检索结果-内蒙古大学图书馆

An examination of different fitness and novelty based selection methods for the evolution of neural networks

SOFT COMPUTING 2013年第5期17卷 753-767页

作者： Inden, Benjamin Jin, Yaochu Haschke, Robert Ritter, Helge Sendhoff, Bernhard Univ Bielefeld Res Inst Cognit & Robot D-33615 Bielefeld Germany Univ Surrey Dept Comp Guildford GU2 5XH Surrey England Univ Bielefeld Neuroinformat Grp D-33615 Bielefeld Germany Honda Res Inst Europe Offenbach Germany

It has been suggested recently that it is a reasonable abstraction of evolutionary processes to use evolutionary algorithms that select individuals based on the novelty of their behavior instead of their fitness. Here we study the performance of fitness- and novelty-based search on several neuroevolution tasks. We also propose several new algorithms that select both for fit and for novel individuals, but without weighting these two criteria directly against each other. We find that behavioral speciation, behavioral near neutral speciation, and behavioral novelty speciation perform best on most tasks. Pure novelty search, as well as a number of hybrid methods without speciation mechanism, do not perform well on most tasks. Using behavioral criteria for speciation often yields better results than using genetic criteria.

关键词： Neuroevolution Selection Novelty search evolutionary robotics NEAT

来源：评论

学校读者我要写书评

暂无评论

Evolution of swarm robotics systems with novelty search

引用

SWARM INTELLIGENCE 2013年第2-3期7卷 115-144页

作者： Gomes, Jorge Urbano, Paulo Christensen, Anders Lyhne FCUL LabMAg Lisbon Portugal Inst Telecommun Lisbon Portugal Inst Univ Lisboa ISCTE IUL Lisbon Portugal

Novelty search is a recent artificial evolution technique that challenges traditional evolutionary approaches. In novelty search, solutions are rewarded based on their novelty, rather than their quality with respect to a predefined objective. The lack of a predefined objective precludes premature convergence caused by a deceptive fitness function. In this paper, we apply novelty search combined with NEAT to the evolution of neural controllers for homogeneous swarms of robots. Our empirical study is conducted in simulation, and we use a common swarm robotics task-aggregation, and a more challenging task-sharing of an energy recharging station. Our results show that novelty search is unaffected by deception, is notably effective in bootstrapping evolution, can find solutions with lower complexity than fitness-based evolution, and can find a broad diversity of solutions for the same task. Even in non-deceptive setups, novelty search achieves solution qualities similar to those obtained in traditional fitness-based evolution. Our study also encompasses variants of novelty search that work in concert with fitness-based evolution to combine the exploratory character of novelty search with the exploitatory character of objective-based evolution. We show that these variants can further improve the performance of novelty search. Overall, our study shows that novelty search is a promising alternative for the evolution of controllers for robotic swarms.

关键词： evolutionary robotics Neuroevolution Swarm robotics Novelty search NEAT Behavioural diversity Deception

来源：评论

学校读者我要写书评

暂无评论

Simulating Robots Without Conventional Physics: A Neural Network Approach

引用

JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS 2013年第3-4期71卷 319-348页

作者： Pretorius, C. J. Du Plessis, M. C. Cilliers, C. B. Nelson Mandela Metropolitan Univ Dept Comp Sci ZA-6031 Port Elizabeth South Africa

The construction of physics-based simulators for use in evolutionary robotics (ER) can be complex and time-consuming. Alternative simulation schemes construct robotic simulators from empirically-collected data. Such empirical simulators, however, also have associated challenges. This paper therefore investigates the potential use of Artificial Neural Networks, henceforth simply referred to as Neural Networks (NNs), as alternative robotic simulators. In contrast to physics models, NN-based simulators can be constructed without requiring an explicit mathematical model of the system being modeled, which can simplify simulator development. The generalization abilities of NNs, along with NNs' noise tolerance, suggest that NNs could be well-suited to application in robotics simulation. Investigating whether NNs can be effectively used as robotic simulators in ER is thus the endeavour of this work. Two robot morphologies were selected on which the NN simulators created in this work were based, namely a differentially steered robot and an inverted pendulum robot. Accuracy tests indicated that NN simulators created for these robots generally trained well and could generalize well on data not presented during simulator construction. In order to validate the feasibility of the created NN simulators in the ER process, these simulators were subsequently used to evolve controllers in simulation, similar to controllers developed in related studies. Encouraging results were obtained, with the newly-evolved controllers allowing experimental robots to exhibit obstacle avoidance, light-approaching behaviour and inverted pendulum stabilization. It was thus clearly established that NN-based robotic simulators can be successfully employed as alternative simulation schemes in the ER process.

