∇SLAM: Dense SLAM meets Automatic Differentiation

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

doi:10.1109/icra40945.2020.9197519

Cited by 24 publications

(18 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Regarding learning and adaptation, the integration of SLAM with deep learning specifically will be another key research area over the next decade. Indeed, end-to-end, datadriven machine learning techniques for SLAM are already starting to enter the literature (136,137,132,140). There is ample opportunity to develop novel deep learning systems that are specifically adapted to the unique features of robotic perception; these include rich, temporally-coherent streams of sensory data available to robots, novel sensing modalities and data types beyond classical vision (e.g., direct 3D perception via lidar, event-based Solid lines represent the basic flow of information through the perception subsystem: Sensor measurements Ỹ are received and processed, and a state estimate X is communicated to the rest of the system.…”

Section: Discussion and Future Perspectivesmentioning

confidence: 99%

Advances in Inference and Representation for Simultaneous Localization and Mapping

Rosen¹,

Doherty²,

Espinoza³

et al. 2021

Preprint

View full text Add to dashboard Cite

Simultaneous localization and mapping (SLAM) is the process of constructing a global model of an environment from local observations of it; this is a foundational capability for mobile robots, supporting such core functions as planning, navigation, and control. This article reviews recent progress in SLAM, focusing on advances in the expressive capacity of the environmental models used in SLAM systems (representation) and the performance of the algorithms used to estimate these models from data (inference). A prominent theme of recent SLAM research is the pursuit of environmental representations (including learned representations) that go beyond the classical attributes of geometry and appearance to model properties such as hierarchical organization, affordance, dynamics, and semantics; these advances equip autonomous agents with a more comprehensive understanding of the world, enabling more versatile and intelligent operation. A second major theme is a revitalized interest in the mathematical properties of the SLAM estimation problem itself (including its computational and informationtheoretic performance limits); this work has led to the development of novel classes of certifiable and robust inference methods that dramatically improve the reliability of SLAM systems in real-world operation. We survey these advances with an emphasis on their ramifications for achieving robust, long-duration autonomy, and conclude with a discussion of open challenges and a perspective on future research directions.

show abstract

Section: Discussion and Future Perspectivesmentioning

confidence: 99%

Advances in Inference and Representation for Simultaneous Localization and Mapping

Rosen¹,

Doherty²,

Espinoza³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Both the goal-directed and mapping-focused methods aim to reconstruct the geometry of their environment, and do not directly benefit from training with prior data. Some approaches have sought to incorporate learning into mapping and reconstruction [21,22], which benefits from prior data, but still aims at dense geometric reconstruction. Our approach uses a model that is trained with data from prior environments to predict traversability rather than geometry, and this model is then used in combination with geographic hints to plan a path to the goal.…”

Section: Roadmap Hintmentioning

confidence: 99%

ViKiNG: Vision-Based Kilometer-Scale Navigation with Geographic Hints

Shah,

Levine

2022

Preprint

View full text Add to dashboard Cite

“…This work introduces a differentiable SLAM approach that fills a gap in this literature. While Jatavallabhula et al [31] have investigated differentiable SLAM pipelines, they focus solely on the effect of differentiable approximations and do not perform learning of any kind.…”

Section: Related Workmentioning

confidence: 99%

Differentiable SLAM-net: Learning Particle SLAM for Visual Navigation

Karkus¹,

Cai²,

Hsu³

2021

Preprint

View full text Add to dashboard Cite

Simultaneous localization and mapping (SLAM) remains challenging for a number of downstream applications, such as visual robot navigation, because of rapid turns, featureless walls, and poor camera quality. We introduce the Differentiable SLAM Network (SLAM-net) along with a navigation architecture to enable planar robot navigation in previously unseen indoor environments. SLAM-net encodes a particle filter based SLAM algorithm in a differentiable computation graph, and learns task-oriented neural network components by backpropagating through the SLAM algorithm. Because it can optimize all model components jointly for the end-objective, SLAM-net learns to be robust in challenging conditions. We run experiments in the Habitat platform with different real-world RGB and RGB-D datasets. SLAM-net significantly outperforms the widely adapted ORB-SLAM in noisy conditions. Our navigation architecture with SLAMnet improves the state-of-the-art for the Habitat Challenge 2020 PointNav task by a large margin (37% to 64% success).

show abstract

∇SLAM: Dense SLAM meets Automatic Differentiation

Cited by 24 publications

References 35 publications

Advances in Inference and Representation for Simultaneous Localization and Mapping

Advances in Inference and Representation for Simultaneous Localization and Mapping

ViKiNG: Vision-Based Kilometer-Scale Navigation with Geographic Hints

Differentiable SLAM-net: Learning Particle SLAM for Visual Navigation

Contact Info

Product

Resources

About