Affordance as general value function: a computational model

نویسندگان

چکیده

General value functions (GVFs) in the reinforcement learning (RL) literature are long-term predictive summaries of outcomes agents following specific policies environment. Affordances as perceived action possibilities with valence may be cast into predicted policy-relative goodness and modeled GVFs. A systematic explication this connection shows that GVFs especially their deep-learning embodiments (1) realize affordance prediction a form direct perception, (2) illuminate fundamental between perception affordance, (3) offer scalable way to learn affordances using RL methods. Through an extensive review existing on GVF applications representative research robotics, we demonstrate provide right framework for real-world applications. In addition, highlight few new avenues opened up by perspective “affordance GVF,” including orchestrating complex behaviors.

برای دانلود باید عضویت طلایی داشته باشید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Affordance as context

The concept of affordance is relatively easy to define, but has proved to be remarkably difficult to engineer. This paradox has sparked numerous debates as to its true nature. The discussion presented here begins with a review of the use of the term from which emerges evidence for a two-fold classification—simple affordance and complex affordance. Simple affordance corresponds to Gibson’s origi...

متن کامل

Mathematics as a Computational Model

Simple basic mathematical notation and natural deduction are used everyday by the average person to do computation. This paper discusses the beauty of this model of computation and questions why this model has not been adopted. A brief description is given of a prototype implementation of this model. The main focus of the paper is on the considerable benefits the current mathematical model has ...

متن کامل

Affordance - based Computational Model of Driver Behavior on Highways : A Colored Petri Net Approach

In this paper, an affordance-based Colored Petri Net (CPN) model for representing driver behavior is proposed. We adopt a simulation-based approach and conduct an analysis of driver affordances on a highway-driving task. The computational CPN model is an extension of the initial conceptual CPN model and allows experimenters to enforce driving preferences as preferential turn probabilities for i...

متن کامل

Value Function Transfer for General Game Playing

We present value function transfer techniques for General Game Playing (GGP) by Reinforcement Learning. We focus on 2 player, alternate-move, complete information board games and use the GGP simulator and framework. Our approach is two-pronged: first we extract knowledge about crucial regions in the value-function space of any game in the genre. Then for each target game, we generate a smaller ...

متن کامل

A Functional Model for Affordance-Based Agents

Today’s mobile artificial agents, such as mobile robots, are based on an object-oriented paradigm. They partition their environment into various objects and act in relation to individual properties of these objects. Such perception and acting is insufficient for goal-directed behavior in dynamic environments, which requires action-relevant information in the form of affordances. Affordances des...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Adaptive Behavior

سال: 2021

ISSN: ['1059-7123', '1741-2633']

DOI: https://doi.org/10.1177/1059712321999421