Unifying Map and Landmark Based Representations for Visual Navigation

TL;DR

This paper introduces a learned visual navigation framework that combines map and landmark representations, enabling efficient and robust navigation in noisy and novel environments using sparse images.

Contribution

It presents a unified, data-driven formulation integrating path planning, feature synthesis, and goal-driven control for visual navigation.

Findings

Achieves performance gains over baseline methods in simulated environments.

Effectively navigates in noisy and novel environments with sparse visual input.

Demonstrates the benefits of unifying map and landmark representations.

Abstract

This works presents a formulation for visual navigation that unifies map based spatial reasoning and path planning, with landmark based robust plan execution in noisy environments. Our proposed formulation is learned from data and is thus able to leverage statistical regularities of the world. This allows it to efficiently navigate in novel environments given only a sparse set of registered images as input for building representations for space. Our formulation is based on three key ideas: a learned path planner that outputs path plans to reach the goal, a feature synthesis engine that predicts features for locations along the planned path, and a learned goal-driven closed loop controller that can follow plans given these synthesized features. We test our approach for goal-driven navigation in simulated real world environments and report performance gains over competitive baseline approaches.