KDI: Learning Complex Motor Tasks in Natural and Artifical Systems
KDI: Learning Complex Motor Tasks in Natural and Artifical Systems
批准号:
9873474
负责人:
Stuart Russell
金额:
$120.0万
依托单位国家:
美国
项目类别:
Standard Grant
财政年份:
1998
资助国家:
美国
项目状态:
已结题
起止时间:
1998-10-01 至 2002-09-30
中文摘要
这个项目将开发一个统一的理论,说明自然系统和人工系统如何学习解决复杂的运动任务,如跑步、跳水、投掷和飞行,这些任务需要大量的感官输入和许多低级别活动的协调、排序和微调。这样的项目之所以可能,是因为我们在理解人类和其他动物的运动控制系统方面取得了重大的实验进展,也因为我们控制学习的数学模型越来越复杂。这些模型不仅将用于分析和预测运动控制中的自然现象,还将用于为执行复杂任务的人工系统推导有效的自适应控制器。要产生复杂的行为,必须将自然系统和人工系统分层组织,并进行多层抽象。因此,第一个研究任务将是确定适当的表示级别,在该级别上可以对物理系统进行建模,并且可以在该级别上定义控制动作。例如,在描述一只从A飞到B的昆虫时,可能的水平可能是1)神经信号和机械特性控制着每一只翼猫的详细形状2)基本的翼拍周期3)11引导“循环”飞行4)起飞、导航、着陆。将在各种实验环境和任务下进行详细的运动、力和/或气流测量,以建立正式模型和物理系统之间的对应关系。这些实验将对各种生物进行,可能包括在昆虫中飞行,在蟑螂中奔跑,在人类中奔跑、潜水和投掷。这些研究(对于昆虫,还有神经生理学研究)还将建立在控制系统的每个级别上可用的感觉输入。给定控制系统的一般结构和适当的感觉输入,下一步是设计能够学习成功执行给定任务的学习算法。要使用的学习方法是强化学习,这是一种旨在调整控制算法以优化目标函数的技术,即指定奖励信号的长期累积值。作为感觉输入的一部分,奖励被提供给学习算法。将开发新的强化学习算法,在分层控制结构中使用局部和全局奖励信号进行操作;此外,即使使用整体目标函数的非线性表示,这些算法也将被证明是收敛的。这项研究应该阐明这样一个核心问题,即这种在动物和人类中的学习形式是否可以被视为由优化或某些其他原则驱动,例如在系统的不同级别之间保持固定的接口特征。在动物身上发现一致的奖励功能,特别是在人类身上,将对一般学习理论产生重大影响。
英文摘要
9873474RussellThis project will develop a unified theory of how natural and artificial systems can learn to solve complex motor tasks, such as running, diving, throwing, and flying, that entail significant sensory input and the coordination., sequencing, and fine-tuning of many low-level activities. Such a project is possible because of significant experimental advances in our understanding of motor control systems in humans and other animals, and because of increased sophistication in our mathematical models of control learning. These models will be used not only to analyze and predict natural phenomena in motor control, but also to derive effective adaptive controllers for artificial systems carrying out complex tasks.To generate complex behaviors, natural and artificial systems must be organized hierarchically with multiple layers of abstraction. The first research task will therefore be to identify appropriate levels of representation at which the physical system can be modclled and at which control actions can be defined. For example, in describing an insect flying from A to B, possible levels might be 1) nerve signals and mechanical properties controlling the detailed shaping of each wingbcat 2) basic wingbeat cycle 3) 11 steering" the cycle to direct flight 4) takeoff, navigation, landing. Detailed motion, force, and/or airflow measurements will be made under a variety of experimental circumstances and tasks to establish the correspondence between formal models and physical systems. These experiments will be carried out for a variety of organisms, possibly including flying in insects, running in cockroaches, and for running, diving, and throwing in humans. These studies (and, in the case of insects, neurophysiological studies) will also establish the sensory inputs that are available at each level of the control system.Given the general structure of the control system and the appropriate sensory inputs, the next step is to design learning algorithms capable of learning to perform the given task successfully. The learning method to be used is reinforcement learning, a technique designed to adjust the control algorithm to optimize an objective function-that is, the long-term accumulated value of a specified reward signal. The reward is supplied to the learning algorithm as part of the sensory input. New reinforcement learning algorithms will be developed that operate using both local and global reward signals within a hierarchical control structure; furthermore, these algorithms will be proved to converge even using nonlinear representations of the overall objective function. This research should shed light on the central question of whether this form of learning in animals and humans can be viewed as driven by optimization or by some other principle, such as the preservation of fixed interface characteristics among the various levels of the system. Discovery of consistent reward functions in animals, especially humans, would have significant consequences for general theories of learning.
