Self-Configuring Robot Path Planning With Obstacle Avoidance via Deep Reinforcement Learning

Sangiovanni, Bianca; Incremona, Gian Paolo; Piastra, Marco; Ferrara, Antonella

doi:10.1109/lcsys.2020.3002852

Cited by 73 publications

(45 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After the simulation, the displacement, velocity, acceleration and force curve of the driving joint of each unit parallel mechanism module can be obtained as outputs. The simulation results are shown in Figures 8,9,10,11,12,13. where x, y, z, φ x , φ y , and φ z denote the position and orientation of the distal moving platform of the HRETR. By observing the kinematic diagram of Figures 8,9,10,11,12,13 the kinematic law of each limb meets the preset requirements (both the initial and terminal velocity and the acceleration are 0).…”

Section: Dynamic Analysis Of the Hretrmentioning

confidence: 99%

“…With the introduction of the concept of Industry 4.0, robots have become more and more essential in intelligent manufacturing [1]. Despite the wide application of robots in the industry [2][3][4][5], medical sector [6][7][8][9][10], aerospace field [11,12], and other fields, consumers' needs are becoming increasingly dynamic and complex [13]. This challenge can be resolved with better motion flexibility, stiffness and load capacity of robots.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Hyper-redundant Elephant’s Trunk Robot with an Open Structure: Design, Kinematics, Control and Prototype

Zhao

Song

Zhang

et al. 2020

Chin. J. Mech. Eng.

View full text Add to dashboard Cite

As for the complex operational tasks in the unstructured environment with narrow workspace and numerous obstacles, the traditional robots cannot accomplish these mentioned complex operational tasks and meet the dexterity demands. The hyper-redundant bionic robots can complete complex tasks in the unstructured environments by simulating the motion characteristics of the elephant’s trunk and octopus tentacles. Compared with traditional robots, the hyper-redundant bionic robots can accomplish complex tasks because of their flexible structure. A hyper-redundant elephant’s trunk robot (HRETR) with an open structure is developed in this paper. The content includes mechanical structure design, kinematic analysis, virtual prototype simulation, control system design, and prototype building. This design is inspired by the flexible motion of an elephant’s trunk, which is expansible and is composed of six unit modules, namely, 3UPS-PS parallel in series. First, the mechanical design of the HRETR is completed according to the motion characteristics of an elephant’s trunk and based on the principle of mechanical bionic design. After that, the backbone mode method is used to establish the kinematic model of the robot. The simulation software SolidWorks and ADAMS are combined to analyze the kinematic characteristics when the trajectory of the end moving platform of the robot is assigned. With the help of ANSYS, the static stiffness of each component and the whole robot is analyzed. On this basis, the materials of the weak parts of the mechanical structure and the hardware are selected reasonably. Next, the extensible structures of software and hardware control system are constructed according to the modular and hierarchical design criteria. Finally, the prototype is built and its performance is tested. The proposed research provides a method for the design and development for the hyper-redundant bionic robot.

show abstract

Section: Dynamic Analysis Of the Hretrmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Hyper-redundant Elephant’s Trunk Robot with an Open Structure: Design, Kinematics, Control and Prototype

Zhao

Song

Zhang

et al. 2020

Chin. J. Mech. Eng.

View full text Add to dashboard Cite

show abstract

“…In [32] a dynamic obstacle avoidance approach for robot manipulator base on a continuous Deep-RL algorithm [33] was proposed and a well-designed reward function was presented to avoid the sparse reward and increase useful exploitation. In Sangiovanni et al [34], presented a hybrid control method that combines the Deep-RL policy and the PRM policy (i.e. a sampling-based method) to achieve collision-free path planning for anthropomorphic robot manipulators.…”

Section: Introductionmentioning

confidence: 99%

Collision-free path planning for welding manipulator via hybrid algorithm of deep reinforcement learning and inverse kinematics

Zhong

Wang

Cheng

2021

Complex Intell. Syst.

View full text Add to dashboard Cite

In actual welding scenarios, an effective path planner is needed to find a collision-free path in the configuration space for the welding manipulator with obstacles around. However, as a state-of-the-art method, the sampling-based planner only satisfies the probability completeness and its computational complexity is sensitive with state dimension. In this paper, we propose a path planner for welding manipulators based on deep reinforcement learning for solving path planning problems in high-dimensional continuous state and action spaces. Compared with the sampling-based method, it is more robust and is less sensitive with state dimension. In detail, to improve the learning efficiency, we introduce the inverse kinematics module to provide prior knowledge while a gain module is also designed to avoid the local optimal policy, we integrate them into the training algorithm. To evaluate our proposed planning algorithm in multiple dimensions, we conducted multiple sets of path planning experiments for welding manipulators. The results show that our method not only improves the convergence performance but also is superior in terms of optimality and robustness of planning compared with most other planning algorithms.

show abstract

“…Between the tested approaches, the authors concluded that reusing experience improves performance greatly. Sangiovanni et al (2020) further extends the work in Sangiovanni et al (2018) by using a hybrid approach to the obstacle avoidance task, using conventional motion planning techniques and DRL. The idea is that if a metric evaluating the risk of collision is below a certain threshold, the system gives the control of the manipulator to a DRL system, which has the task of avoiding collisions.…”

Section: The Reward Functionmentioning

confidence: 89%

“…A video showing their experiments can be seen at <https://youtu.be/Zqc6sm76ZCg>. Sangiovanni et al (2018) proposes a system using DRL to perform motion control of a robot arm with obstacle avoidance. Their goal was to design a model that could be used in physical Human-Robot Interaction (pHRI), where the robot must avoid collisions with humans in its workspace.…”

Section: The Reward Functionmentioning

confidence: 99%

End-to-End Visual Obstacle Avoidance for a Robotic Manipulator using Deep Reinforcement Learning

Sanches¹

View full text Add to dashboard Cite

I also want to thank my laboratory partners, especially José Pedro and Iury Batista, for their friendship and support, even during this very difficult time of isolation. Our daily conversations brought a bit of the laboratory into my home and made my days more enjoyable.Additionally, I want to thank the friends I made during my journey at the ICMC, having fond memories that I will carry with me forever. Likewise, my childhood friends, whose bonds of friendship endure despite the time.And last but not least, I want to thank all the research funding agencies, specially CNPq, CAPES, and FAPESP, for funding this research and many others, essential for the development of science in Brazil."I would rather have questions that can't be answered than answers that can't be questioned."(Richard Feynman

show abstract

Self-Configuring Robot Path Planning With Obstacle Avoidance via Deep Reinforcement Learning

Cited by 73 publications

References 28 publications

A Hyper-redundant Elephant’s Trunk Robot with an Open Structure: Design, Kinematics, Control and Prototype

A Hyper-redundant Elephant’s Trunk Robot with an Open Structure: Design, Kinematics, Control and Prototype

Collision-free path planning for welding manipulator via hybrid algorithm of deep reinforcement learning and inverse kinematics

End-to-End Visual Obstacle Avoidance for a Robotic Manipulator using Deep Reinforcement Learning

Contact Info

Product

Resources

About