Parallel Gym Gazebo: a Scalable Parallel Robot Deep Reinforcement Learning Platform

2019 ◽  
Author(s):  
Zhen Liang ◽  
Zhongxuan Cai ◽  
Minglong Li ◽  
Wenjing Yang
Keyword(s):  
Reinforcement Learning ◽  
Parallel Robot ◽  
Learning Platform

Vibration Suppression for Large-Scale Flexible Structures Using Deep Reinforcement Learning Based on Cable-Driven Parallel Robots

2020 ◽  
Author(s):  
Haining Sun ◽  
Xiaoqiang Tang ◽  
Jinhao Wei
Keyword(s):  
Reinforcement Learning ◽  
Large Scale ◽  
Vibration Suppression ◽  
External Disturbance ◽  
Flexible Structures ◽  
Parallel Robot ◽  
Parallel Robots ◽  
Gradient Algorithm ◽  
Normal Operation ◽  
Control Scheme

Abstract Specific satellites with ultra-long wings play a crucial role in many fields. However, external disturbance and self-rotation could result in undesired vibrations of flexible wings, which affects the normal operation of the satellites. In severe cases, the satellites will be damaged. Therefore, it is imperative to conduct vibration suppression for these flexible structures. Utilizing deep reinforcement learning (DRL), an active control scheme is presented in this paper to rapidly suppress the vibration of flexible structures with quite small controllable force based on a cable-driven parallel robot (CDPR). To verify the controller’s effectiveness, three groups of simulation with different initial disturbance are implemented. Besides, to enhance the contrast, a passive pre-tightening scheme is also tested. First, the dynamic model of the CDPR that is comprised of four cables and a flexible structure is established using the finite element method. Then, the dynamic behavior of the model under the controllable cable force is analyzed by Newmark-ß method. Furthermore, the agent of DRL is trained by the deep deterministic policy gradient algorithm (DDPG). Finally, the control scheme is conducted on Simulink environment to evaluate its performance, and the results are satisfactory, which validates the controller’s ability to suppress vibrations.

A Reinforcement Learning Platform for Small and Medium-sized Enterprises in Logistics

2021 ◽  
Author(s):  
Jean Paul Sebastian Piest ◽  
Maria-Eugenia Iacob ◽  
Marten van Sinderen ◽  
Martijn Gemmink ◽  
Bjorn Goossens
Keyword(s):  
Reinforcement Learning ◽  
Learning Platform

Development of an Actor-Critic Deep Reinforcement Learning Platform for Robotic Grasping in Real World

2020 ◽  
Vol 15 (2) ◽  
pp. 197-204
Author(s):  
Taewon Kim ◽  
◽  
Yeseong Park ◽  
Jong Bok Kim ◽  
Youngbin Park ◽  
...  
Keyword(s):  
Reinforcement Learning ◽  
Real World ◽  
Robotic Grasping ◽  
Learning Platform

scholarly journals Trajectory Planning of a CableBased Parallel Robot using Reinforcement Learning and Soft Actor-Critic

2020 ◽  
Vol 15 ◽  
Keyword(s):  
Artificial Intelligence ◽  
Reinforcement Learning ◽  
Parallel Robot ◽  
Parallel Robots ◽  
Reward Function ◽  
Serial Robots ◽  
Manufacturing Efficiency ◽  
Serial Robot ◽  
Active Research ◽  
Set Up

Industry 4.0 introduces the use of modular stations and better communication between agents to improve manufacturing efficiency and to lower the downtime between the customer and its final product. Among novel mechanisms that have a high potential in this new industrial paradigm are cable­suspended parallel robot (CSPR): their payload­to­mass ratio is high compared to their serial robot counterpart and their setup is quick compared to other types of parallel robots such as Gantry system, popular in the automotive industry but difficult to set up and to adapt while the production line changes. A CSPR can cover the workspace of a manufacturing hall and providing assistance to operators before they arrive at their workstation. One challenge is to generate the desired trajectories, so that the CSPR could move to the desired area. Reinforcement Learning (RL) is a branch of Artificial Intelligence where the agent interacts with an environment to maximize a reward function. This paper proposes the use of a RL algorithm called Soft Actor­Critic (SAC) to train a two degrees­of­freedom (DOFs) CSPR to perform pick­and­place trajectory. Even though the pick­and­place trajectory based on artificial intelligence has been an active research with serial robots, this technique has yet to be applied to parallel robots.

scholarly journals Multi-Task Deep Reinforcement Learning with PopArt

2019 ◽  
Vol 33 ◽  
pp. 3796-3803 ◽  
Author(s):  
Matteo Hessel ◽  
Hubert Soyer ◽  
Lasse Espeholt ◽  
Wojciech Czarnecki ◽  
Simon Schmitt ◽  
...  
Keyword(s):  
Reinforcement Learning ◽  
Human Performance ◽  
Learning Algorithm ◽  
State Of The Art ◽  
Single Agent ◽  
Learning System ◽  
Learning Platform ◽  
Art Performance ◽  
The One ◽  
First Time

The reinforcement learning (RL) community has made great strides in designing algorithms capable of exceeding human performance on specific tasks. These algorithms are mostly trained one task at the time, each new task requiring to train a brand new agent instance. This means the learning algorithm is general, but each solution is not; each agent can only solve the one task it was trained on. In this work, we study the problem of learning to master not one but multiple sequentialdecision tasks at once. A general issue in multi-task learning is that a balance must be found between the needs of multiple tasks competing for the limited resources of a single learning system. Many learning algorithms can get distracted by certain tasks in the set of tasks to solve. Such tasks appear more salient to the learning process, for instance because of the density or magnitude of the in-task rewards. This causes the algorithm to focus on those salient tasks at the expense of generality. We propose to automatically adapt the contribution of each task to the agent’s updates, so that all tasks have a similar impact on the learning dynamics. This resulted in state of the art performance on learning to play all games in a set of 57 diverse Atari games. Excitingly, our method learned a single trained policy - with a single set of weights - that exceeds median human performance. To our knowledge, this was the first time a single agent surpassed human-level performance on this multi-task domain. The same approach also demonstrated state of the art performance on a set of 30 tasks in the 3D reinforcement learning platform DeepMind Lab.

scholarly journals MRDRL-ROS: a Multi Robot Deep Reinforcement Learning Platform based on Robot Operating System

2021 ◽  
Vol 2113 (1) ◽  
pp. 012086
Author(s):  
Zhongxuan Cai ◽  
Zhen Liang ◽  
Jing Ren
Keyword(s):  
Operating System ◽  
Reinforcement Learning ◽  
Software Engineering ◽  
Learning Environment ◽  
Building Blocks ◽  
Experimental Results ◽  
Robot Operating System ◽  
Engineering Approach ◽  
Learning Platform ◽  
Diverse Environment

Abstract Deep reinforcement learning (DRL) has greatly improved the intelligence of AI in recent years and the community has proposed several common software to facilitate the development of DRL. However, in robotics the utility of common DRL software is limited and the development is time-consuming due to the complexity of various robot software. In this paper, we propose a software engineering approach leveraging modularity to facilitate robot DRL development. The platform decouples learning environment into task, simulator and hierarchical robot modules, which in turn enables diverse environment generation using existing modules as building blocks, regardless of the underlying robot software details. Experimental results show that our platform provides composable environment building, introduces high module reuse and efficiently facilitates robot DRL.

Sign in / Sign up

Export Citation Format

Share Document