Combining optimal control and learning for autonomous aerial navigation in novel indoor environments

TL;DR

This paper presents a framework combining optimal control and learning for MAV indoor navigation, enabling obstacle avoidance using only onboard sensors in novel environments, validated in simulation.

Contribution

It introduces a novel approach that integrates optimal control with perception learning to navigate indoor environments without prior maps.

Findings

Effective obstacle avoidance in novel scenes

Perception module learns from privileged information

Framework works with onboard sensors alone

Abstract

This report proposes a combined optimal control and perception framework for Micro Aerial Vehicle (MAV) autonomous navigation in novel indoor enclosed environments, relying exclusively on on-board sensor data. We use privileged information from a simulator to generate optimal waypoints in 3D space that our perception system learns to imitate. The trained learning based perception module in turn is able to generate similar obstacle avoiding waypoints from sensor data (RGB + IMU) alone. We demonstrate the efficacy of the framework across novel scenes in the iGibson simulation environment.