logo SBA

ETD

Archivio digitale delle tesi discusse presso l’Università di Pisa

Tesi etd-06132018-152640


Tipo di tesi
Tesi di laurea magistrale
Autore
MASSI, ELISA
URN
etd-06132018-152640
Titolo
Design and implementation of a bio-inspired locomotion control architecture for a modular robot
Dipartimento
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
BIONICS ENGINEERING
Relatori
relatore Prof.ssa Laschi, Cecilia
relatore Dott. Falotico, Egidio
correlatore Dott.ssa Tolu, Silvia
Parole chiave
  • Brain-inspired control architecture
  • Central Pattern Generator
  • Cerebellum
  • Spiking neural network
Data inizio appello
05/07/2018
Consultabilità
Non consultabile
Data di rilascio
05/07/2088
Riassunto
Current robotic control strategies are mainly based on trajectory planning that adjust the movements based on the next desired state. These control policies do not efficiently perform where the dimensionality of the control problem increases or disturbances from the external environment affect the system behaviour. These issues are critical in locomotion tasks and the need for different control methods arises. To make autonomous robots able to move in a real and dynamic environment, controllers inspired by brain mechanisms have been proposed.
This research project focuses on the design and validation of a brain-inspired control architecture for robotic locomotion. The proposed control architecture involves the interaction among different motion controllers whose communication represents a simplified model of the neural locomotion control in vertebrates.
The presented solution combines classical control strategies with reservoir computing and spiking neural networks to obtain a scalable and adaptable controller by taking advantage of the different learning properties of neural networks. The features and the advantages of a central pattern generator (CPG) as a trajectory planner for robotic locomotion are described. We integrated a cerebellar-like neural network in the control loop in order to provide an adaptive contribution at the joint level. The experiments are performed on the Neurorobotic Platform (NRP) which offers a simulation environment where an interface between spiking neural network and simulated or real robots is easily achievable. The experiments on a simulated modular robot (Fable) show improvements on the locomotion performances of the robot in terms of position error, time delay and covered distance, compared to a control strategy where just the CPG network is taken into consideration. In conclusion, we can derive that a bio-inspired control strategies to design new efficient and adaptable robotic locomotion controller is possible.
File