Learning GP-BayesFilters via Gaussian process latent variable models

Ko, Jonathan; Fox, Dieter

doi:10.1007/s10514-010-9213-0

Cited by 61 publications

(60 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…GP latent variable models (Lawrence 2004;Wang et al 2006) were developed to handle cases in which no ground truth is available for GP input values. They are thus well suited for cases with uncertain training data, and their extension to learning GPBayesFilters has recently been investigated with extremely promising results (Ko and Fox 2009). Finally, the predictive uncertainty of the GPs used in this paper assumed diagonal error covariance matrices.…”

Section: Discussionmentioning

confidence: 99%

GP-BayesFilters: Bayesian filtering using Gaussian process prediction and observation models

2009

Self Cite

View full text Add to dashboard Cite

Bayesian filtering is a general framework for recursively estimating the state of a dynamical system. Key components of each Bayes filter are probabilistic prediction and observation models. This paper shows how non-parametric Gaussian process (GP) regression can be used for learning such models from training data. We also show how Gaussian process models can be integrated into different versions of Bayes filters, namely particle filters and extended and unscented Kalman filters. The resulting GP-BayesFilters can have several advantages over standard (parametric) filters. Most importantly, GP-BayesFilters do not require an accurate, parametric model of the system. Given enough training data, they enable improved tracking accuracy compared to parametric models, and they degrade gracefully with increased model uncertainty. These advantages stem from the fact that GPs consider both the noise in the system and the uncertainty in the model. If an approximate parametric model is available, it can be incorporated into the GP, resulting in further performance improvements. In experiments, we show different properties of GP-BayesFilters using data collected with an autonomous micro-blimp as well as synthetic data.

show abstract

Section: Discussionmentioning

confidence: 99%

GP-BayesFilters: Bayesian filtering using Gaussian process prediction and observation models

2009

Self Cite

View full text Add to dashboard Cite

show abstract

“…More informed dynamics models will help improve the accuracy of the reconstructed node locations and the tracking accuracy after reconstruction. The use of the recently developed GP-Bayes filters (Ko and Fox 2011) would also allow for better incorporation of motion.…”

Section: Discussionmentioning

confidence: 99%

Target tracking without line of sight using range from radio

2011

View full text Add to dashboard Cite

We propose a framework for utilizing fixed ultrawideband ranging radio nodes to track a moving target radio node in an environment without guaranteed line of sight or accurate odometry. For the case where the fixed nodes' locations are known, we derive a Bayesian room-level tracking method that takes advantage of the structural characteristics of the environment to ensure robustness to noise. For the case of unknown fixed node locations, we present a twostep approach that first reconstructs the target node's path using Gaussian Process Latent Variable models (GPLVMs) and then uses that path to determine the locations of the fixed nodes. We present experiments verifying our algorithm in an office environment, and we compare our results to those generated by online and batch SLAM methods, as well as odometry mapping. Our algorithm is successful at tracking a moving target node without odometry and mapping the locations of fixed nodes using radio ranging data that are both noisy and intermittent. Electronic supplementary materialThe online version of this article (

show abstract

“…Finally, learning synergistic motion models that enable reasoning about redundancy in the state spaces of interacting objects, hands and body parts, could improve tracking performance of both the observed and occluded parts of the articulated models. Very promising results in that direction have been presented by Schröder et al (2013), and the incorporation of more complex, non-linear models such as Gaussian process latent variable models (Ko and Fox 2011;Damianou et al 2011) might yield even further improvements. In g the points observed on the tip of the pinky finger have been data associated to the ring finger, while the points observed on the ring finger have been data associated to the middle finger, meaning there is no gradient to pull the ring finger back into the correct location.…”

Section: Limitations and Possible Extensionsmentioning

confidence: 99%

DART: dense articulated real-time tracking with consumer depth cameras

2015

Self Cite

View full text Add to dashboard Cite

This paper introduces DART, a general framework for tracking articulated objects composed of rigid bodies connected through a kinematic tree. DART covers a broad set of objects encountered in indoor environments, including furniture and tools, and human and robot bodies, hands and manipulators. To achieve efficient and robust tracking, DART extends the signed distance function representation to articulated objects and takes full advantage of highly parallel GPU algorithms for data association and pose optimization. We demonstrate the capabilities of DART on different types of objects that have each required dedicated tracking techniques in the past.

show abstract

Learning GP-BayesFilters via Gaussian process latent variable models

Cited by 61 publications

References 22 publications

GP-BayesFilters: Bayesian filtering using Gaussian process prediction and observation models

GP-BayesFilters: Bayesian filtering using Gaussian process prediction and observation models

Target tracking without line of sight using range from radio

DART: dense articulated real-time tracking with consumer depth cameras

Contact Info

Product

Resources

About