Trajectory-Based Skill Learning Using Generalized Cylinders.

S Reza Ahmadzadeh,Sonia Chernova

doi:10.3389/frobt.2018.00132

Abstract

In this article, we introduce Trajectory Learning using Generalized Cylinders (TLGC), a novel trajectory-based skill learning approach from human demonstrations. To model a demonstrated skill, TLGC uses a Generalized Cylinder—a geometric representation composed of an arbitrary space curve called the spine and a surface with smoothly varying cross-sections. Our approach is the first application of Generalized Cylinders to manipulation, and its geometric representation offers several key features: it identifies and extracts the implicit characteristics and boundaries of the skill by encoding the demonstration space, it supports for generation of multiple skill reproductions maintaining those characteristics, the constructed model can generalize the skill to unforeseen situations through trajectory editing techniques, our approach also allows for obstacle avoidance and interactive human refinement of the resulting model through kinesthetic correction. We validate our approach through a set of real-world experiments with both a Jaco 6-DOF and a Sawyer 7-DOF robotic arm.

Highlights

Learning from Demonstration (LfD) approaches provide the ability to interactively teach robots new skills, eliminating the need for manual programming of the desired behavior (Argall et al, 2009b)
We conducted eight experiments on two robotic platforms to demonstrate the encoding of the Generalized Cylinder (GC) model, as well as its reproduction and generalization capabilities, on multiple trajectory-based skills8
The results show that both Dynamic Movement Primitives (DMPs) and Gaussian Mixture Models (GMM)/Gaussian Mixture Regression (GMR) are capable of learning the skill

Summary

Introduction

Learning from Demonstration (LfD) approaches provide the ability to interactively teach robots new skills, eliminating the need for manual programming of the desired behavior (Argall et al, 2009b). By observing a set of human-provided examples and constructing a model, LfD approaches can reproduce the skill and generalize it to novel situations autonomously. These capabilities make LfD a powerful approach that has the potential to enable even non-expert users to teach new skills to robots with minimum effort. In robotics, generalized cylinders have been used for finding flyable paths for unmanned aerial vehicles (Shanmugavel et al, 2007) They have been used for collision detection during physical human-robot interaction (Martínez-Salvador et al, 2003; Corrales et al, 2011). We first outline the mathematical definition and parameterized formulation of Canal Surfaces (CS) (Hilbert and Cohn-Vossen, 1952), which are a simpler form of GCs, and extend the formulae to generalized cylinders

Objectives

Methods

Results

Conclusion