9,898 Hits in 2.9 sec

Learning modular policies for robotics

Gerhard Neumann, Christian Daniel, Alexandros Paraschos, Andras Kupcsik, Jan Peters
2014 Frontiers in Computational Neuroscience  
In this paper we present our work on unified approach for learning such a modular control architecture.  ...  Finally, we summarize our experiments for learnin modular control architectures in simulation and with real robots. l g , k t a y n w d g  ...  INFORMATION THEORETIC POLICY SEARCH FOR LEARNING MODULAR CONTROL POLICIES In this section we will sequentially introduce our information theoretic policy search framework used for learning modular control  ... 
doi:10.3389/fncom.2014.00062 pmid:24966830 pmcid:PMC4052508 fatcat:thnozh7pofdznlr6hxxhrypfqa

Learning Modular Neural Network Policies for Multi-Task and Multi-Robot Transfer [article]

Coline Devin, Abhishek Gupta, Trevor Darrell, Pieter Abbeel, Sergey Levine
2016 arXiv   pre-print
Reinforcement learning (RL) can automate a wide variety of robotic skills, but learning each new skill requires considerable real-world data collection and manual representation engineering to design policy  ...  Using deep reinforcement learning to train general purpose neural network policies alleviates some of the burden of manual representation engineering by using expressive policy classes, but exacerbates  ...  devise a novel modular approach to policy learning.  ... 
arXiv:1609.07088v1 fatcat:dq57nzd5ajafhbjqcydpdhnlou

Hierarchical Learning for Modular Robots [article]

Risto Kojcev, Nora Etxezarreta, Alejandro Hernández, Víctor Mayoral
2018 arXiv   pre-print
We present a hierarchical approach for modular robots that allows a robot to simultaneously learn multiple tasks.  ...  We argue that hierarchical methods can become the key for modular robots achieving reconfigurability.  ...  The goal of this work is to evaluate the meta-learning shared hierarchies (MLSH) method and its applicability to modular robots.  ... 
arXiv:1802.04132v1 fatcat:oqb5b377sbhkdorrmofbb4u5yi

Learning Modular Robot Control Policies [article]

Julian Whitman, Matthew Travers, Howie Choset
2021 arXiv   pre-print
Modular robots can be rearranged into a new design, perhaps each day, to handle a wide variety of tasks by forming a customized robot for each new task.  ...  We show the modular policy generalizes to a large number of designs that were not seen during training without any additional learning.  ...  For modular robots, developing a new control policy for each design and task is not a scalable solution.  ... 
arXiv:2105.10049v2 fatcat:h4ywzjonrfbsjnooeql3v63fp4

PyRoboLearn: A Python Framework for Robot Learning Practitioners

Brian Delhaisse, Leonel Dario Rozo, Darwin G. Caldwell
2019 Conference on Robot Learning  
On the quest for building autonomous robots, several robot learning frameworks with different functionalities have recently been developed.  ...  Our framework provides a plethora of robotic environments, learning models and algorithms.  ...  We also use PyTorch and Numpy for our learning models and algorithms. PyTorch is chosen because of its Pythonic nature, modularity and popularity in research.  ... 
dblp:conf/corl/DelhaisseRC19 fatcat:7gxssvdlm5dovcx2p5lq5enyau

Adaptable Automation with Modular Deep Reinforcement Learning and Policy Transfer [article]

Zohreh Raziei, Mohsen Moghaddam
2020 arXiv   pre-print
Recent advances in deep Reinforcement Learning (RL) have created unprecedented opportunities for intelligent automation, where a machine can autonomously learn an optimal policy for performing a given  ...  This article tackles this limitation by developing and testing a Hyper-Actor Soft Actor-Critic (HASAC) RL framework based on the notions of task modularization and transfer learning.  ...  In complex industrial applications of robots, adaptability can be achieved by modularizing the learning tasks, learning optimal policies for each individual module, and then transferring learned policies  ... 
arXiv:2012.01934v1 fatcat:ex3i3gfa4jf55huswl5pz5kc2a

MetaMorph: Learning Universal Controllers with Transformers [article]

Agrim Gupta, Linxi Fan, Surya Ganguli, Li Fei-Fei
2022 arXiv   pre-print
In this work, we propose MetaMorph, a Transformer based approach to learn a universal controller over a modular robot design space.  ...  However, modular robot systems now allow for the flexible combination of general-purpose building blocks into task optimized morphologies.  ...  JOINT POLICY OPTIMIZATION The problem of learning a universal controller for a set of K robots drawn from a modular robot design space is a multi-task RL problem.  ... 
arXiv:2203.11931v1 fatcat:lzhqm4sk75fjjb6p3klb5b36dy

ROS2Learn: a reinforcement learning framework for ROS 2 [article]

