68 resultados para MOTOR CONTROL
Resumo:
Animals repeat rewarded behaviors, but the physiological basis of reward-based learning has only been partially elucidated. On one hand, experimental evidence shows that the neuromodulator dopamine carries information about rewards and affects synaptic plasticity. On the other hand, the theory of reinforcement learning provides a framework for reward-based learning. Recent models of reward-modulated spike-timing-dependent plasticity have made first steps towards bridging the gap between the two approaches, but faced two problems. First, reinforcement learning is typically formulated in a discrete framework, ill-adapted to the description of natural situations. Second, biologically plausible models of reward-modulated spike-timing-dependent plasticity require precise calculation of the reward prediction error, yet it remains to be shown how this can be computed by neurons. Here we propose a solution to these problems by extending the continuous temporal difference (TD) learning of Doya (2000) to the case of spiking neurons in an actor-critic network operating in continuous time, and with continuous state and action representations. In our model, the critic learns to predict expected future rewards in real time. Its activity, together with actual rewards, conditions the delivery of a neuromodulatory TD signal to itself and to the actor, which is responsible for action choice. In simulations, we show that such an architecture can solve a Morris water-maze-like navigation task, in a number of trials consistent with reported animal performance. We also use our model to solve the acrobot and the cartpole problems, two complex motor control tasks. Our model provides a plausible way of computing reward prediction error in the brain. Moreover, the analytically derived learning rule is consistent with experimental evidence for dopamine-modulated spike-timing-dependent plasticity.
Resumo:
At the crossing between motor control neuroscience and robotics system theory, the paper presents a rhythmic experiment that is amenable both to handy laboratory implementation and simple mathematical modeling. The experiment is based on an impact juggling task, requiring the coordination of two upper-limb effectors and some phase-locking with the trajectories of one or several juggled objects. We describe the experiment, its implementation and the mathematical model used for the analysis. Our underlying research focuses on the role of sensory feedback in rhythmic tasks. In a robotic implementation of our experiment, we study the minimum feedback that is required to achieve robust control. A limited source of feedback, measuring only the impact times, is shown to give promising results. A second field of investigation concerns the human behavior in the same impact juggling task. We study how a variation of the tempo induces a transition between two distinct control strategies with different sensory feedback requirements. Analogies and differences between the robotic and human behaviors are obviously of high relevance in such a flexible setup. © 2008 Elsevier Ltd. All rights reserved.
Resumo:
In order to understand the underlying mechanisms of animals' agility, dexterity and efficiency in motor control, there has been an increasing interest in the study of gait patterns in biological and artificial legged systems. This paper presents a novel approach to the study of gait patterns which makes use of intrinsic mechanical dynamics of robotic systems. Each of these robots consists of a U-shape elastic beam and exploits free vibration to generate different gait patterns. We developed a conceptual model for these robots, and through simulation and real-world experiments, we show three distinct mechanisms for generating four different gait patterns in these robots. © 2012 IEEE.
Resumo:
Legged locomotion of biological systems can be viewed as a self-organizing process of highly complex system-environment interactions. Walking behavior is, for example, generated from the interactions between many mechanical components (e.g., physical interactions between feet and ground, skeletons and muscle-tendon systems), and distributed informational processes (e.g., sensory information processing, sensory-motor control in central nervous system, and reflexes) [21]. An interesting aspect of legged locomotion study lies in the fact that there are multiple levels of self-organization processes (at the levels of mechanical dynamics, sensory-motor control, and learning). Previously, the self-organization of mechanical dynamics was nicely demonstrated by the so-called Passive Dynamic Walkers (PDWs; [18]). The PDW is a purely mechanical structure consisting of body, thigh, and shank limbs that are connected by passive joints. When placed on a shallow slope, it exhibits natural bipedal walking dynamics by converting potential to kinetic energy without any actuation. An important contribution of these case studies is that, if designed properly, mechanical dynamics can generate a relatively complex locomotion dynamics, on the one hand, and the mechanical dynamics induces self-stability against small disturbances without any explicit control of motors, on the other. The basic principle of the mechanical self-stability appears to be fairly general that there are several different physics models that exhibit similar characteristics in different kinds of behaviors (e.g., hopping, running, and swimming; [2, 4, 9, 16, 19]), and a number of robotic platforms have been developed based on them [1, 8, 13, 22]. © 2009 Springer London.
