Learning Human Search Behavior from Egocentric Visual Inputs

TL;DR

This paper presents a method for developing virtual agents that search for objects in 3D environments using egocentric vision and locomotion, resulting in human-like navigation behaviors without relying on privileged 3D data.

Contribution

It introduces a novel approach combining a search policy and online replanning to enable natural, human-like search behaviors in virtual agents using only RGBD inputs.

Findings

Effective search of occluded objects in indoor scenes

Policy generalizes across different characters without retraining

Agents exhibit natural navigation behaviors

Abstract

"Looking for things" is a mundane but critical task we repeatedly carry on in our daily life. We introduce a method to develop a human character capable of searching for a randomly located target object in a detailed 3D scene using its locomotion capability and egocentric vision perception represented as RGBD images. By depriving the privileged 3D information from the human character, it is forced to move and look around simultaneously to account for the restricted sensing capability, resulting in natural navigation and search behaviors. Our method consists of two components: 1) a search control policy based on an abstract character model, and 2) an online replanning control module for synthesizing detailed kinematic motion based on the trajectories planned by the search policy. We demonstrate that the combined techniques enable the character to effectively find often occluded household items in indoor environments. The same search policy can be applied to different full-body characters without the need for retraining. We evaluate our method quantitatively by testing it on randomly generated scenarios. Our work is a first step toward creating intelligent virtual agents with humanlike behaviors driven by onboard sensors, paving the road toward future robotic applications.