Skip to main content
main-content

About this book

This book presents interdisciplinary research that pursues the mutual enrichment of neuroscience and robotics. Building on experimental work, and on the wealth of literature regarding the two cortical pathways of visual processing - the dorsal and ventral streams - we define and implement, computationally and on a real robot, a functional model of the brain areas involved in vision-based grasping actions.

Grasping in robotics is largely an unsolved problem, and we show how the bio-inspired approach is successful in dealing with some fundamental issues of the task. Our robotic system can safely perform grasping actions on different unmodeled objects, denoting especially reliable visual and visuomotor skills.

The computational model and the robotic experiments help in validating theories on the mechanisms employed by the brain areas more directly involved in grasping actions. This book offers new insights and research hypotheses regarding such mechanisms, especially for what concerns the interaction between the dorsal and ventral streams. Moreover, it helps in establishing a common research framework for neuroscientists and roboticists regarding research on brain functions.

Table of Contents

Frontmatter

Chapter 1. Introduction

This book introduces a full framework of the sequence of sensorimotor transformations required to plan and execute hand movements suitable to grasp nearby objects. Present day research on robotic vision-based grasping has been compared with up-to-date neuroscience findings. The proposed framework has been conceived to be applied on a robotic setup, and the analysis of neuroscience findings has been performed taking into account not only biological plausibility, but also practical issues related to implementation constraints.

Eris Chinellato, Angel P. del Pobil

Chapter 2. The Neuroscience of Action and Perception

The visual cortex of humans and other primates is composed of two main information pathways, called

ventral stream

and

dorsal stream

in relation to their location in the brain. The traditional distinction (Ungerleider and Mishkin

1982

; Goodale and Milner

1992

) talks about the ventral “what” and the dorsal “where/how” visual pathways. In fact, the ventral stream is devoted to perceptual analysis of the visual input, such as in recognition, categorization, assessment tasks. The dorsal stream is instead concerned with providing the subject the ability of interacting with its environment in a fast, effective and reliable way. This second stream is directly involved in estimating position, shape and orientation of target objects for reaching and grasping purposes. The tasks performed by the two streams, their duality and interaction, constitute the neuroscientific basis of the research described in this book, and this chapter is devoted to a detailed explanation of the related concepts.

Eris Chinellato, Angel P. del Pobil

Chapter 3. Intelligent Robotic Grasping?

Mutual interest between the fields of robotics and cognitive sciences has been steadily growing in the recent years, especially through the bridging of artificial intelligence research. Nevertheless, the differences in goals and methodology and the lack of a common language make of true interdisciplinary research still a pioneering work. As the following review will expose, grasping is not an exception to this situation. A brief description of traditional and bio-inspired research in robotic vision-based grasping is presented and critically discussed in this chapter, with the purpose of defining a few important guidelines required to achieve proficuous cross-disciplinary research.

Eris Chinellato, Angel P. del Pobil

Chapter 4. Vision-Based Grasping, Where Robotics Meets Neuroscience

This chapter describes how we model the integration of on-line, action-oriented visual information (dorsal pathway) with knowledge about the target object and memories of previous grasping experiences and object characteristics (ventral pathway). Previous models of vision-based grasping have built so far mainly, when not exclusively, on monkey data. Recent neuropsychological and neuroimaging research has shed a new light on how visuomotor coordination is organized and performed in the human brain. Thanks to such research, a model of vision-based grasping which integrates knowledge coming from single-cell monkey studies with human data can be developed. The basic framework of the proposed model is outlined in this chapter. Final goal of the proposal is to mimic, in a robotic setup, the coordination between sensory, associative and motor cortex of the human brain in vision-based grasping actions.

Eris Chinellato, Angel P. del Pobil

Chapter 5. Extraction of Grasp-Related Visual Features

This chapter describes how we model the integration of on-line, action-oriented visual information (dorsal pathway) with knowledge about the target object and memories of previous grasping experiences and object characteristics (ventral pathway). Previous models of vision-based grasping have built so far mainly, when not exclusively, on monkey data. Recent neuropsychological and neuroimaging research has shed a new light on how visuomotor coordination is organized and performed in the human brain. Thanks to such research, a model of vision-based grasping which integrates knowledge coming from single-cell monkey studies with human data can be developed. The basic framework of the proposed model is outlined in this chapter. Final goal of the proposal is to mimic, in a robotic setup, the coordination between sensory, associative and motor cortex of the human brain in vision-based grasping actions.

Eris Chinellato, Angel P. del Pobil

Chapter 6. Visuomotor Transformations for Grasp Planning and Execution

This chapter deals with the tasks of transforming object properties, relevant for grasping purposes, into suitable hand configurations, and executing an appropriate grasping action on the target object. Analytical expressions of the transfer functions realized by particular types of neurons (surface and axis orientation selective) are derived and discussed. The obtained representations are used in grasp planning and execution, and the different projections to AIP and its job as the fundamental hub in programming and monitoring grasping actions are discussed.

Eris Chinellato, Angel P. del Pobil

Chapter 7. An Ever-Developing Research Framework

This chapter presents a number of issues and developments related to possible extensions of the model in Chaps.

4

,

5

and

6

. Some aspects are controversial and need additional neuroscience data to be modeled, others would need a different robotic setup to be properly tested. Some ideas have simply not been implemented yet, or have been implemented only partially, and thus were not described in the main modeling framework.

Eris Chinellato, Angel P. del Pobil
Additional information