gradSLAM: Automagically differentiable SLAM

Jatavallabhula, Krishna Murthy; Saryazdi, Saeı̈d; Iyer, Ganesh; Paull, Liam

doi:10.48550/arxiv.1910.10672

Cited by 1 publication

(6 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, differentiable SLAM methods have been also proposed [19], [26]. Differentiable programming methods are Fig.…”

Section: B Learning-based Odometrymentioning

confidence: 99%

“…often used to make a computation model differentiable for optimization and control; when it comes to SLAM, instead of training neural networks from scratch, we can train them to improve the existing SLAM algorithm's performance. Even though those methods focused on visual SLAM, Murty et al [19] proposed a differentiable ICP algorithm that is based on the differentiable Levenberg-Marquardt (LM) optimization algorithm. However, its differentiability is limited by the KNN algorithm.…”

Section: B Learning-based Odometrymentioning

confidence: 99%

“…In this section, we briefly introduce GICP and present our method to differentiate it. Our method is based on the differentiable ICP algorithm [19], and we use the differentiable KNN algorithm [28] to improve the overall differentiability. We aim to compute reliable gradient information with our differentiable GICP method, so we can use it for our WGICP described in Section IV-B.…”

Section: A Differentiable Gicpmentioning

confidence: 99%

“…We chose GICP because it is one of the most effective extensions of ICP and computes the optimal alignment by assuming local planar distributions in the point clouds. Our differentiable GICP algorithm is based on the differentiable ICP algorithm [19]. To enhance the differentiability, we use the differentiable KNN algorithm in our method.…”

Section: Introduction 3d Simultaneous Localization and Mapping (Slam)mentioning

confidence: 99%

“…• Differentiable GICP: We propose a differentiable GICP algorithm that shows better differentiability than [19].…”

Section: Introduction 3d Simultaneous Localization and Mapping (Slam)mentioning

confidence: 99%

See 4 more Smart Citations

WGICP: Differentiable Weighted GICP-Based Lidar Odometry

Son¹,

Ji²,

Ma³

et al. 2022

Preprint

View full text Add to dashboard Cite

We present a novel differentiable weighted generalized iterative closest point (WGICP) method applicable to general 3D point cloud data, including that from lidar. Our method builds on differentiable generalized ICP (GICP) and we use the differentiable K-Nearest Neighbors (KNN) algorithm to enhance differentiability. Our differentiable GICP algorithm provides the gradient of output pose estimation with respect to each input point. This allows us to train a neural network to predict its importance, or weight, in order to estimate the correct pose. In contrast to the other ICP-based methods, our formulation reduces the number of points used for GICP by only selecting those with the highest weights and ignoring redundant ones with lower weights. We show that our method improves both the accuracy (up to 30%) and the speed (up to 2×) over the GICP algorithm on the KITTI dataset. We also demonstrate the benefit of WGICP in terms of developing an improved SLAM system.

show abstract

“…Recently, differentiable SLAM methods have been also proposed [19], [26]. Differentiable programming methods are Fig.…”

Section: B Learning-based Odometrymentioning

confidence: 99%

Section: B Learning-based Odometrymentioning

confidence: 99%

Section: A Differentiable Gicpmentioning

confidence: 99%

Section: Introduction 3d Simultaneous Localization and Mapping (Slam)mentioning

confidence: 99%

“…• Differentiable GICP: We propose a differentiable GICP algorithm that shows better differentiability than [19].…”

Section: Introduction 3d Simultaneous Localization and Mapping (Slam)mentioning

confidence: 99%

See 3 more Smart Citations

WGICP: Differentiable Weighted GICP-Based Lidar Odometry

Son¹,

Ji²,

Ma³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

gradSLAM: Automagically differentiable SLAM

Cited by 1 publication

References 0 publications

WGICP: Differentiable Weighted GICP-Based Lidar Odometry

WGICP: Differentiable Weighted GICP-Based Lidar Odometry

Contact Info

Product

Resources

About