Motor-Primed Visual Attention for Humanoid Robots

We present a novel, biologically inspired, approach to an efficient allocation of visual resources for humanoid robots in a form of a motor-primed visual attentional landscape. The attentional landscape is a more general, dynamic and a more complex concept of an arrangement of spatial attention than the popular "attentional spotlight" or "zoom-lens" models of attention. Motor-priming of attention is a mechanism for prioritizing visual processing to motor-relevant parts of the visual field, in contrast to other, motor-irrelevant, parts. In particular, we present two techniques for constructing a visual "attentional landscape". The first, more general, technique, is to devote visual attention to the reachable space of a robot (peripersonal space-primed attention). The second, more specialized, technique is to allocate visual attention with respect to motor plans of the robot (motor plans-primed attention). Hence, in our model, visual attention is not exclusively defined in terms of visual saliency in color, texture or intensity cues, it is rather modulated by motor information. This computational model is inspired by recent findings in visual neuroscience and psychology. In addition to two approaches to constructing the attentional landscape, we present two methods for using the attentional landscape for driving visual processing. We show that motor-priming of visual attention can be used to very efficiently distribute limited computational resources devoted to the visual processing. The proposed model is validated in a series of experiments conducted with the iCub robot, both using the simulator and the real robot.

Enactive robot vision

Mototaka Suzuki

The complexity of today's autonomous robots poses a major challenge for Artificial Intelligence. These robots are equipped with sophisticated sensors and mechanical abilities that allow them to enter our homes and interact with humans. For example, today's robots are almost all equipped with vision and several of them can move over rough terrain with wheels or legs. The methods developed so far in Artificial Intelligence, however, are not yet ready to cope with the complexity of the information gathered through the robot sensors and the need for rapid action in partially unknown and dynamic environments. In this thesis, I will argue that the apparent complexity of the environment and of the robot brain can be significantly simplified if perception, behavior, and learning are allowed to co-develop on the same time scale. In doing so, robots become sensitive to, and actively exploit, characteristics of the environment that they can tackle within their own computational and physical constraints. This line of work is grounded on philosophical and psychological research showing that perception is an active process mediated by behavior. However, computational models of active vision are very rare and often rely on architectures that are preprogrammed to detect certain characteristics of the environment. Previous work have shown that complex visual tasks, such as position and size invariant shape recognition as well as navigation, can be tackled with remarkably simple neural architectures generated by a coevolutionary process of active vision and feature selection. Behavioral machines equipped with primitive vision systems and direct pathways between visual and motor neurons were evolved while freely interacting with their environments. I proceed further on this line of investigation and describe the application of this methodology in three situations, namely car driving with an omnidirectional camera, goal-oriented navigation of a humanoid robot, and cooperative tasks by two agents. I will show that these systems develop sensitivity to a number of oriented, retinotopic, visual features – oriented edges, corners, height – and a behavioral repertoire to locate, bring, and keep these features in sensitive regions of the vision system that allow them to accomplish their goals. In a second set of experiments, I will show that active vision can be exploited by the robot to perform anticipatory exploration of the environment in a task that requires landmark-based navigation. Evolved robots exploit an internal expectation system that makes use of active exploration to check for expected events in the environment. I will describe a third set of experiments where, in addition to an evolutionary process, the visual system of the robot can develop receptive fields by means of unsupervised Hebbian learning and show that these receptive fields are significantly affected by the behavior of the system and differ from those predicted by most computational models of visual cortex. Finally, I will show that these robots replicate the performance deficiencies observed in experiments of motor deprivation with kitten when they are exposed to the same type of motor deprivations. Furthermore, the analyses of our robot brains suggest an explanation for the deficiencies observed in kitten that have not yet been fully understood.

EPFL2007

Adaptive sensorimotor peripersonal space representation and motor learning for a humanoid robot

Micha Hersch

This thesis presents possible computational mechanisms by which a humanoid robot can develop a coherent representation of the space within its reach (its peripersonal space), and use it to control its movements. Those mechanisms are inspired by current theories of peripersonal space representation and motor control in humans, targeting a cross-fertilization between robotics on one side, and cognitive science on the other side. This research addresses the issue of adaptivity the sensorimotor level, at the control level and at the level of simple task learning. First, this work considers the concept of body schema and suggests a computational translation of this concept, appropriate for controlling a humanoid robot. This model of the body schema is adaptive and evolves as a result of the robot sensory experience. It suggests new avenues for understanding various psychophysical and neuropsychological phenomenons of human peripersonal space representation such as adaptation to distorted vision and tool use, fake limbs experiments, body-part centered receptive fields, and multimodal neurons. Second, it is shown how the motor modality can be added to the body schema. The suggested controller is inspired by the dynamical system theory of motor control and allows the robot to simultaneously and robustly control its limbs in joint angles space and in end-effector location space. This amounts to controlling the robot in both proprioceptive and visual modalities. This multimodal control can benefit from the advantages offered by each modality and is better than traditional robotic controllers in several respects. It offers a simple and elegant solution to the singularity and joint limit avoidance problems and can be seen as a generalization of the Damped Least Square approach to robot control. The controller exhibits several properties of human reaching movements, such as quasi-straight hand paths and bell-shaped velocity profiles and non-equifinality. In a third step, the motor modalities is endowed with a statistical learning mechanism, based on Gaussian Mixture Models, that enables the humanoid to learn motor primitives from demonstrations. The robot is thus able to learn simple manipulation tasks and generalize them to various context, in a way that is robust to perturbations occurring during task execution. In addition to simulation results, the whole model has been implemented and validated on two humanoid robots, the Hoap3 and the iCub, enabling them to learn their arm and head geometries, perform reaching movements, adapt to unknown tools, and visual distortions, and learn simple manipulation tasks in a smooth, robust and adaptive way. Finally, this work hints at possible computational interpretations of the concepts of body schema, motor perception and motor primitives.

EPFL2009

Visuomotor Coordination in Reach-To-Grasp Tasks: From Humans to Humanoids and Vice Versa

Luka Lukic

Understanding the principles involved in visually-based coordinated motor control is one of the most fundamental and most intriguing research problems across a number of areas, including psychology, neuroscience, computer vision and robotics. Not very much is known regarding computational functions that the central nervous system performs in order to provide a set of requirements for visually-driven reaching and grasping. Additionally, in spite of several decades of advances in the field, the abilities of humanoids to perform similar tasks are by far modest when needed to operate in unstructured and dynamically changing environments. More specifically, our first focus is understanding the principles involved in human visuomotor coordination. Not many behavioral studies considered visuomotor coordination in natural, unrestricted, head-free movements in complex scenarios such as obstacle avoidance. To fill this gap, we provide an assessment of visuomotor coordination when humans perform prehensile tasks with obstacle avoidance, an issue that has received far less attention. Namely, we quantify the relationships between the gaze and arm-hand systems, so as to inform robotic models, and we investigate how the presence of an obstacle modulates this pattern of correlations. Second, to complement these observations, we provide a robotic model of visuomotor coordination, with and without the presence of obstacles in the workspace. The parameters of the controller are solely estimated by using the human motion capture data from our human study. This controller has a number of interesting properties. It provides an efficient way to control the gaze, arm and hand movements in a stable and coordinated manner. When facing perturbations while reaching and grasping, our controller adapts its behavior almost instantly, while preserving coordination between the gaze, arm, and hand. In the third part of the thesis, we study the neuroscientific literature of the primates. We here stress the view that the cerebellum uses the cortical reference frame representation. The cerebellum by taking into account this representation performs closed-loop programming of multi-joint movements and movement synchronization between the eye-head system, arm and hand. Based on this investigation, we propose a functional architecture of the cerebellar-cortical involvement. We derive a number of improvements of our visuomotor controller for obstacle-free reaching and grasping. Because this model is devised by carefully taking into account the neuroscientific evidence, we are able to provide a number of testable predictions about the functions of the central nervous system in visuomotor coordination. Finally, we tackle the flow of the visuomotor coordination in the direction from the arm-hand system to the visual system. We develop two models of motor-primed attention for humanoid robots. Motor-priming of attention is a mechanism that implements prioritizing of visual processing with respect to motor-relevant parts of the visual field. Recent studies in humans and monkeys have shown that visual attention supporting natural behavior is not exclusively defined in terms of visual saliency in color or texture cues, rather the reachable space and motor plans present the predominant source of this attentional modulation. Here, we show that motor-priming of visual attention can be used to efficiently distribute robot's computational resources devoted to visual processing.

EPFL2015