Dataset Information

Learning Trajectory Distributions for Assisted Teleoperation and Path Planning.

ABSTRACT: Several approaches have been proposed to assist humans in co-manipulation and teleoperation tasks given demonstrated trajectories. However, these approaches are not applicable when the demonstrations are suboptimal or when the generalization capabilities of the learned models cannot cope with the changes in the environment. Nevertheless, in real co-manipulation and teleoperation tasks, the original demonstrations will often be suboptimal and a learning system must be able to cope with new situations. This paper presents a reinforcement learning algorithm that can be applied to such problems. The proposed algorithm is initialized with a probability distribution of demonstrated trajectories and is based on the concept of relevance functions. We show in this paper how the relevance of trajectory parameters to optimization objectives is connected with the concept of Pearson correlation. First, we demonstrate the efficacy of our algorithm by addressing the assisted teleoperation of an object in a static virtual environment. Afterward, we extend this algorithm to deal with dynamic environments by utilizing Gaussian Process regression. The full framework is applied to make a point particle and a 7-DoF robot arm autonomously adapt their movements to changes in the environment as well as to assist the teleoperation of a 7-DoF robot arm in a dynamic environment.

SUBMITTER: Ewerton M

PROVIDER: S-EPMC7806068 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Learning Trajectory Distributions for Assisted Teleoperation and Path Planning.

Ewerton Marco M Arenz Oleg O Maeda Guilherme G Koert Dorothea D Kolev Zlatko Z Takahashi Masaki M Peters Jan J

Frontiers in robotics and AI 20190924

Several approaches have been proposed to assist humans in co-manipulation and teleoperation tasks given demonstrated trajectories. However, these approaches are not applicable when the demonstrations are suboptimal or when the generalization capabilities of the learned models cannot cope with the changes in the environment. Nevertheless, in real co-manipulation and teleoperation tasks, the original demonstrations will often be suboptimal and a learning system must be able to cope with new situat ...[more]

PMID: 33501104

Similar Datasets

Project description:Objective: Stereoelectroencephalography (SEEG) is a procedure in which many electrodes are stereotactically implanted within different regions of the brain to estimate the epileptogenic zone in patients with drug-refractory focal epilepsy. Computer-assisted planning (CAP) improves risk scores, gray matter sampling, orthogonal drilling angles to the skull and intracerebral length in a fraction of the time required for manual planning. Due to differences in planning practices, such algorithms may not be generalizable between institutions. We provide a prospective validation of clinically feasible trajectories using "spatial priors" derived from previous implantations and implement a machine learning classifier to adapt to evolving planning practices. Methods: Thirty-two patients underwent consecutive SEEG implantations utilizing computer-assisted planning over 2 years. Implanted electrodes from the first 12 patients (108 electrodes) were used as a training set from which entry and target point spatial priors were generated. CAP was then prospectively performed using the spatial priors in a further test set of 20 patients (210 electrodes). A K-nearest neighbor (K-NN) machine learning classifier was implemented as an adaptive learning method to modify the spatial priors dynamically. Results: All of the 318 prospective computer-assisted planned electrodes were implanted without complication. Spatial priors developed from the training set generated clinically feasible trajectories in 79% of the test set. The remaining 21% required entry or target points outside of the spatial priors. The K-NN classifier was able to dynamically model real-time changes in the spatial priors in order to adapt to the evolving planning requirements. Conclusions: We provide spatial priors for common SEEG trajectories that prospectively integrate clinically feasible trajectory planning practices from previous SEEG implantations. This allows institutional SEEG experience to be incorporated and used to guide future implantations. The deployment of a K-NN classifier may improve the generalisability of the algorithm by dynamically modifying the spatial priors in real-time as further implantations are performed.

Dataset Information

Learning Trajectory Distributions for Assisted Teleoperation and Path Planning.

Publications

Learning Trajectory Distributions for Assisted Teleoperation and Path Planning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets