Learning for Active 3D Mapping

TL;DR

This paper introduces an active 3D mapping approach that learns to reconstruct dense occupancy maps from sparse depth data and optimizes the control of depth sensors, significantly improving map accuracy.

Contribution

It presents a novel method combining learning-based reconstruction with greedy control optimization for depth sensors like solid-state lidars.

Findings

Significant improvement in 3D map accuracy on KITTI dataset.

Efficient greedy algorithm with proven approximation ratio.

Effective joint learning and control of depth measurements.

Abstract

We propose an active 3D mapping method for depth sensors, which allow individual control of depth-measuring rays, such as the newly emerging solid-state lidars. The method simultaneously (i) learns to reconstruct a dense 3D occupancy map from sparse depth measurements, and (ii) optimizes the reactive control of depth-measuring rays. To make the first step towards the online control optimization, we propose a fast prioritized greedy algorithm, which needs to update its cost function in only a small fraction of pos- sible rays. The approximation ratio of the greedy algorithm is derived. An experimental evaluation on the subset of the KITTI dataset demonstrates significant improve- ment in the 3D map accuracy when learning-to-reconstruct from sparse measurements is coupled with the optimization of depth-measuring rays.