Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 48 of the 52 references it cites were found in a registry or an index; 18 of the 25 sources its claims cite were checked for support.

Catching Objects in Flight

Open access
Kim, Seungsu, Shukla, Ashwini, Billard, Aude · 2014 · IEEE Transactions on Robotics

Abstract

We address the difficult problem of catching in-flight objects with uneven shapes. This requires the solution of three complex problems: accurate prediction of the trajectory of fastmoving objects, predicting the feasible catching configuration, and planning the arm motion, and all within milliseconds. We follow a programming-by-demonstration approach in order to learn, from throwing examples, models of the object dynamics and arm movement. We propose a new methodology to find a feasible catching configuration in a probabilistic manner. We use the dynamical systems approach to encode motion from several demonstrations. This enables a rapid and reactive adaptation of the arm motion in the presence of sensor uncertainty. We validate the approach in simulation with the iCub humanoid robot and in real-world experiments with the KUKA LWR 4+ (7-degree-of-freedom arm robot) to catch a hammer, a tennis racket, an empty bottle, a partially filled bottle, and a cardboard box.

Overview

CoChat Summary

## TL;DR

This paper presents a real-time robot-catching framework that learns object flight dynamics and feasible grasping/reachability from demonstrations, then combines them to predict where and when to catch in-flight objects. It is validated in iCub simulation and on a KUKA LWR 4+ with an Allegro hand, achieving 100% success in simulation and 73.2% in real-world trials.

## Key findings

  • The paper reports a 100% success rate for catching two objects in iCub simulation, based on 50 throws; 3 of 50 throws were excluded because they did not enter the robot’s reachable space.
  • In real-robot experiments, the system caught 52 out of 71 total trials, yielding a 73.2% success rate; 9 of 80 throws were excluded because they never entered reachable space.
  • The system predicts the best catching configuration in about 0.2 ms on a 2.7-GHz quad-core PC, enabling real-time replanning.
  • For the real robot, the paper reports mean flying time across all trials of 4.97 ± 0.46 s, and the model predicts trajectories better than a rigid-body model, especially for the partially filled bottle.

## Methods

  • The framework has two iterative threads: one continuously predicts object trajectory and updates the best catching configuration and timing; the other adapts arm and finger motion to the updated target.
  • Object dynamics are learned from demonstrations using support vector regression with an RBF kernel and coupled with an EKF for noisy sensing; the paper states the model uses about 20 throwing examples per object.
  • Graspable space and reachable space are each modeled as Gaussian mixture models, with the optimal catching posture found by maximizing the joint probability of the combined models; arm/hand motion is generated using coupled dynamical systems (CDS) and a timing controller.
  • Validation uses iCub simulation with a hammer and tennis racket, and a KUKA LWR 4+ with Allegro hand in real-world experiments on an empty bottle, partially filled bottle, tennis racket, and cardboard box.

## Limitations

  • The paper states that the trajectory model is local: if an object is thrown very differently from the demonstrations, prediction can become inaccurate, especially for highly nonlinear motions such as a partially filled bottle.
  • The authors say they do not explicitly compute collisions between the object and the hand/arm before catching; they use heuristics instead, and this leads to some failures.
  • They state they model only task dynamics, not robot dynamics, so some generated trajectories produce velocities the robot cannot follow.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/tro.2014.2316022

Cites

48 of 48 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

14 Valid · 9 Not valid · 2 Not checked · 81 Own results and interpretations

50 more claims · 5 with their cited sources checked

Create a free account to see all 106 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 31
  • Probably not a substantial claim. 22
  • The quote only partly states this claim. 4

Create a free account to see these claims.

Citations

Cites

Research works printed but not linked: 1

  1. schaal, 0, Programmable pattern generators, Proc Int Conf Comput Intell Neurosci, 48

Cited by

See all 220

Metadata

Authors
Kim, S., Shukla, A., & Billard, A.
Affiliations
École Polytechnique Fédérale de Lausanne (CH)
Venue
IEEE Transactions on Robotics
Published
2014-10
Volume
30
Issue
5
Pages
1049-1065
ISSN
  • 1552-3098 (ISSN-L)
  • 1941-0468
OpenAlex
W2018020788
Cited by
220
References
52
Citations by year
2026: 8 · 2025: 14 · 2024: 27 · 2023: 19 · 2022: 20 · 2021: 29 · 2020: 20 · 2019: 12 · 2018: 23 · 2017: 17 · 2016: 18 · 2015: 10 · 2014: 3
FWCI
17.81
Type
journal article
Language
en
Open access
Locations
  • IEEE Transactions on Robotics (publishedVersion)
  • Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion)
  • Infoscience (Ecole Polytechnique Fédérale de Lausanne) (submittedVersion, cc-by-nc-nd)
Funders
  • EU Projects AMARSI (FP7-ICT-248311)
  • FirstMM (FP7-248258)
Topics
  • Physical Sciences › Engineering › Control and Systems Engineering › Robot Manipulation and Learning
  • Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Motor Control and Adaptation
Keywords
trajectory prediction, programming by demonstration, dynamical systems, sensor uncertainty, humanoid robots, robot manipulators, iCub

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.