Robotic manipulation of multiple objects as a POMDP

Pajarinen, Joni; Kyrki, Ville

doi:10.1016/j.artint.2015.04.001

Cited by 45 publications

(40 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, even the smallest tasks we investigated had a discrete time horizon an order of magnitude greater. Therefore, we suspect that a fundamentally different mechanism from the one presented in [37] will be required for these tasks, and believe that the approximation algorithms presented in this paper may serve as a starting point for a new approach.…”

Section: Discussionmentioning

confidence: 99%

“…One possibility may be coupling replanning with an online POMDP process and/or a true state estimator that can fix the policy. There is promising active research in online POMDP planning [37]. However, this robotic POMDP application is at a short-term reactive cycle stage, only looking forward in time at most 4 actions in the examples provided.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Extending the Applicability of POMDP Solutions to Robotic Tasks

2015

View full text Add to dashboard Cite

Abstract-Partially-Observable Markov Decision Processes (POMDPs) are used in many robotic task classes from soccer to household chores. Determining an approximately optimal action policy for POMDPs is PSPACE-complete, and the exponential growth of computation time prohibits solving large tasks. This paper describes two techniques to extend the range of robotic tasks that can be solved using a POMDP. Our first technique reduces the motion constraints of a robot, and then uses state-of-the-art robotic motion planning techniques to respect the true motion constraints at runtime. We then propose a novel task decomposition that can be applied to some indoor robotic tasks. This decomposition transforms a long time horizon task into a set of shorter tasks.We empirically demonstrate the performance gain provided by these two techniques through simulated execution in a variety of environments. Comparing a direct formulation of a POMDP to solving our proposed reductions, we conclude that the techniques proposed in this paper can provide significant enhancement to current POMDP solution techniques, extending the POMDP instances that can be solved to include large, continuous-state robotic tasks.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Extending the Applicability of POMDP Solutions to Robotic Tasks

2015

View full text Add to dashboard Cite

show abstract

“…Pajarinen and Kyrki presented a partially observable MDP (POMDP) for robot manipulation such as cleaning dishes in the dishwasher. 97 The POMDP is used to estimate different action choices based on optimization reward function with the probabilistic model in an uncertain world.…”

Section: Supervised Learning Sl Is the Machine Learning Task That Maps An Input To An Output Based On A Labeled Set Of Training Examples mentioning

confidence: 99%

Control strategies for cleaning robots in domestic applications: A comprehensive review

Kim

Mishra

Limosani

et al. 2019

International Journal of Advanced Robotic Systems

View full text Add to dashboard Cite

Service robots are built and developed for various applications to support humans as companion, caretaker, or domestic support. As the number of elderly people grows, service robots will be in increasing demand. Particularly, one of the main tasks performed by elderly people, and others, is the complex task of cleaning. Therefore, cleaning tasks, such as sweeping floors, washing dishes, and wiping windows, have been developed for the domestic environment using service robots or robot manipulators with several control approaches. This article is primarily focused on control methodology used for cleaning tasks. Specifically, this work mainly discusses classical control and learning-based controlled methods. The classical control approaches, which consist of position control, force control, and impedance control , are commonly used for cleaning purposes in a highly controlled environment. However, classical control methods cannot be generalized for cluttered environment so that learning-based control methods could be an alternative solution. Learning-based control methods for cleaning tasks can encompass three approaches: learning from demonstration (LfD), supervised learning (SL), and reinforcement learning (RL). These control approaches have their own capabilities to generalize the cleaning tasks in the new environment. For example, LfD, which many research groups have used for cleaning tasks, can generate complex cleaning trajectories based on human demonstration. Also, SL can support the prediction of dirt areas and cleaning motion using large number of data set. Finally, RL can learn cleaning actions and interact with the new environment by the robot itself. In this context, this article aims to provide a general overview of robotic cleaning tasks based on different types of control methods using manipulator. It also suggest a description of the future directions of cleaning tasks based on the evaluation of the control approaches.

show abstract

“…Other works focus on handling occlusion and noisy sensors while performing the task of dishwasher loading. 24 They model multi-object manipulation of crowded occluded objects as a partially observable Markov decision process in order to improve planning process. This study was, however, limited to a single object and ignored the complexities of object detection.…”

Section: Introductionmentioning

confidence: 99%

Autonomous dishwasher loading from cluttered trays using pre‐trained deep neural networks

Voysey

Thuruthel

Iida

2020

Engineering Reports

View full text Add to dashboard Cite

Autonomous dishwasher loading is a benchmark problem in robotics that highlights the challenges of robotic perception, planning, and manipulation in an unstructured environment. Current approaches resort to a specialized solution, however, these technologies are not viable in a domestic setting. Learning-based solutions seem promising for a general purpose solutions; however, they require large amounts of catered data to be applied in real-world scenarios. This article presents a novel learning-based solution without a training phase using pre-trained object detection networks. By developing a perception, planning, and manipulation framework around an off-the-shelf object detection network, we are able to develop robust pick-and-place solutions that are easy to develop and general purpose requiring only a RGB feedback and a pinch gripper. Analysis of a real-world canteen tray data is first performed and used for developing our in-lab experimental setup. Our results obtained from real-world scenarios indicate that such approaches are highly desirable for plug-and-play domestic applications with limited calibration. All the associated data and code of this work are shared in a public repository.

show abstract

Robotic manipulation of multiple objects as a POMDP

Cited by 45 publications

References 32 publications

Extending the Applicability of POMDP Solutions to Robotic Tasks

Extending the Applicability of POMDP Solutions to Robotic Tasks

Control strategies for cleaning robots in domestic applications: A comprehensive review

Autonomous dishwasher loading from cluttered trays using pre‐trained deep neural networks

Contact Info

Product

Resources

About