Learning grounded finite-state representations from unstructured demonstrations

By Scott Niekum, Sarah Osentoski, George Dimitri Konidaris, Sachin Chitta, Bhaskara Marthi, Andrew Barto. 2015

228 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Control and Systems Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

Citations in the corpus

Benchmark for Skill Learning from Demonstration: Impact of User Experience, Task Complexity, and Start Configuration on Performance (2019, methodology): “Motion Segmentation for Pressing Task Since the pressing task includes two consecutive pressing primitives, we conducted experiments of the task once without and once with motion segmentation [24]–[27].”
Modular Multitask Reinforcement Learning with Policy Sketches (2016, methodology): “An alternative family of approaches employs post-hoc analysis of demonstrations or pretrained policies to extract reusable sub-components (Stolle & Pre-cup, 2002; Konidaris et al., 2011; Niekum et al., 2015).”
An Algorithmic Perspective on Imitation Learning (2018, methodology): “As seen from these previous studies, advances in trajectory segmentation in imitation learning [Kulić et al., 2008, Konidaris et al., 2011, Niekum et al., 2014] are closely related to the methodological advances [Fearnhead and Liu, 2007, Fox et al., 2009] in the machine learning community.”
Unsupervised Perceptual Rewards for Imitation Learning (2016, methodology): “[26, 19, 21], which like our method attempts to decompose”
Using dVRK teleoperation to facilitate deep learning of automation tasks for an industrial robot (2017, background): “In kinesthetic teaching, an operator physically moves the robot in a gravity-balanced “teach mode” to complete the given task [8], [9], [4], [10] and the robot records the motion.”
A tutorial on task-parameterized movement learning and retrieval (2015, background): “An option is to deﬁne a minimal admissible eigenvalue λ min and adjust each covariance matrix Σ ( j ) i so that with i is a matrix containing the stacked eigenvectors of Σ ( j ) i , with λ ( j ) i,k the corresponding eigenvalues.”
Real-time natural language corrections for assistive robotic manipulators (2017, background): “Corrections are also provided through kinesthetic demonstration as described by Niekum et al. (2015) and graphical interfaces (Argall et al., 2012).”
A Review of Robot Learning for Manipulation: Challenges, Representations, and Algorithms (2019, background): “Some approaches actively ask users for additional (partial) demonstrations in areas of the state space in which confidence is low [59] or risk is high [46], while others rely on a human user to identify when a mistake has been made [281].”
KeyIn: Discovering Subgoal Structure with Keyframe-based Video Prediction (2019, background): “L G ] 11 A p r 2019 once a subgoal sequence is discovered (Konidaris et al., 2012; Niekum et al., 2015; Kroemer et al., 2015; Hausman et al., 2017).”
Unsupervised Learning of Sensorimotor Affordances by Stochastic Future Prediction (2018, background): “Methods for aggregating low-level controls into higher-level representations on which planning and RL can be performed are well-studied in the robotics literature: notable examples include the options framework for hierarchical RL [43, 3], dynamic motion primitives for manipulation [38, 37, 31] and ”
Autonomously constructing hierarchical task networks for planning and human-robot collaboration (2016, background): “Recent work has extended skill generalization within LfD, a highly desirable property, by leveraging Bayesian nonparametric models and dynamic movement primitives to extract flexible skills from unstructured demonstrations [11].”
Modeling Long-horizon Tasks as Sequential Interaction Landscapes (2020, background): “Another direction is to learn policies for robot-object interaction from demonstrations with the goal to reduce uncertainty with expert demonstrations [64, 30, 44, 43, 42].”