Filters








13,007 Hits in 5.3 sec

Smooth Exploration for Robotic Reinforcement Learning [article]

Antonin Raffin, Jens Kober, Freek Stulp
2021 arXiv   pre-print
Reinforcement learning (RL) enables robots to learn skills from interactions with the real world.  ...  The noise sampling interval of gSDE permits to have a compromise between performance and smoothness, which allows training directly on the real robots without loss of performance.  ...  When learning robotic skills with deep reinforcement learning (Deep RL), the de facto standard for exploration is to sample a noise vector t from a Gaussian distribution independently at each time step  ... 
arXiv:2005.05719v2 fatcat:zrh4yt3edfbynecj2imvp3hqou

A fuzzy controller with supervised learning assisted reinforcement learning algorithm for obstacle avoidance

Cang Ye, N.H.C. Yung, Danwei Wang
2003 IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics)  
For sufficient learning, a new learning method using modified Sutton and Barto's model is proposed to strengthen the exploration.  ...  After sufficient training, fine learning is applied which employs reinforcement learning algorithm to fine-tune the membership functions for the output variables.  ...  For sufficient learning, a new learning method using modified Sutton and Barto's model is proposed to strengthen the exploration.  ... 
doi:10.1109/tsmcb.2003.808179 pmid:18238153 fatcat:eg3ycz6i7jftfhd7sj3xjhfx6m

Constrained-Space Optimization and Reinforcement Learning for Complex Tasks

Ya-Yen Tsai, Bo Xiao, Edward Johns, Guang-Zhong Yang
2020 IEEE Robotics and Automation Letters  
Learning from Demonstration is increasingly used for transferring operator manipulation skills to robots.  ...  This paper presents a constrained-space optimization and reinforcement learning scheme for managing complex tasks.  ...  Index Terms-Medical robotics, Learn from Demonstration (LfD), Reinforcement Learning (RL), Robot learning, Robotic suturing. I.  ... 
doi:10.1109/lra.2020.2965392 fatcat:4upr4a6gdreytgdiqbirnjy4fi

Reinforcement Learning Neural Network to the Problem of Autonomous Mobile Robot Obstacle Avoidance

Bing-Qiang Huang, Guang-Yi Cao, Min Guo
2005 2005 International Conference on Machine Learning and Cybernetics  
An approach to the problem of autonomous mobile robot obstacle avoidance using reinforcement learning neural network is proposed in this paper.  ...  The simulation results show that the simulated robot using the reinforcement learning neural network can enhance its learning ability obviously and can finish the given task in a complex environment.  ...  Online learning and adaptation are desirable traits for any robot learning algorithm operating in changing and unstructured environments where the robot explores its environment to collect sufficient samples  ... 
doi:10.1109/icmlc.2005.1526924 fatcat:xgefoyokargt7b3htnsf3trl6y

Exploration in structured space of robot movements for autonomous augmentation of action knowledge

Denis Forte, Bojan Nemec, Ales Ude
2015 2015 International Conference on Advanced Robotics (ICAR)  
Imitation learning has been proposed as the basis for fast and efficient acquisition of new sensorimotor behaviors.  ...  Efficient exploration becomes possible by exploiting the structure of the search space defined by the previously acquired example movements.  ...  Based on these early works, reinforcement learning of robot movements has started to be seen as a viable approach to motion learning in robotics [3] - [7] , even in the case of high degree freedom humanoid  ... 
doi:10.1109/icar.2015.7251464 dblp:conf/icar/ForteNU15 fatcat:4fznifokefegpda3cl6ism2aiu

Effect of human guidance and state space size on Interactive Reinforcement Learning

Halit Bener Suay, Sonia Chernova
2011 2011 RO-MAN  
We present the first study of Interactive Reinforcement Learning in realworld robotic systems.  ...  The Interactive Reinforcement Learning algorithm enables a human user to train a robot by providing rewards in response to past actions and anticipatory guidance to guide the selection of future actions  ...  REINFORCEMENT LEARNING Reinforcement learning algorithms enable a robot to learn from its experience.  ... 
doi:10.1109/roman.2011.6005223 dblp:conf/ro-man/SuayC11 fatcat:numbqfpuyjd4hoxshtckfasdtq

Using Approximate Models in Robot Learning [article]

Ali Lenjani
2019 arXiv   pre-print
For solving this type of problem, if we have an appropriate reward function and dynamics model; finding an optimal control policy is possible by using model-based reinforcement learning and optimal control  ...  The problem has significant difficulties including a large number of trials required for data collection and a massive volume of computations required to find a closed-loop controller for high dimensional  ...  Consequently, it is vital to use exploration methods that maximize the knowledge expanded during learning and minimize the costs of exploration and time for learning.  ... 
arXiv:1902.04696v1 fatcat:aqlrofh6sbgehktekmau4zaqpe

2018年度日本神経回路学会学術賞受賞者のことば/論文賞の研究概要/最優秀研究賞の研究概要/優秀研究賞の研究概要

2019 The Brain & Neural Networks  
Our motivation stems from this challenge, so we aim to explore practical model-free reinforcement learning algorithms in complex robot systems to fulfill stable learning with insufficient real world samples  ...  With the capability of exploring unknown worlds, reinforcement learning has become a popular approach that expresses a remarkably broad range of robot control problems in a natural manner.  ... 
doi:10.3902/jnns.26.27 fatcat:dcfqvfwc5ze2faikazeue2ksly

