GAPLE: Generalizable Approaching Policy LEarning for Robotic Object Searching in Indoor Environment

TL;DR

This paper introduces GAPLE, a novel approach for training a generalizable policy for robotic object searching in indoor environments, leveraging depth and semantic segmentation features to improve transferability across different scenes.

Contribution

GAPLE is a new method that uses depth and semantic features to enable robots to actively approach objects with better generalization across environments.

Findings

Validated on House3D dataset and real-world platform

Demonstrated improved generalization over scene-specific methods

Provided qualitative analysis of approach behaviors

Abstract

We study the problem of learning a generalizable action policy for an intelligent agent to actively approach an object of interest in an indoor environment solely from its visual inputs. While scene-driven or recognition-driven visual navigation has been widely studied, prior efforts suffer severely from the limited generalization capability. In this paper, we first argue the object searching task is environment dependent while the approaching ability is general. To learn a generalizable approaching policy, we present a novel solution dubbed as GAPLE which adopts two channels of visual features: depth and semantic segmentation, as the inputs to the policy learning module. The empirical studies conducted on the House3D dataset as well as on a physical platform in a real world scenario validate our hypothesis, and we further provide in-depth qualitative analysis.