Survey of imitation learning for robotic manipulation

Fang, Bin; Jia, Shidong; Guo, Di; Xu, Muhua; Wen, Shuhuan; Sun, Fuchun

doi:10.1007/s41315-019-00103-5

Cited by 111 publications

(47 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…With the advancement of hardware and machine learning technology, data-driven methods that can reduce the complexity of the computation without listing all possibilities are widely used in robot manipulation [ 11 ]. The ability of perception and understanding are improved via feature recognition and classification, then the probability model of manipulation can be learned to perform the task [ 19 ]. Nowadays, there are two main solutions for data-driven methods.…”

Section: The Backgroundmentioning

confidence: 99%

Learning for a Robot: Deep Reinforcement Learning, Imitation Learning, Transfer Learning

Jiang

Zeng

et al. 2021

Sensors

135

View full text Add to dashboard Cite

Dexterous manipulation of the robot is an important part of realizing intelligence, but manipulators can only perform simple tasks such as sorting and packing in a structured environment. In view of the existing problem, this paper presents a state-of-the-art survey on an intelligent robot with the capability of autonomous deciding and learning. The paper first reviews the main achievements and research of the robot, which were mainly based on the breakthrough of automatic control and hardware in mechanics. With the evolution of artificial intelligence, many pieces of research have made further progresses in adaptive and robust control. The survey reveals that the latest research in deep learning and reinforcement learning has paved the way for highly complex tasks to be performed by robots. Furthermore, deep reinforcement learning, imitation learning, and transfer learning in robot control are discussed in detail. Finally, major achievements based on these methods are summarized and analyzed thoroughly, and future research challenges are proposed.

show abstract

Section: The Backgroundmentioning

confidence: 99%

Learning for a Robot: Deep Reinforcement Learning, Imitation Learning, Transfer Learning

Jiang

Zeng

et al. 2021

Sensors

135

View full text Add to dashboard Cite

show abstract

“…There are two kinds of contour error, namely tracking error that measures how far the current tool position away from the expected one, and absolute error that represents the minimal distance between the current tool position and the pre-defined trajectory. With an optical MoCap one can perform assembly demonstration with the robot (direct demonstration) or without the attendance of any robot (indirect demonstration) [28]. For 3C product assembly applications, no matter which LfD method is selected to learn the assembly skill, it is more efficient to carry out the indirect demonstration.…”

Section: ) Trajectory Performance Evaluationmentioning

confidence: 99%

Performance Evaluation of Optical Motion Capture Sensors for Assembly Motion Capturing

Cao

Yang

et al. 2021

IEEE Access

View full text Add to dashboard Cite

The optical motion capture (MoCap) sensor provides an effective way to capture human motions and transform them into valuable data that can be applied to certain tasks, e.g. robot learning from demonstration (LfD). In spite of the wide utilization of optical MoCaps in LfD studies, there are few works that explore their potentiality in small parts robotic assembly. Robot manipulation skill learning from demonstration has gained the attention of researchers recently and robotic 3C (Computer, Communication, and Consumer Electronics) product assembly turns out to be a promising application thanks to the increasing consumption of 3C products. To further explore the potential of optical MoCaps in robotic 3C product assembly. This work proposes a performance evaluation protocol that takes the characters of both optical MoCaps and 3C product assembly operations into account. The proposed evaluation protocol includes static and trajectory evaluations. The former refers to the widely used evaluation indicators such as precision and accuracy. Meanwhile, the trajectory evaluation takes contour error as an error metric. Three popular optical MoCaps are studied in the experiment. Experiment results show that the static performance of all of the three optical MoCaps can meet the requirements of the 3C product assembly task. What's more, Prime X41 possesses the best trajectory performance. This work sheds light on the wider usage of optical MoCaps in manufacturing industries.

show abstract

“…Using raw demonstration is also important in hastening the process where learning samples are immediately acquired for the robot so that the robot can perform the action precisely. However, this type of robotic learning is appropriate only for cases with direct contact and a passive controller [195] [196]. Thus, it is not suitable for dealing with multiple DoFs (e.g.…”

Section: Robots Learning From Demonstrationmentioning

confidence: 99%

Review of Deep Reinforcement Learning-Based Object Grasping: Techniques, Open Challenges, and Recommendations

2020

View full text Add to dashboard Cite

The motivation behind our work is to review and analyze the most relevant studies on deep reinforcement learning-based object manipulation. Various studies are examined through a survey of existing literature and investigation of various aspects, namely, the intended applications, techniques applied, challenges faced by researchers and recommendations for minimizing obstacles. This review refers to all relevant articles on deep reinforcement learning-based object manipulation and solutions. The object grasping issue is a major manipulation challenge. Object grasping requires detection systems, methods and tools to facilitate efficient and fast agent training. Several studies have proposed that object grasping and its subtypes are the main elements in dealing with the environment and agent. Unlike other review articles, this review article provides different observations on deep reinforcement learning-based manipulation. The results of this comprehensive review of deep reinforcement learning in the manipulation field may be valuable for researchers and practitioners because they can expedite the establishment of important guidelines.

show abstract

Survey of imitation learning for robotic manipulation

Cited by 111 publications

References 31 publications

Learning for a Robot: Deep Reinforcement Learning, Imitation Learning, Transfer Learning

Learning for a Robot: Deep Reinforcement Learning, Imitation Learning, Transfer Learning

Performance Evaluation of Optical Motion Capture Sensors for Assembly Motion Capturing

Review of Deep Reinforcement Learning-Based Object Grasping: Techniques, Open Challenges, and Recommendations

Contact Info

Product

Resources

About