Introducing a proprio-ceptive feedback in the bio-inspired Tegotae control approach : enhanced learning and energy efficiency

In bio-inspired robotic control, CPGs (Central Pattern Generators) architectures have been extensively adopted to generate periodic patterns in the actuation, due to the interesting properties of nonlinear oscillators which they make use of. Although sensory feedback in CPGs is not necessary for the generation of the patterns, it plays a central role in guaranteeing adaptivity to the environmental conditions. On the other hand, feedbacks permit to derive information about the state of the system, but their inclusion do not usually lead to a different formulation of the control policy. This generally remains unvaried with respect to the open-loop case. Finally their inclusion greatly modifies the dynamics of the CPGs architecture, often leading to bifurcations and other interesting dynamical phenomena, which remain untreated. A study about a novel proprio-ceptive feedback design in CGPs is performed, together with the definition of a control policy directly based on it. The additional effects in terms of learning and energy efficiency are shown, and the occurring dynamical phenomena are outlined.

In robotica bioispirata, le architetture CPGs (Central Pattern Generators)sono usate largamente per generare pattern periodici nell'attuazione. Il nostro studio si concentra sul design di un nuovo feedback propriocettivo insieme ad una nuova legge di controllo su esso basato. Sono mostrate le proprietà addizionali in termini di apprendimento e efficienza energetica, e sono delineati i fenomeni dinamici che avvengono.