Memory Unscented Particle Filter for 6-DOF Tactile Localization

TL;DR

This paper introduces the Memory Unscented Particle Filter (MUPF), a real-time algorithm for 6-DOF tactile object localization that effectively handles multimodal distributions using contact measurements.

Contribution

The paper presents a novel MUPF algorithm combining particle filtering with memory and unscented Kalman filtering for improved tactile localization.

Findings

Accurate localization with few particles

Real-time performance demonstrated on a humanoid robot

Effective handling of multimodal distributions

Abstract

This paper addresses 6-DOF (degree-of-freedom) tactile localization, i.e. the pose estimation of tridimensional objects given tactile measurements. This estimation problem is fundamental for the operation of autonomous robots that are often required to manipulate and grasp objects whose pose is a-priori unknown. The nature of tactile measurements, the strict time requirements for real-time operation and the multimodality of the involved probability distributions pose remarkable challenges and call for advanced nonlinear filtering techniques. Following a Bayesian approach, this paper proposes a novel and effective algorithm, named Memory Unscented Particle Filter (MUPF), which solves the 6-DOF localization problem recursively in real-time by only exploiting contact point measurements. MUPF combines a modified particle filter that incorporates a sliding memory of past measurements to better handle multimodal distributions, along with the unscented Kalman filter that moves the particles towards regions of the search space that are more likely with the measurements. The performance of the proposed MUPF algorithm has been assessed both in simulation and on a real robotic system equipped with tactile sensors (i.e., the iCub humanoid robot). The experiments show that the algorithm provides accurate and reliable localization even with a low number of particles and, hence, is compatible with real-time requirements.