Object Goal Navigation Based on Semantics and RGB Ego View

TL;DR

This paper introduces a semantic and RGB ego view-based navigation system enabling service robots to find objects in unknown indoor environments, outperforming humans in completion time during tests.

Contribution

It proposes a novel architecture combining semantic understanding with RGB ego perception for object goal navigation in unknown environments.

Findings

Outperforms human users in navigation completion time

Effective in both simulation and real-world indoor settings

Utilizes GeoSem map for semantic and geometric integration

Abstract

This paper presents an architecture and methodology to empower a service robot to navigate an indoor environment with semantic decision making, given RGB ego view. This method leverages the knowledge of robot's actuation capability and that of scenes, objects and their relations -- represented in a semantic form. The robot navigates based on GeoSem map - a relational combination of geometric and semantic map. The goal given to the robot is to find an object in a unknown environment with no navigational map and only egocentric RGB camera perception. The approach is tested both on a simulation environment and real life indoor settings. The presented approach was found to outperform human users in gamified evaluations with respect to average completion time.