Resumo:
In human and animal running spring-like leg behavior is found, and similar concepts have been demonstrated by various robotic systems in the past. In general, a spring-mass model provides self-stabilizing characteristics against external perturbations originated in leg-ground interactions and motor control. Although most of these systems made use of linear spring-like legs. The question addressed in this paper is the influence of leg segmentation (i.e. the use of rotational joint and two limb-segments) to the self-stability of running, as it appears to be a common design principle in nature. This paper shows that, with the leg segmentation, the system is able to perform self-stable running behavior in significantly broader ranges of running speed and control parameters (e.g. control of angle of attack at touchdown, and adjustment of spring stiffness) by exploiting a nonlinear relationship between leg force and leg compression. The concept is investigated by using a two-segment leg model and a robotic platform, which demonstrate the plausibility in the real world. ©2008 IEEE.
Resumo:
This article discusses the issues of adaptive autonomous navigation as a challenge of artificial intelligence. We argue that, in order to enhance the dexterity and adaptivity in robot navigation, we need to take into account the decentralized mechanisms which exploit physical system-environment interactions. In this paper, by introducing a few underactuated locomotion systems, we explain (1) how mechanical body structures are related to motor control in locomotion behavior, (2) how a simple computational control process can generate complex locomotion behavior, and (3) how a motor control architecture can exploit the body dynamics through a learning process. Based on the case studies, we discuss the challenges and perspectives toward a new framework of adaptive robot control. © Springer-Verlag Berlin Heidelberg 2007.
Resumo:
There is an increasing attention of exploiting compliant materials for the purpose of legged locomotion, because they provide significant advantages in locomotion performance with respect to energy efficiency and stability. Toward establishing a fundamental basis for this line of research, a minimalistic locomotion model of a single legged system is explored in this paper. By analyzing the dynamic behavior of the system in simulation and a physical robotic platform, it is shown that a stable locomotion process can be achieved without the necessity of sensory feedback. In addition, further analysis characterizes the relation between motor control and the natural body dynamics determined by morphological properties such as body mass and spring constant. © 2006 The authors.
Resumo:
It has been shown that sensory morphology and sensory-motor coordination enhance the capabilities of sensing in robotic systems. The tasks of categorization and category learning, for example, can be significantly simplified by exploiting the morphological constraints, sensory-motor couplings and the interaction with the environment. This paper argues that, in the context of sensory-motor control, it is essential to consider body dynamics derived from morphological properties and the interaction with the environment in order to gain additional insight into the underlying mechanisms of sensory-motor coordination, and more generally the nature of perception. A locomotion model of a four-legged robot is used for the case studies in both simulation and real world. The locomotion model demonstrates how attractor states derived from body dynamics influence the sensory information, which can then be used for the recognition of stable behavioral patterns and of physical properties in the environment. A comprehensive analysis of behavior and sensory information leads to a deeper understanding of the underlying mechanisms by which body dynamics can be exploited for category learning of autonomous robotic systems. © 2006 Elsevier Ltd. All rights reserved.
Resumo:
There is ample evidence that humans are able to control the endpoint impedance of their arms in response to active destabilizing force fields. However, such fields are uncommon in daily life. Here, we examine whether the CNS selectively controls the endpoint impedance of the arm in the absence of active force fields but in the presence of instability arising from task geometry and signal-dependent noise (SDN) in the neuromuscular system. Subjects were required to generate forces, in two orthogonal directions, onto four differently curved rigid objects simulated by a robotic manipulandum. The endpoint stiffness of the limb was estimated for each object curvature. With increasing curvature, the endpoint stiffness increased mainly parallel to the object surface and to a lesser extent in the orthogonal direction. Therefore, the orientation of the stiffness ellipses did not orient to the direction of instability. Simulations showed that the observed stiffness geometries and their pattern of change with instability are the result of a tradeoff between maximizing the mechanical stability and minimizing the destabilizing effects of SDN. Therefore, it would have been suboptimal to align the stiffness ellipse in the direction of instability. The time course of the changes in stiffness geometry suggests that modulation takes place both within and across trials. Our results show that an increase in stiffness relative to the increase in noise can be sufficient to reduce kinematic variability, thereby allowing stiffness control to improve stability in natural tasks.