Dissertação (mestrado)—Universidade de Brasília, Faculdade de Tecnologia, Departamento de Engenharia Elétrica, 2014. / Submitted by Ana Cristina Barbosa da Silva (annabds@hotmail.com) on 2014-11-25T17:23:31Z No. of bitstreams: 1 2014_RafaelCortesdePaiva.pdf: 7660330 bytes, checksum: eaad53db8e1c76edec638a3e30ee5f3e (MD5) / Approved for entry into archive by Raquel Viana(raquelviana@bce.unb.br) on 2014-11-25T17:58:53Z (GMT) No. of bitstreams: 1 2014_RafaelCortesdePaiva.pdf: 7660330 bytes, checksum: eaad53db8e1c76edec638a3e30ee5f3e (MD5) / Made available in DSpace on 2014-11-25T17:58:54Z (GMT). No. of bitstreams: 1 2014_RafaelCortesdePaiva.pdf: 7660330 bytes, checksum: eaad53db8e1c76edec638a3e30ee5f3e (MD5) / Nesse trabalho foi realizado o estudo de técnicas bio-inspiradas para gerar a marcha de um robô bípede. Foi utilizado o conceito de CPG, Central Pattern Generator (CPG), que é uma rede neural capaz de produzir respostas rítmicas. Elas foram modeladas como osciladores acoplados chamados de osciladores neurais. Para tanto foram utilizados alguns modelos de osciladores, o modelo de Matsuoka, o modelo de Kuramoto e o modelo de Kuramoto com acoplamento entre a dinâmica do oscilador e a dinâmica da marcha. Foram usados dois modelos de robôs, o Bioloid e o NAO. Para otimizar os parâmetros dos osciladores foram utilizados o Algoritmo Genético (AG), o Particle Swarm Optimization (PSO) e o Nondominated sorting Genetic Algorithm II (NSGA-II). Foi utilizada uma função de custo que através de determinadas condições tem como objetivo obter uma marcha eficiente. No NSGA-II, além dessa função de custo, foi utilizada outra função de custo que considera o trabalho realizado pelo robô. Além disso, também foi utilizada a aprendizagem por reforço para treinar um controlador que corrige a postura do robô durante a marcha. Foi possível propor um framework para obter os parâmetros dos osciladores e através dele obter uma marcha estável em ambas as plataformas. Também foi possível propor um framework utilizando aprendizagem por reforço para treinar um controlador para corrigir a postura do robô com a marcha sendo gerado pelo oscilador de Kuramoto com acoplamento. O objetivo do algoritmo foi minimizar a velocidade do ângulo de arfagem do corpo do robô, dessa forma, a variação do ângulo de arfagem também foi minimizada consequentemente. Além disso, o robô andou mais “cautelosamente” para poder manter a postura e dessa forma percorreu uma distância menor do que se estivesse sem o controlador. ______________________________________________________________________________ ABSTRACT / This document describes computational optimized bipedal robot gait generators. Thegaits are applied by a neural oscillator, composed of coupled central pattern generators(CPG), which are neural networks capable of producing rhythmic output. The models ofthe oscillators used were the Matsuoka model, Kuramoto model and Kura moto model withcoupling between the dynamics of the oscillator and dynamics of the gait. Two bipedalrobots, a NAO and a Bioloid, were used. The neural oscillators were optimized with threealgorithms, a Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Nondominatedsorting Genetic Algorithm II (NSGA-II). It was used a fitness function that has theobjective to obtain an efficient gait through some conditions. In NSGA-II, besides this fitnessfunction, another one was used that has the objective to minimize the work done by therobot. Additionally, reinforcement learning techniques were used to train a controller thatcorrects the robots gait posture. It was proposed a framework to obtain the parameters of theoscillators used and obtain efficient gaits in both robots. Also, it was proposed a frameworkusing reinforcement learning to train a controller to correct the robots gait posture. The objective of the algorithm was to minimize the pitch angular velocity, consequently the pitchangle standard deviation was minimized. Additionally, the robot moved with more “caution” and walked less compared with the walk without the posture controller.

Networks of interneurons within the spinal cord coordinate the rhythmic activation of muscles during locomotion. These networks are subject to extensive neuromodulation, ensuring appropriate behavioural output. Astrocytes are proposed to detect neuronal activity via Gαq-linked G-protein coupled receptors and to secrete neuromodulators in response. However, there is currently a paucity of evidence that astrocytic information processing of this kind is important in behaviour. Here, it is shown that protease-activated receptor-1 (PAR1), a Gαq-linked receptor, is preferentially expressed by glia in the spinal cords of postnatal mice. During ongoing locomotor-related network activity in isolated spinal cords, PAR1 activation stimulates release of adenosine triphosphate (ATP), which is hydrolysed to adenosine extracellularly. Adenosine then activates A1 receptors to reduce the frequency of locomotor-related bursting recorded from ventral roots. This entails inhibition of D1 dopamine receptors, activation of which enhances burst frequency. The effect of A1 blockade scales with network activity, consistent with activity-dependent production of adenosine by glia. Astrocytes also regulate activity by controlling the availability of D-serine or glycine, both of which act as co-agonists of glutamate at N-methyl-D-aspartate receptors (NMDARs). The importance of NMDAR regulation for locomotor-related activity is demonstrated by blockade of NMDARs, which reduces burst frequency and amplitude. Bath-applied D-serine increases the frequency of locomotor-related bursting but not intense synchronous bursting produced by blockade of inhibitory transmission, implying activity-dependent regulation of co-agonist availability. Depletion of endogenous D-serine increases the frequency of locomotor-related but not synchronous bursting, indicating that D-serine is required at a subset of NMDARs expressed by inhibitory interneurons. Blockade of the astrocytic glycine transporter GlyT1 increases the frequency of locomotor-related activity, but application of glycine has no effect, indicating that GlyT1 regulates glycine at excitatory synapses. These results indicate that glia play an important role in regulating the output of spinal locomotor networks.