Learning mobile manipulation actions from human demonstrations

Welschehold, Tim; Dornhege, Christian; Burgard, Wolfram

doi:10.1109/iros.2017.8206152

Cited by 36 publications

(38 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…During optimization the teacher's demonstrations are adapted towards trajectories that are feasible for robot execution. For details on the graph structure and the implementation we refer to Welschehold et al [22], [23].…”

Section: Action Learningmentioning

confidence: 99%

“…We evaluate our recently proposed graph-based approach [23] for learning a mobile manipulation task from human demonstrations on data acquired with the approach for 3D human pose estimation presented in this work. We evaluate the methods on the same four tasks as in our previous work [23]: one task of opening and moving through a room door and three tasks of opening small furniture pieces. The tasks will be referred to as room door, swivel door, drawer, and sliding door.…”

Section: Experiments -Action Learningmentioning

confidence: 99%

“…Specifically we compare the optimization for all aforementioned tasks for two different teacher pose estimation methods. The first relies on detecting markers attached to the teachers hand and torso and was conducted for our previous work [23]. The second follows the approach presented in this work.…”

Section: A Adapting Human Demonstrations To Robot Requirementsmentioning

confidence: 99%

See 2 more Smart Citations

3D Human Pose Estimation in RGBD Images for Robotic Task Learning

Zimmermann

Welschehold

Dornhege

et al. 2018

2018 IEEE International Conference on Robotics and Automation (ICRA)

Self Cite

132

View full text Add to dashboard Cite

We propose an approach to estimate 3D human pose in real world units from a single RGBD image and show that it exceeds performance of monocular 3D pose estimation approaches from color as well as pose estimation exclusively from depth. Our approach builds on robust human keypoint detectors for color images and incorporates depth for lifting into 3D. We combine the system with our learning from demonstration framework to instruct a service robot without the need of markers. Experiments in real world settings demonstrate that our approach enables a PR2 robot to imitate manipulation actions observed from a human teacher.

show abstract

Section: Action Learningmentioning

confidence: 99%

Section: Experiments -Action Learningmentioning

confidence: 99%

See 1 more Smart Citation

3D Human Pose Estimation in RGBD Images for Robotic Task Learning

Zimmermann

Welschehold

Dornhege

et al. 2018

2018 IEEE International Conference on Robotics and Automation (ICRA)

Self Cite

132

View full text Add to dashboard Cite

show abstract

“…IV-B.2. The grasping trajectory is generated based on the current pose in s t of the object associated with action a t (the grasp pose is learned as part of the model, see [2]), and the motion is performed with respect to the reference frame of the object defined by the corresponding template γ t . We set the end pose of each trajectory as the initial state of the next motion model.…”

Section: Solving the Task Through Search-based Optimizationmentioning

confidence: 99%

“…We evaluate our contributions thoroughly in real-world experiments with a PR2 robot. Note that an earlier version of this work with preliminary results was presented in [4].…”

Section: Introductionmentioning

confidence: 99%

Combined Task and Action Learning from Human Demonstrations for Mobile Manipulation Applications

Welschehold

Abdo

Dornhege

et al. 2019

2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Self Cite

View full text Add to dashboard Cite

Learning from demonstrations is a promising paradigm for transferring knowledge to robots. However, learning mobile manipulation tasks directly from a human teacher is a complex problem as it requires learning models of both the overall task goal and of the underlying actions. Additionally, learning from a small number of demonstrations often introduces ambiguity with respect to the intention of the teacher, making it challenging to commit to one model for generalizing the task to new settings. In this paper, we present an approach to learning flexible mobile manipulation action models and task goal representations from teacher demonstrations. Our action models enable the robot to consider different likely outcomes of each action and to generate feasible trajectories for achieving them. Accordingly, we leverage a probabilistic framework based on Monte Carlo tree search to compute sequences of feasible actions imitating the teacher intention in new settings without requiring the teacher to specify an explicit goal state. We demonstrate the effectiveness of our approach in complex tasks carried out in real-world settings.

show abstract

Haptic Programming

Stumm¹,

Brell-Çokcan²

2018

Robotic Fabrication in Architecture, Art and Design 2018

View full text Add to dashboard Cite

Learning mobile manipulation actions from human demonstrations

Cited by 36 publications

References 17 publications

3D Human Pose Estimation in RGBD Images for Robotic Task Learning

3D Human Pose Estimation in RGBD Images for Robotic Task Learning

Combined Task and Action Learning from Human Demonstrations for Mobile Manipulation Applications

Haptic Programming

Contact Info

Product

Resources

About