Yue Leire Erro Nuin, Nestor Gonzalez Lopez, Elias Barba Moral, Lander Usategui San Juan, Alejandro Solano Rueda, Víctor Mayoral Vilches, Risto Kojcev
2019 arXiv   pre-print
We propose a novel framework for Deep Reinforcement Learning (DRL) in modular robotics to train a robot directly from joint states, using traditional robotic tools.  ...  We use an state-of-the-art implementation of the Proximal Policy Optimization, Trust Region Policy Optimization and Actor-Critic Kronecker-Factored Trust Region algorithms to learn policies in four different  ...  Hence, the scalability of previous methods for modular robots is questionable. Modular robots can extend their components seamlessly by just adding modules to the robotic system.  ... 
arXiv:1903.06282v2 fatcat:ylij26vstbdzdn6qukeqxwag3i

Acquisition of Competitive Behaviors in Multi-Agent System Based on a Modular Learning System

Yasutake Takahashi, Kazuhiro Edazawa, Kentaro Noma, Minoru Asada
2009 Journal of the Robotics Society of Japan  
constancy is needed for behavior learning to converge.  ...  Scheduling for learning is introduced to avoid the complexity in autonomous situation assignment.  ...  Hosoda: "Pur- posive behavior acquisition for a real robot by vision-based re- inforcement learning," Machine Learning, vol.23, pp.279-303, 1996. [ 3  ... 
doi:10.7210/jrsj.27.350 fatcat:hc6qnm45arcaxihlmqyfivhlji

Multi-Task Reinforcement Learning with Soft Modularization [article]

Ruihan Yang, Huazhe Xu, Yi Wu, Xiaolong Wang
2020 arXiv   pre-print
Instead of directly selecting routes for each task, our task-specific policy uses a method called soft modularization to softly combine all the possible routes, which makes it suitable for sequential tasks  ...  Multi-task learning is a very challenging problem in reinforcement learning.  ...  Figure 1 : 1 We design a multi-task policy network with soft modularization for robotics manipulation.  ... 
arXiv:2003.13661v2 fatcat:6j4ghklahzfddi6qirwcy3zv6u

Automated Design of Adaptive Controllers for Modular Robots using Reinforcement Learning

Paulina Varshavskaya, Leslie Pack Kaelbling, Daniela Rus
2008 The international journal of robotics research  
Designing distributed controllers for self-reconfiguring modular robots has been consistently challenging.  ...  modular robots.  ...  of a modular robot as a multi-agent POMDP.  ... 
doi:10.1177/0278364907084983 fatcat:7g4cra6lhvgnzmixeskviznski

Adaptive Modular Reinforcement Learning for Robot Controlled in Multiple Environments

Teppei Iwata, Takeshi Shibuya
2021 IEEE Access  
This paper proposes an adaptive modular reinforcement learning architecture and an algorithm for robot control operating in multiple environments.  ...  Consequently, reinforcement learning is expected to be applied to robot control where model identification is difficult. These robots are often expected to operate in multiple environments.  ...  ACKNOWLEDGMENT We would like to thank Editage ( for English language editing.  ... 
doi:10.1109/access.2021.3070704 fatcat:ebk3hbljg5hb3l47s4pdz6fpvm

Evaluation of Deep Reinforcement Learning Methods for Modular Robots [article]

Risto Kojcev, Nora Etxezarreta, Alejandro Hernández, Víctor Mayoral
2018 arXiv   pre-print
We propose a novel framework for Deep Reinforcement Learning (DRL) in modular robotics using traditional robotic tools that extend state-of-the-art DRL implementations and provide an end-to-end approach  ...  Our results show that extending the modular robot from 3 degrees-of-freedom (DoF), to 4 DoF, does not affect the robot's learning.  ...  Thus, the scalability of previous methods for modular robots is questionable. Modular robots can extend their components seamlessly.  ... 
arXiv:1802.02395v1 fatcat:7pkt6w5gabbylnmli2hlypnbea

Online Gait Learning for Modular Robots with Arbitrary Shapes and Sizes [chapter]

Massimiliano D'Angelo, Berend Weel, A. E. Eiben
2013 Lecture Notes in Computer Science  
To evolve robot morphologies and controllers in real-space and real-time we need a generic learning mechanism that enables arbitrary modular shapes to obtain a suitable gait quickly after 'birth'.  ...  The experiments give insights into the online dynamics of gait learning, the distribution of lucky / unlucky runs and their dependence on the size and complexity of the modular robotic organisms.  ...  Related Work The design of locomotion for modular robotics is a difficult problem.  ... 
doi:10.1007/978-3-642-45008-2_4 fatcat:mcslqfmsrfhwrdwzccip5o3mz4

Robot_gym: accelerated robot training through simulation in the cloud with ROS and Gazebo [article]

Víctor Mayoral Vilches, Alejandro Hernández Cordero, Asier Bilbao Calvo, Irati Zamalloa Ugarte, Risto Kojcev
2018 arXiv   pre-print
We unveil that, for simple tasks, simple 3DoF robots require more than 140 attempts to learn. For more complex, 6DoF robots, the number of attempts increases to more than 900 for the same task.  ...  However, such training requires a big amount of experimentation which is not always feasible for a physical robot.  ...  However, the largest bottleneck for applying RL techniques to real robot systems is the amount of training time and hand-crafted policy for the robot to learn certain behaviour.  ... 
arXiv:1808.10369v1 fatcat:qheoetn35nb63njmh5gi6tqwfa
« Previous Showing results 1 — 15 out of 9,898 results