Spatially-dependent Bayesian semantic perception under model and localization uncertainty

Feldman, Yuri; Indelman, Vadim

doi:10.1007/s10514-020-09921-0

Cited by 3 publications

(1 citation statement)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, the classifier output depends on the relative viewpoint between camera and object. This dependency can be modeled [4][5][6][7], and then be used to improve classification and localization accuracy within a SLAM setting. But, this kind of model was not used in epistemic uncertainty aware classification.…”

Section: Introductionmentioning

confidence: 99%

Epistemic Uncertainty Aware Semantic Localization and Mapping for Inference and Belief Space Planning

Turetsky¹,

Indelman²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

We investigate the problem of autonomous object classification and semantic SLAM, which in general exhibits a tight coupling between classification, metric SLAM and planning under uncertainty. We contribute a unified framework for inference and belief space planning (BSP) that addresses prominent sources of uncertainty in this context: classification aliasing (classifier cannot distinguish between candidate classes from certain viewpoints), classifier epistemic uncertainty (classifier receives data "far" from its training set), and localization uncertainty (camera and object poses are uncertain). Specifically, we develop two methods for maintaining a joint distribution over robot and object poses, and over posterior class probability vector that consider epistemic uncertainty in a Bayesian fashion. The first approach is Multi-Hybrid (MH), where multiple hybrid beliefs over poses and classes are maintained to approximate the joint belief over poses and posterior class probability. The second approach is Joint Lambda Pose (JLP), where the joint belief is maintained directly using a novel JLP factor. Furthermore, we extend both methods to BSP, planning while reasoning about future posterior epistemic uncertainty indirectly, or directly via a novel information-theoretic reward function. Both inference methods utilize a novel viewpointdependent classifier uncertainty model that leverages the coupling between poses and classification scores, and predicts the epistemic uncertainty from certain viewpoints. In addition, this model is used to generate predicted measurements during planning. To the best of our knowledge, this is the first work that reasons about classifier epistemic uncertainty within semantic SLAM and BSP. We evaluate extensively our inference and BSP approaches in simulation and using real data from the Active Vision Dataset. Results clearly indicate superior classification performance of our methods compared to an approach that is not epistemic uncertainty aware.

show abstract

Section: Introductionmentioning

confidence: 99%

Epistemic Uncertainty Aware Semantic Localization and Mapping for Inference and Belief Space Planning

Turetsky¹,

Indelman²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Epistemic uncertainty aware semantic localization and mapping for inference and belief space planning

Turetsky

Indelman

2023

Artificial Intelligence

View full text Add to dashboard Cite

Enfoque general y sistemático para percepción activa semántica en robótica

Morilla Cabello,

Montijano

2024

JA-CEA

View full text Add to dashboard Cite

En este artículo, abordamos el problema de la percepción activa de información semántica, centrado en determinar las acciones que un robot móvil debe realizar para obtener información semántica de calidad del entorno. Con el auge de los algoritmos de percepción semántica, surgen nuevas oportunidades para los sistemas de planificación robóticos. Sin embargo, para aprovechar estas oportunidades, es crucial identificar los elementos esenciales que cualquier sistema de control orientado a la percepción debe tener. Para ello, proponemos una arquitectura general aplicable a cualquier sistema de percepción activa y analizamos las diferencias fundamentales que surgen al considerar información semántica en su diseño. Además, describimos una implementación preliminar de la arquitectura propuesta. Nuestro objetivo principal es proporcionar a los investigadores una formulación general y un sistema unificado y modular que facilite el avance en el campo de la percepción activa semántica.

show abstract

Spatially-dependent Bayesian semantic perception under model and localization uncertainty

Cited by 3 publications

References 31 publications

Epistemic Uncertainty Aware Semantic Localization and Mapping for Inference and Belief Space Planning

Epistemic Uncertainty Aware Semantic Localization and Mapping for Inference and Belief Space Planning

Epistemic uncertainty aware semantic localization and mapping for inference and belief space planning

Enfoque general y sistemático para percepción activa semántica en robótica

Contact Info

Product

Resources

About