关键词： Robotic simulation Artificial Neural Networks evolutionary robotics Controller development

来源：评论

学校读者我要写书评

暂无评论

Robotic locomotion of three generations of a family tree of dynamical systems. Part I: Passive gait patterns

引用

NONLINEAR DYNAMICS 2013年第3期73卷 1969-1989页

作者： Tavakoli, Ali Hurmuzlu, Yildirim So Methodist Univ Dept Mech Engn Dallas TX 75205 USA

This paper includes the latest results of our research effort in developing a family of energetically efficient robots that can produce stable gait. Gravity powered bipeds provide ample proof for the possibility of designing robots that can walk using minimal energy. It is conceivable that mechanisms that are much simpler than legged robots can also produce gravity powered locomotion. In this paper, we present the results of our study of the first three generations of the family: a single mass, a two-mass, and a three-mass system. We show that these three systems can generate a rich set of passive gaits such as hopping, tapping, and walking each including various gait modes. These periodic passive gaits are sustained based on conditions for the physical parameters of the systems. We have developed a preliminary "contact based rule of passive gait patterns" that seems to work well for the first three generations of the family. We think that this rule can be extended to more complex generations of the family. The control issues are presented in Part II of this paper.

关键词： Multi-body dynamical systems Hybrid-systems Passive systems Impulsive control evolutionary robotics

来源：评论

学校读者我要写书评

暂无评论

Linear reactive control for efficient 2D and 3D bipedal walking over rough terrain

引用

ADAPTIVE BEHAVIOR 2013年第1期21卷 29-46页

作者： Solomon, Joseph H. Locascio, Mark A. Hartmann, Mitra J. Z. Northwestern Univ Dept Biomed Engn Evanston IL 60208 USA Northwestern Univ Dept Mech Engn Evanston IL 60208 USA

The kinematics of human walking are largely driven by passive dynamics, but adaptation to varying terrain conditions and responses to perturbations require some form of active control. The basis for this control is often thought to take the form of entrainment between a neural oscillator (i.e., a central pattern generator and/or distributed counterparts) and the mechanical system. Here we use techniques in evolutionary robotics to explore the potential of a purely reactive, linear controller to control bipedal locomotion over rough terrain. In these simulation studies, joint torques are computed as weighted linear sums of sensor states, and the weights are optimized using an evolutionary algorithm. We show that linear reactive control can enable a seven-link 2D biped and a nine-link 3D biped to walk over rough terrain (steps of similar to 5% leg length or more in the 2D case). In other words, the simulated walker gradually learns the appropriate weights to achieve stable locomotion. The results indicate that oscillatory neural structures are not necessarily a requirement for robust bipedal walking. The study of purely reactive control through linear feedback may help to reveal some basic control principles of stable walking.

关键词： Bipedal walking dynamic walking biped reactive control rough terrain evolutionary robotics

来源：评论

学校读者我要写书评

暂无评论

Cooperative Transport by a Swarm Robotic System Based on CMA-NeuroES Approach

引用

JOURNAL OF ADVANCED COMPUTATIONAL INTELLIGENCE AND INTELLIGENT INFORMATICS 2013年第6期17卷 932-942页

作者： Yu, Tian Yasuda, Toshiyuki Ohkura, Kazuhiro Matsumura, Yoshiyuki Goka, Masanori Hiroshima Univ Grad Sch Engn 1-4-1 Kagamiyama Higashihiroshima Hiroshima 7398527 Japan Shinshu Univ Grad Sch Sci & Technol Ueda Nagano 3868567 Japan Fukuyama Univ Grad Sch Engn Hiroshima 7290251 Japan

Swarm robotic systems consist of many homogeneous autonomous robots with no type of global controllers. It is difficult to design controllers for such behavioral systems, since the behavior of system level is the emergent results of the dynamical interaction between the systems and the environment. This paper uses a method that in which robot controllers are designed by a covariance matrix adaptation evolution strategy (CMA-ES) with artificial neural networks. This approach is referred to as CMA-NeuroES. Among the many evolutionary algorithms for evolving artificial neural networks, two conventionally representation approaches, fast evolution strategies and differential evolution, are used for comparison. The cooperative transport problem is used as a benchmark of swarm robotic systems to test their performance. Results show that CMA-NeuroES has the overall best performance of the three.

