The cerebellum has a central role in fine motor control and motor learning tasks. In this work, a bioinspired adaptive model, developed by means of a spiking neural network made of thousands of artificial neurons, has been leveraged to control a humanoid NAO robot in real-time. The learning properties of the system have been challenged in a classic cerebellum-driven paradigm, a perturbed upper limb reaching protocol. The neurophysiological principles used to develop the model, succeeded in driving an adaptive motor control protocol with warm up (baseline), acquisition and extinction phases. The spiking neural network model showed learning behaviors similar to the ones experimentally measured with human subjects in the same task in the acquisition phase, while resorted to other strategies in the extinction phase. The model processed in real-time external inputs, encoded as spikes, and the generated spiking activity of its output neurons was decoded, in order to provide the proper correction effect on the motor actuators.Three bidirectional long-term plasticity rules have been embedded for different connections and with different time-scales. The plasticities shaped the firing activity of the output layer neurons of the network. In the perturbed upper limb reaching protocol the neurorobot successfully learned how to compensate for the external perturbation generating an appropriate correction. Therefore, the spiking cerebellar model was able to reproduce in the robotic platform how biological systems deal with external sources of error, in both ideal and real (noisy) environment.
Il cervelletto ha un ruolo fondamentale nel controllo motorio fine e nell'apprendimento motorio. In questo lavoro, è stato utilizzato un modello adattativo bio-ispirato, sviluppando una rete neurale di tipo "spiking" composta da migliaia di neuroni artificiali, per controllare un robot umanoide in tempo reale. Le proprietà di apprendimento del sistema sono state verificate tramite un protocollo di raggiungimento con l'arto superiore, sottoposto ad una perturbazione esterna. I principi neurofisiologici utilizzati per sviluppare il modello hanno reagito con successo nel protocollo motorio adattativo composto da fasi alterne di acquisizione ed estinzione. Il modello di rete neurale ha esibito un comportamento simile a quello osservato in soggetti umani sottoposti allo stesso protocollo. Il modello ha elaborato in tempo reale gli input esterni codificandoli in potenziali d'azione, e decodificando l'output della rete per fornire un segnale di controllo appropriato per i motori del robot. Tre regole di plasticità bidirezionali e a lungo termine sono state implementate fra le diverse connessioni e con differenti scale temporali. Queste plasticità hanno modificato l'attività dei neuroni di output della rete. Nel protocollo di raggiungimento con arto superiore il neurorobot ha imparato con successo a compensare la perturbazione esterna, generando un'opportuna correzione. Il modello cerebellare utilizzato è stato quindi in grado di riprodurre in una piattaforma robotica il comportamento di sistemi biologici per la compensazione di disturbi esterni, sia in ambienti ideali (simulazione) che in ambienti reali (robot fisico), soggetti quindi a rumore.
Control of a humanoid robot by a bioinspired cerebellar simulator
MARTINA, DARIO
2017/2018
Abstract
The cerebellum has a central role in fine motor control and motor learning tasks. In this work, a bioinspired adaptive model, developed by means of a spiking neural network made of thousands of artificial neurons, has been leveraged to control a humanoid NAO robot in real-time. The learning properties of the system have been challenged in a classic cerebellum-driven paradigm, a perturbed upper limb reaching protocol. The neurophysiological principles used to develop the model, succeeded in driving an adaptive motor control protocol with warm up (baseline), acquisition and extinction phases. The spiking neural network model showed learning behaviors similar to the ones experimentally measured with human subjects in the same task in the acquisition phase, while resorted to other strategies in the extinction phase. The model processed in real-time external inputs, encoded as spikes, and the generated spiking activity of its output neurons was decoded, in order to provide the proper correction effect on the motor actuators.Three bidirectional long-term plasticity rules have been embedded for different connections and with different time-scales. The plasticities shaped the firing activity of the output layer neurons of the network. In the perturbed upper limb reaching protocol the neurorobot successfully learned how to compensate for the external perturbation generating an appropriate correction. Therefore, the spiking cerebellar model was able to reproduce in the robotic platform how biological systems deal with external sources of error, in both ideal and real (noisy) environment.File | Dimensione | Formato | |
---|---|---|---|
2018_10_Martina.pdf
Open Access dal 18/09/2019
Descrizione: Testo della tesi
Dimensione
21.55 MB
Formato
Adobe PDF
|
21.55 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/142663