RecNet: An Invertible Point Cloud Encoding through Range Image Embeddings for Multi-Robot Map Sharing and Reconstruction

TL;DR

This article introduces RecNet, a novel approach that concurrently addresses both challenges of resource-constrained robots and the need for effective place recognition in multi-robotic systems, and assesses its efficacy and potential for real-world applications.

Abstract

In the field of resource-constrained robots and the need for effective place recognition in multi-robotic systems, this article introduces RecNet, a novel approach that concurrently addresses both challenges. The core of RecNet's methodology involves a transformative process: it projects 3D point clouds into range images, compresses them using an encoder-decoder framework, and subsequently reconstructs the range image, restoring the original point cloud. Additionally, RecNet utilizes the latent vector extracted from this process for efficient place recognition tasks. This approach not only achieves comparable place recognition results but also maintains a compact representation, suitable for sharing among robots to reconstruct their collective maps. The evaluation of RecNet encompasses an array of metrics, including place recognition performance, the structural similarity of the reconstructed point clouds, and the bandwidth transmission advantages, derived from sharing only the latent vectors. Our proposed approach is assessed using both a publicly available dataset and field experiments$^1$, confirming its efficacy and potential for real-world applications.

Figures (5)

• Figure 1: The original and the reconstructed map using RecNet on an experiment around Luleå University of Technology.
• Figure 2: The overall RecNet pipeline is composed of two identical encoder legs that share weights, a single decoder leg and a tail network, responsible for estimating the similarity between the two latent bottleneck vectors.
• Figure 3: Two LiDAR scans from the KITTI dataset, where (a) depicts the original scan, (b) is the reconstructed using the MSE and (c) is the reconstructed using the Gradient MSE.
• Figure 4: The place recognition performance of RecNet compared to others, in the 00 sequence of KITTI.
• Figure 5: The original and reconstructed point cloud maps from the 00 sequence of KITTI. The highlighted segments (a) - (c) provide a closer look at the original map, while the segments (d) - (f) demonstrate the corresponding reconstructed parts. RecNet facilitates a reconstruction that resembles the original map by only transmitting the latent vectors of the network.