关键词： swarm robotic systems evolving artificial neural network evolutionary robotics covariance matrix adaptation evolution strategy

来源：评论

学校读者我要写书评

暂无评论

Evolving locomotion for a 12-DOF quadruped robot in simulated environments

引用

BIOSYSTEMS 2013年第2期112卷 102-106页

作者： Klaus, Gordon Glette, Kyrre Hoyin, Mats Univ Oslo Dept Informat N-0373 Oslo Norway

We demonstrate the power of evolutionary robotics (ER) by comparing to a more traditional approach its performance and cost on the task of simulated robot locomotion. A novel quadruped robot is introduced, the legs of which - each having three non-coplanar degrees of freedom - are very maneuverable. Using a simplistic control architecture and a physics simulation of the robot, gaits are designed both by hand and using a highly parallel evolutionary algorithm (EA). It is found that the EA produces, in a small fraction of the time that takes to design by hand, gaits that travel at two to four times the speed of the hand-designed one. The flexibility of this approach is demonstrated by applying it across a range of differently configured simulators. (C) 2013 Elsevier Ireland Ltd. All rights reserved.

关键词： evolutionary robotics Simulation Locomotion

来源：评论

学校读者我要写书评

暂无评论

Generating Cooperative Collective Behavior in Swarm Robotic Systems

引用

JOURNAL OF ADVANCED COMPUTATIONAL INTELLIGENCE AND INTELLIGENT INFORMATICS 2013年第5期17卷 699-706页

作者： Ohkura, Kazuhiro Yasuda, Toshiyuki Matsumura, Yoshiyuki Hiroshima Univ Grad Sch Engn 1-4-1 Kagamiyama Higashihiroshima 7398527 Japan Shinshu Univ Fac Text Sci & Technol Ueda Nagano 3868567 Japan

Swarm robotics research involves multirobot systems that consist of many homogeneous autonomous robots but no global controller. In this paper, an evolutionary robotics approach using an artificial neural network is applied to a swarm robotic system. Conventionally, the neural network evolved using only synaptic weights under the condition of a fixed topology. Our research group has been developing a novel approach to a topology and weight evolving artificial neural network named Mutation-Based Evolving Artificial Neural Network (MBEANN). A series of computer simulations shows that MBEANN yields better results in terms of flexibility than conventional solutions to the cooperative package-pushing problem.

关键词： swarm robotics evolutionary robotics topology and weight evolving artificial neural network (TWEANN) MBEANN

来源：评论

学校读者我要写书评

暂无评论

Evolution of Station Keeping as a Response to Flows in an Aquatic Robot 13

Evolution of Station Keeping as a Response to Flows in an Aq...

引用

15th Genetic and evolutionary Computation Conference (GECCO)

作者： Moore, Jared M. Clark, Anthony J. McKinley, Philip K. Michigan State Univ Dept Comp Sci & Engn E Lansing MI 48824 USA

ISBN: (纸本)9781450319638

Developing complex behaviors for aquatic robots is a difficult engineering challenge due to the uncertainty of an underwater environment. Neuroevolution provides one method of dealing with this type of problem. Artificial neural networks discern different conditions by mapping sensory input to responses, and evolutionary computation provides a training algorithm suitable to the high dimensionality of the problem. In this paper, we present results of applying neuroevolution to an aquatic robot tasked with station keeping, that is, maintaining a given position despite surrounding water flow. The virtual device exposed to evolution is modeled after a physical counterpart that has been fabricated with a 3D printer and tested in physical environments. Evolved behaviors exhibit a variety of unexpected, complex fin/flipper movements that enable the robot to achieve and maintain station, despite water flow from different directions. Moreover, the results show that evolved controllers are able to effectively carry out this task using only information from a simulated accelerometer and gyroscope, matching the inertial measurement unit (IMU) on the actual robot.

关键词： evolutionary robotics Neural Network Application Simulation Station Keeping Aquatic robotics Neuroevolution

来源：评论

学校读者我要写书评

暂无评论

A Coevolutionary Approach to Learn Animal Behavior Through Controlled Interaction 13

A Coevolutionary Approach to Learn Animal Behavior Through C...

引用

15th Genetic and evolutionary Computation Conference (GECCO)

作者： Li, Wei Gauci, Melvin Gross, Roderich Univ Sheffield Dept Automat Control & Syst Engn Sheffield S10 2TN S Yorkshire England

ISBN: (纸本)9781450319638

This paper proposes a method that allows a machine to infer the behavior of an animal in a fully automatic way. In principle, the machine does not need any prior information about the behavior. It is able to modify the environmental conditions and observe the animal;therefore it can learn about the animal through controlled interaction. Using a competitive coevolutionary approach, the machine concurrently evolves animats, that is, models to approximate the animal, as well as classifiers to discriminate between animal and animat. We present a proof-of-concept study conducted in computer simulation that shows the feasibility of the approach. Moreover, we show that the machine learns significantly better through interaction with the animal than through passive observation. We discuss the merits and limitations of the approach and outline potential future directions.

关键词： Science automation animal behavior coevolution Turing test evolutionary robotics interaction artificial life

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：