期刊论文(0)
专著(0)
科研奖励(0)
会议论文
Conference: Inaugural Workshop on Provably Safe and Beneficial AI (PSBAI)
-
批准号:2230996
-
项目类别:Standard Grant
-
资助金额:$9.39万
-
财政年份:2022
-
负责人:Stuart Russell
-
依托单位:
RI: Medium: Hierarchical Decision Making for Physical Agents
-
批准号:0904672
-
项目类别:Standard Grant
-
资助金额:$120.0万
-
财政年份:2009
-
负责人:Stuart Russell
-
依托单位:
REU Site: Computer Science in the Interest of Society (CSIS)
-
批准号:0754843
-
项目类别:Continuing Grant
-
资助金额:$23.76万
-
财政年份:2008
-
负责人:Stuart Russell
-
依托单位:
Learning Complex Probabilistic Models from Data
-
批准号:9634215
-
项目类别:Continuing Grant
-
资助金额:$33.94万
-
财政年份:1997
-
负责人:Stuart Russell
-
依托单位:
Research on Real-Time Decision Making: The Ralph Project
-
批准号:9211512
-
项目类别:Continuing Grant
-
资助金额:$34.68万
-
财政年份:1993
-
负责人:Stuart Russell
-
依托单位:
Real-Time Intelligent Control for an Automated Taxi
-
批准号:9309729
-
项目类别:Standard Grant
-
资助金额:$4.62万
-
财政年份:1993
-
负责人:Stuart Russell
-
依托单位:
Japanese Language Award for Gary Ogasawara
-
批准号:9207213
-
项目类别:Standard Grant
-
资助金额:$1.18万
-
财政年份:1992
-
负责人:Stuart Russell
-
依托单位:
Collaborative Research: Solving Chess with Probabilistic Planning and Control
-
批准号:9024557
-
项目类别:Standard Grant
-
资助金额:$0.95万
-
财政年份:1991
-
负责人:Stuart Russell
-
依托单位:
PYI: Architectures and Algorithms for Autonomous Intelligent Systems
-
批准号:9058427
-
项目类别:Continuing Grant
-
资助金额:$26.25万
-
财政年份:1990
-
负责人:Stuart Russell
-
依托单位:
Research on Real-Time Decision Making: The RALPH Project
-
批准号:8903146
-
项目类别:Continuing Grant
-
资助金额:$30.2万
-
财政年份:1989
-
负责人:Stuart Russell
-
依托单位:
国内基金
海外基金
登录
查看更多内容
Scalable Learning and Optimization: High-dimensional Models and Online Decision-Making Strategies for Big Data Analysis
-
批准号:--
-
项目类别:合作创新研究团队
-
资助金额:--
-
批准年份:2024
-
负责人:姚韬
-
依托单位:
Understanding structural evolution of galaxies with machine learning
-
批准号:
-
项目类别:省市级项目
-
资助金额:10.0万元
-
批准年份:2022
-
负责人:Nicola Rosario Napolitano
-
依托单位:
煤矿安全人机混合群智感知任务的约束动态多目标Q-learning进化分配
-
批准号:--
-
项目类别:青年科学基金项目
-
资助金额:30万元
-
批准年份:2022
-
负责人:吉建娇
-
依托单位:
基于领弹失效考量的智能弹药编队短时在线Q-learning协同控制机理
-
批准号:62003314
-
项目类别:青年科学基金项目
-
资助金额:24.0万元
-
批准年份:2020
-
负责人:沈剑
-
依托单位:
集成上下文张量分解的e-learning资源推荐方法研究
-
批准号:61902016
-
项目类别:青年科学基金项目
-
资助金额:24.0万元
-
批准年份:2019
-
负责人:万珊珊
-
依托单位:
具有时序迁移能力的Spiking-Transfer learning (脉冲-迁移学习)方法研究
-
批准号:61806040
-
项目类别:青年科学基金项目
-
资助金额:20.0万元
-
批准年份:2018
-
负责人:解修蕊
-
依托单位:
基于Deep-learning的三江源区冰川监测动态识别技术研究
-
批准号:51769027
-
项目类别:地区科学基金项目
-
资助金额:38.0万元
-
批准年份:2017
-
负责人:张大奇
-
依托单位:
具有时序处理能力的Spiking-Deep Learning(脉冲深度学习)方法研究
-
批准号:61573081
-
项目类别:面上项目
-
资助金额:64.0万元
-
批准年份:2015
-
负责人:屈鸿
-
依托单位:
基于有向超图的大型个性化e-learning学习过程模型的自动生成与优化
-
批准号:61572533
-
项目类别:面上项目
-
资助金额:66.0万元
-
批准年份:2015
-
负责人:孙雪冬
-
依托单位:
E-Learning中学习者情感补偿方法的研究
-
批准号:61402392
-
项目类别:青年科学基金项目
-
资助金额:26.0万元
-
批准年份:2014
-
负责人:秦继伟
-
依托单位: