Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pose-free object classification from surface contact features in sequences of Robotic grasps

Teresa Alves, Alexandre Bernardino, Plinio Moreno

TL;DR

This work tackles pose-free haptic object recognition using a multi-finger hand by proposing two methods: PN, which uses contact positions and normals, and P, which uses only contact positions. Both methods rely on per-object hash tables built from a limited number of grasps and a Bayesian voting scheme to update object probabilities, enabling identification without knowing the hand–object pose. The authors also present an active-exploration variant that reduces grasp requirements when pose information is available. In GraspIt! simulations with five YCB objects, PN consistently achieves higher accuracy and requires fewer grasps than P, and active exploration further minimizes grasp counts and classification errors, demonstrating the practicality of pose-free haptic recognition for robust manipulation.

Abstract

In this work, we propose two cost efficient methods for object identification, using a multi-fingered robotic hand equipped with proprioceptive sensing. Both methods are trained on known objects and rely on a limited set of features, obtained during a few grasps on an object. Contrary to most methods in the literature, our methods do not rely on the knowledge of the relative pose between object and hand, which greatly expands the domain of application. However, if that knowledge is available, we propose an additional active exploration step that reduces the overall number of grasps required for a good recognition of the object. One of the methods depends on the contact positions and normals and the other depends on the contact positions alone. We test the proposed methods in the GraspIt! simulator and show that haptic-based object classification is possible in pose-free conditions. We evaluate the parameters that produce the most accurate results and require the least number of grasps for classification.

Pose-free object classification from surface contact features in sequences of Robotic grasps

TL;DR

This work tackles pose-free haptic object recognition using a multi-finger hand by proposing two methods: PN, which uses contact positions and normals, and P, which uses only contact positions. Both methods rely on per-object hash tables built from a limited number of grasps and a Bayesian voting scheme to update object probabilities, enabling identification without knowing the hand–object pose. The authors also present an active-exploration variant that reduces grasp requirements when pose information is available. In GraspIt! simulations with five YCB objects, PN consistently achieves higher accuracy and requires fewer grasps than P, and active exploration further minimizes grasp counts and classification errors, demonstrating the practicality of pose-free haptic recognition for robust manipulation.

Abstract

In this work, we propose two cost efficient methods for object identification, using a multi-fingered robotic hand equipped with proprioceptive sensing. Both methods are trained on known objects and rely on a limited set of features, obtained during a few grasps on an object. Contrary to most methods in the literature, our methods do not rely on the knowledge of the relative pose between object and hand, which greatly expands the domain of application. However, if that knowledge is available, we propose an additional active exploration step that reduces the overall number of grasps required for a good recognition of the object. One of the methods depends on the contact positions and normals and the other depends on the contact positions alone. We test the proposed methods in the GraspIt! simulator and show that haptic-based object classification is possible in pose-free conditions. We evaluate the parameters that produce the most accurate results and require the least number of grasps for classification.
Paper Structure (16 sections, 7 equations, 7 figures, 3 tables)

This paper contains 16 sections, 7 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Approach
  4. Methods
  5. Data Collection and processing
  6. Exploration Techniques
  7. Experiments
  8. Robotic Simulator and Data collection
  9. Creating Hash Tables
  10. Processing grasp values for identification
  11. Results
  12. Passive learning approach
  13. Active learning approach
  14. Comparing Passive and Active Approaches
  15. Overall results
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Objects used for classification
  • Figure 2: Representation of the average number of grasps for the classification of the objects, for both methods - PN (continuous) and P (dashed)
  • Figure 3: Percentual perception errors obtained for each object's classification, using Passive learning
  • Figure 4: Representation of the average number of grasps for the classification of the objects for both methods - PN (continuous) and P (dashed)
  • Figure 5: Percentual perception errors obtained for each object's classification, using Active learning
  • ...and 2 more figures