The Path Planning of Mobile Robot by Neural Networks and Hierarchical Reinforcement Learning

Jinglun Yu, Yuancheng Su, Yifan Liao
2020 Frontiers in Neurorobotics  
Deep Deterministic Policy Gradient (DDPG), a path planning algorithm for mobile robots based on neural networks and hierarchical reinforcement learning, performed better in all aspects than other algorithms  ...  By mapping the current state of these actions through Hierarchical Reinforcement Learning (HRL), the needs of mobile robots are met.  ...  An Agent uses reinforcement learning methods to learn, which is to acquire knowledge from a sequence of actions obtained by exploration.  ... 
doi:10.3389/fnbot.2020.00063 pmid:33132890 pmcid:PMC7561669 fatcat:nidntdemf5ddhow4mzd3h3k3bm

Self-Organisation of Generic Policies in Reinforcement Learning

Simón Smith, J. Michael Herrmann
2013 Advances in Artificial Life, ECAL 2013  
The approach is illustrated by a learning tasks in a six-legged robot.  ...  We propose the use of an exploratory self-organised policy to initialise the parameters of the function approximation in the reinforcement learning policy based on the value function of the exploratory  ...  As the main contribution of this study, we propose a combination of homeokinetic and reinforcement learning which uses for high-dimensional reinforcement learning tasks a combination of autonomous exploration  ... 
doi:10.7551/978-0-262-31709-2-ch091 dblp:conf/ecal/SmithH13 fatcat:ac3c6atjpfcp5pya3fvyru2vei

Learning manipulation skills from a single demonstration

Peter Englert, Marc Toussaint
2017 The international journal of robotics research  
We consider the scenario where a robot is demonstrated a manipulation skill once and should then use only few own trials to learn to reproduce, optimize and generalize that same skill.  ...  The algorithm is evaluated on a synthetic benchmark experiment and compared to state-of-the-art learning methods. We also demonstrate the performance on real robot experiments with a PR2.  ...  We then use a combination of optimal control and reinforcement learning to improve the skill with respect to smoothness and force efficiency.  ... 
doi:10.1177/0278364917743795 fatcat:slnvwjgfdzdv3ksjbwz7ys3haa

Sensor-Based Navigation Using Hierarchical Reinforcement Learning [article]

Christopher Gebauer, Nils Dengler, Maren Bennewitz
2022 arXiv   pre-print
This makes deep reinforcement learning (DRL) especially interesting, as these algorithms promise a self-learning system only relying on feedback from the environment.  ...  Robotic systems are nowadays capable of solving complex navigation tasks.  ...  The authors used a model-based reinforcement learning approach to find an optimal meeting point for two mobile robots without centralized communication.  ... 
arXiv:2108.13268v2 fatcat:roe2qpv5jbhbzgmqdtbjtvff64

Bézier Curve Based Continuous and Smooth Motion Planning for Self-Learning Industrial Robots

Christian Scheiderer, Timo Thun, Tobias Meisen
2019 Procedia Manufacturing  
A prominent example is motion planning for industrial robotics, where reinforcement-learning solutions to date result in non-fluent trajectories.  ...  A prominent example is motion planning for industrial robotics, where reinforcement-learning solutions to date result in non-fluent trajectories.  ...  Acknowledgements The authors would like to thank the German Research Foundation DFG for the kind support within the Cluster of Excellence Internet of Production (IoP). Project-ID: 390621612  ... 
doi:10.1016/j.promfg.2020.01.054 fatcat:pct5hl6yjng4fhhn2mjsglquku

Learning a DFT-based sequence with reinforcement learning: a NAO implementation

Boris Durán, Gauss Lee, Robert Lowe
2012 Paladyn: Journal of Behavioral Robotics  
AbstractThe implementation of sequence learning in robotic platforms offers several challenges.  ...  Results from the comparison of two reinforcement learning methods applied to sequence generation, for both simulation and implementation, are provided.  ...  in an 'online' manner Reinforcement learning is a paradigm that has been employed for both autonomous, nonsupervised learning in robots and for learning sequences of states or state-action pairs (cf  ... 
doi:10.2478/s13230-013-0109-5 fatcat:7rx34qn2hvdu3cicxxuuazsima

Improved Path Planning for Indoor Patrol Robot Based on Deep Reinforcement Learning

Jianfeng Zheng, Shuren Mao, Zhenyu Wu, Pengcheng Kong, Hao Qiang
2022 Symmetry  
To solve the problems of poor exploration ability and convergence speed of traditional deep reinforcement learning in the navigation task of the patrol robot under indoor specified routes, an improved  ...  deep reinforcement learning algorithm based on Pan/Tilt/Zoom(PTZ) image information was proposed in this paper.  ...  While reinforcement learning is responsible for the patrol robot to explore and analyze the acquired environmental information, which helps the patrol robot to make correct decisions and makes the patrol  ... 
doi:10.3390/sym14010132 fatcat:exlqmscogjeoxefod7w5hie7qa
« Previous Showing results 1 — 15 out of 13,007 results