Robot Learning of Object Manipulation Task Actions From Human Demonstrations

Kyrarini, Maria; Haseeb, Muhammad; Ristić-Durrant, Danijela; Gräser, Axel

doi:10.22190/fume170515010k

Cited by 5 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The method proposed in Kyrarini et al (2017) uses Gaussian Mixture Models (GMMs) to learn a motion from several human demonstrations. Similarly to the proposed approach they used kinesthetic teaching to record the demonstrations.…”

Section: Related Workmentioning

confidence: 99%

Guided Stochastic Optimization for Motion Planning

et al. 2019

View full text Add to dashboard Cite

Learning from Demonstration (LfD) is a family of methods used to teach robots specific tasks. It is used to assist them with the increasing difficulty of performing manipulation tasks in a scalable manner. The state-of-the-art in collaborative robots allows for simple LfD approaches that can handle limited parameter changes of a task. These methods however typically approach the problem from a control perspective and therefore are tied to specific robot platforms. In contrast, this paper proposes a novel motion planning approach that combines the benefits of LfD approaches with generic motion planning that can provide robustness to the planning process as well as scaling task learning both in number of tasks and number of robot platforms. Specifically, it introduces Dynamical Movement Primitives (DMPs) based LfD as initial trajectories for the Stochastic Optimization for Motion Planning (STOMP) framework. This allows for successful task execution even when the task parameters and the environment change. Moreover, the proposed approach allows for skill transfer between robots. In this case a task is demonstrated to one robot via kinesthetic teaching and can be successfully executed by a different robot. The proposed approach, coined Guided Stochastic Optimization for Motion Planning (GSTOMP) is evaluated extensively using two different manipulator systems in simulation and in real conditions. Results show that GSTOMP improves task success compared to simple LfD approaches employed by the state-of-the-art collaborative robots. Moreover, it is shown that transferring skills is feasible and with good performance. Finally, the proposed approach is compared against a plethora of state-of-the-art motion planners. The results show that the motion planning performance is comparable or better than the state-of-the-art.

show abstract

Section: Related Workmentioning

confidence: 99%

Guided Stochastic Optimization for Motion Planning

et al. 2019

View full text Add to dashboard Cite

show abstract

“…A widely used solution for the first problem is Dynamic Time Warping (DTW) (Akgun et al 2012;Sabbaghi et al 2014). Using of DTW also was proposed by the authors as possible solution of the second problem (Kyrarini et al 2016(Kyrarini et al , 2017. However, one disadvantage of DTW is the complexity of the algorithm which is time consuming (Movchan and Zymbler 2015).…”

Section: Selection Of Similar Demonstrationsmentioning

confidence: 99%

Robot learning of industrial assembly task via human demonstrations

et al. 2018

Self Cite

View full text Add to dashboard Cite

Human-robot collaboration in industrial applications is a challenging robotic task. Human working together with the robot at a workplace to complete a task may create unpredicted events for the robot, as humans can act unpredictably. Humans tend to perform a task in a not fully repetitive manner using their expertise and cognitive capabilities. The traditional robot programming cannot cope with these challenges of human-robot collaboration. In this paper, a framework for robot learning by multiple human demonstrations is introduced. Through the demonstrations, the robot learns the sequence of actions for an assembly task (high-level learning) without the need of pre-programming. Additionally, the robot learns every path as needed for object manipulation (low-level learning). Once the robot has the knowledge of the demonstrated task, it can perform the task in collaboration with the human. However, the need for adaptation of the learned knowledge may arise as the human collaborator could introduce changes in the environment, such as placing an object to be manipulated in a position and orientation different from the demonstrated ones. In this paper, a novel real-time adaptation algorithm to cope with these changes in the environment, introduced by the human factor, is proposed. The proposed algorithm is able to identify the sequence of actions needed to be performed in a new environment. A Gaussian Mixture Model-based modification algorithm is able to adapt the learned path in order to enable robot to successfully complete the task without the need of additional training by demonstration. The proposed framework copes with changes in the position and orientation of the objects to be manipulated and also provides obstacle avoidance. Moreover, the framework enables the human collaborator to suggest different sequence of actions for the learned task, which will be performed by the robot. The proposed algorithm was tested on a dual-arm industrial robot in an assembly scenario and the results are presented. Shown results demonstrate a potential of the proposed robot learning framework to enable continuous human-robot collaboration.

show abstract

“…Identifying transformations to the task frame is part of the embodiment mapping. Let us keep in mind that for some tasks that involve, for instance, manipulated object [2,18,19,10], tools [20], etc. ; other elements should be recorded during the demonstrations with respect to the task frame.…”

Section: Introductionmentioning

confidence: 99%