Solving peg-in-hole tasks by human demonstration and exception strategies

Abu‐Dakka, Fares J.; Nemec, Bojan; Kramberger, Aljaž; Buch, Anders Glent; Krüger, Norbert; Ude, Aleš

doi:10.1108/ir-07-2014-0363

Cited by 59 publications

(48 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The reward functions include the terms of the force, moment, z-axis robot position, xyz robot rotation, and time step. Meta-parameters α, β, and γ control the weights of the terms individually because they deal with different information types (see Equations (8) and (9)). We can change the importance of each term in the reward function by adjusting these meta-parameters.…”

Section: Discussionmentioning

confidence: 99%

“…We attempt to obtain these optimal solutions with fewer trials-and-errors by providing near-optimal solutions learned from human demonstrations. In this paper, robots improve motor skills to optimize them-referred to as improvement-and generalize them so that they are widely applicable-known as generalization-through self-learning.The peg-in-hole task has also been addressed through several imitation learning studies [8,9]. However, the peg-in-hole task is not easy to learn with this method alone.…”

mentioning

confidence: 99%

“…The peg-in-hole task has also been addressed through several imitation learning studies [8,9]. However, the peg-in-hole task is not easy to learn with this method alone.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Learning, Improving, and Generalizing Motor Skills for the Peg-in-Hole Tasks Based on Imitation Learning and Self-Learning

et al. 2020

View full text Add to dashboard Cite

We propose a framework based on imitation learning and self-learning to enable robots to learn, improve, and generalize motor skills. The peg-in-hole task is important in manufacturing assembly work. Two motor skills for the peg-in-hole task are targeted: "hole search" and "peg insertion". The robots learn initial motor skills from human demonstrations and then improve and/or generalize them through reinforcement learning (RL). An initial motor skill is represented as a concatenation of the parameters of a hidden Markov model (HMM) and a dynamic movement primitive (DMP) to classify input signals and generate motion trajectories. Reactions are classified as familiar or unfamiliar (i.e., modeled or not modeled), and initial motor skills are improved to solve familiar reactions and generalized to solve unfamiliar reactions. The proposed framework includes processes, algorithms, and reward functions that can be used for various motor skill types. To evaluate our framework, the motor skills were performed using an actual robotic arm and two reward functions for RL. To verify the learning and improving/generalizing processes, we successfully applied our framework to different shapes of pegs and holes. Moreover, the execution time steps and path optimization of RL were evaluated experimentally. of motor skills. However, for acquiring complete motor skills, it has one evident limitation: it does not ensure that robots acquire motor skills that are optimized for their goals (that is, it generally provides near-optimal solutions) [5]. Furthermore, it is not easy for human performers to provide a demonstration dataset that can cover all situations arising during the execution of a motor skill [6]. Nonetheless, these human demonstrations can be used as a solid starting point for robots to acquire motor skills [7]. To obtain optimal motor skills, robots must be able to improve motor skills through self-learning. However, this self-learning is a time-consuming and expensive process in the absence of references. We attempt to obtain these optimal solutions with fewer trials-and-errors by providing near-optimal solutions learned from human demonstrations. In this paper, robots improve motor skills to optimize them-referred to as improvement-and generalize them so that they are widely applicable-known as generalization-through self-learning.The peg-in-hole task has also been addressed through several imitation learning studies [8,9]. However, the peg-in-hole task is not easy to learn with this method alone. The main reason is that it is difficult for human performers to provide a complete demonstration dataset to robots, because it is not feasible to prepare all possible reaction situations. In addition, unintended reaction information may be included in the dataset during the demonstration process. These problems may prevent robots from acquiring the complete motor skills. Thus, initial motor skills are learned to classify reaction force/moment signals and generate reaction motion trajectories from human demonstrations, and thei...

show abstract

Section: Discussionmentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Learning, Improving, and Generalizing Motor Skills for the Peg-in-Hole Tasks Based on Imitation Learning and Self-Learning

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Inverse reinforcement learning (IRL), on the other hand, estimates a reward function based on the expert's demonstration [17,18], then utilizes the learned reward function to obtain the policy. Existing demonstration methods include visual demonstration [19][20][21][22][23], force demonstration [24][25][26], visual and force demonstration [27,28], and trajectory demonstration [29][30][31][32][33][34][35].…”

Section: Robot Learning From Demonstrationmentioning

confidence: 99%

Teaching Robots to Perform Construction Tasks via Learning from Demonstration

Liang¹,

Kamat²,

Menassa³

2019

Proceedings of the International Symposium on Automation and Robotics in Construction (IAARC)

View full text Add to dashboard Cite

Robots are expected to be widely used on future construction sites to assist human workers in the performance of repetitive physically-demanding tasks. Unlike typical manufacturing assembly lines, where parts are delivered to robots and workers in stationary workstations, construction robots and human workers must accumulate all necessary resources and repeatedly navigate to desired assembly locations on-site to perform useful work. The condition of such resources and the geometry of the environment are constantly changing and generally unstructured. As a result, the motion trajectories of any robot arms cannot be programmed beforehand. The robots must define the trajectory based on the encountered workspace geometry. Learning from Demonstration (LfD) methods have the potential to be used in teaching robots specific skills through human demonstration such that the robots can repeat the same process in different conditions. In this research, we explore the LfD method to teach robots to perform repetitive but geometrically adaptive construction tasks of installing suspended ceiling tiles within pre-assembled ceiling grids. The developed method translates the work context from the set of training videos demonstrated by humans to the target scene, then applies the reinforcement learning method to generate the policy for the robot to perform the subsequent ceiling tile installation. The first phase of the proposed method, i.e., the context translation model, is implemented and evaluated by characterizing whether robot-installed ceiling tiles are successfully moved to the grid area. The experiments demonstrate promising results that show the applicability of the LfD method in teaching robots to perform geometrically-adaptive construction tasks.

show abstract

“…Stenmetz et al use an external force/torque sensor mounted on the robot arm to record force and position simultaneously and learn the force profiles for a PID-controller. Abu-Dakka et al redeploy pegin-hole skills in new settings by adding exception strategies that randomly search for the hole to improve robustness [4]. Nemec et al use a priori knowledge and reinforcement learning to reduce the number of demonstrations needed to teach a flip task to a robot [24] and Pastor et al use reinforcement learning to optimize a pool strike [27].…”

Section: Related Workmentioning

confidence: 99%

Simplified Programming of Re-usable Skills on a Safe Industrial Robot

Stenmark

Haaae

Topp

2017

Proceedings of the 2017 ACM/IEEE International Conference on Human-Robot Interaction

View full text Add to dashboard Cite

This paper presents a study on iconic programming support for mainly position-based lead-through programming of an ABB YuMi collaborative robot. A prototype tool supporting a hybrid programming and execution mode was developed and evaluated with 21 non-expert users with varying programming and robotics experience. We also present a comparison of the programming times for an expert robot programmer using traditional tools versus the new tool. The expert programmed the same tasks in 1/5 of the time compared to traditional tools and the non-experts were able to program and debug a LEGO building task using the robot within 30 minutes.

show abstract

Solving peg-in-hole tasks by human demonstration and exception strategies

Cited by 59 publications

References 23 publications

Learning, Improving, and Generalizing Motor Skills for the Peg-in-Hole Tasks Based on Imitation Learning and Self-Learning

Learning, Improving, and Generalizing Motor Skills for the Peg-in-Hole Tasks Based on Imitation Learning and Self-Learning

Teaching Robots to Perform Construction Tasks via Learning from Demonstration

Simplified Programming of Re-usable Skills on a Safe Industrial Robot

Contact Info

Product

Resources

About