generalize an insect brain inspired computational model. This is a completely new architecture for action-oriented perception, inspired by the basic principles of information processing by living systems and based on the concept of "self-organization".
It will take the advantage of the new insights offered by experimental neurobiology regarding the structure and function of relevant centers in the insect brain, devoted to action-oriented perception. These insights will be enriched by the addition of nonlinear spatial temporal dynamical systems able to show emerging patterns, used as “perceptual states”.
After a background phase, performed in the former FP6 funded project SPARK (herewith called SPARK I), where relevant centers in insects were focused, and also different types of spatial-temporal dynamics were investigated, the challenging idea within SPARK II is to introduce a new computational infrastructure mimicking an insect brain architecture. This architecture will be assessed, optimized and applied to different robotic structures, in order to prove its generality.
The architecture is envisaged to be hierarchical, based on parallel sensory-motor pathways, implementing reflex-driven basic behaviors, enriched with higher and more complex structures, where a mix of bio-inspired artificial neuropills (e.g. for attention-like processes, a short-term memory for planned paths, a memory for dangerous and for rewarding objects) and of more physically-inspired nonlinear lattices, able to generate complex dynamics, work concurrently to generate complex behaviors at the output motor layer.
The architecture will exploit a number of different sensors, processing signals distributed in space and time and also showing nonlinear dynamics. Perceptual processes are conceived as emerging pattern flows (result of a nonlinear spatial-temporal dynamics). Pattern meaning (concept generation) will be incrementally built upon information derived from sensors. It will influence the particular associated motor behavior with the concurrent dynamics generated by models of the relevant perception centers in insects.
The investigation will be focused on the following two main points:
1) theoretical insight, both from the analytical details of neural networks organized in lattices, and, from the biological point of view, with attention to relevant neural areas in insects, like the Mushroom Bodies and the Central Complex, devoted to the multimodal interaction and perception purposes;
2) application of the action-oriented perception models to different robotics architectures, like manipulators, mobile and bio inspired robots. This will show the generality of the proposed approach, suitable for application, with a minor effort, to very different artefacts, so addressing the perceptual process as based on general rules.
Finally, a really challenging task will be studied. It will consist in using the same model, applied to different rovers, to lead to the emergence of cooperation capabilities in order to perform tasks where one robot alone cannot succeed. The ultimate concept that will be proven is that merging complex dynamics and biological inspection in insect brain leads to the emergence of a powerful general system: a new insect brain computational model for perception.