From WWWwikiEn<span style="display: none;">[[Image:SDF11_Huber]][[Image:IROS11_Gehrig]][[Image:Fusion11_Krauthausen]][[Image:Fusion11_Ruoff]][[Image:ACC11_Krauthausen]][[Image:IROS10_Krauthausen]][[Image:KI10_Krauthausen.pdf]][[Image:MFI10_KrauthausenEberhardt.pdf]][[Image:MFI10_Krauthausen.pdf]][[Image:Fusion10_KrauthausenHuber]][[Image:ACC10_Krauthausen.pdf]][[Image:Fusion09_Krauthausen.pdf]]</span> <span style="display: none;">[[Image:SDF11_Huber]][[Image:IROS11_Gehrig]][[Image:Fusion11_Krauthausen]][[Image:Fusion11_Ruoff]][[Image:ACC11_Krauthausen]][[Image:IROS10_Krauthausen]][[Image:KI10_Krauthausen.pdf]][[Image:MFI10_KrauthausenEberhardt.pdf]][[Image:MFI10_Krauthausen.pdf]][[Image:Fusion10_KrauthausenHuber]][[Image:ACC10_Krauthausen.pdf]][[Image:Fusion09_Krauthausen.pdf]]</span>

Peter Krauthausen

	Dipl.-Inform. Research Assistant, Ph.D. Student
	Address:	Karlsruher Institut für TechnologieInstitut für AnthropomatikGebäude 50.20Raum 130Adenauerring 2D-76131 Karlsruhe
	Walk-in hours:	on arrangement
	Phone:	+49-721-608-46340
	E-mail:	Peter.Krauthausen@kit.edu

Academic Career

Research Interests

• Statistical learning theory,
• Machine learning,
• System and estimator theory.

My research is performed as part of the Collaborative Research Center 588 "Humanoid Robots - Learning and Cooperating Multimodal Robots" in the field of intention recognition funded by the Deutsche Forschungsgemeinschaft (DFG).

Teaching

• Assistant and coordinator for the lecture Stochastische Informationsverarbeitung (Probabilistic Information Processing) (winter terms 2007/2008, 2008/2009, 2009/2010 and 2010/2011),
• Supervision of small groups of students in the Praktikum: Forschungsprojekt "Intelligente Sensor-Aktor-Systeme" laboratory (winter term 2007/2008 to summer term 2009).

Publications

Proceedings of the IEEE ISIF Workshop on Sensor Data Fusion: Trends, Solutions, Applications (SDF 2011), Berlin, Germany, October, 2011.

Author : Marco Huber, Peter Krauthausen, Uwe D. Hanebeck
Title : Superficial Gaussian Mixture Reduction
In : Proceedings of the IEEE ISIF Workshop on Sensor Data Fusion: Trends, Solutions, Applications (SDF 2011)
Date : October 2011

Abstract

Many information fusion tasks involve the processing of Gaussian mixtures
with simple underlying shape, but many components. This paper addresses the problem
of reducing the number of components, allowing for faster density processing.
The proposed approach is based on identifying components irrelevant for the overall
density\'s shape by means of the curvature of the density\'s surface. The key idea is to
minimize an upper bound of the curvature while maintaining a low global reduction
error by optimizing the weights of the original Gaussian mixture only. The mixture is
reduced by assigning zero weights to reducible components. The main advantages are
an alleviation of the model selection problem, as the number of components is chosen
by the algorithm automatically, the derivation of simple curvature-based penalty terms,
and an easy, efficient implementation. A series of experiments shows the approach to
provide a good trade-off between quality and sparsity.

Proceedings of the 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2011), San Francisco, California, USA, September, 2011.

Author : Dirk Gehrig, Peter Krauthausen, Lukas Rybok, Hildegard Kühne, Tanja Schultz, Uwe D. Hanebeck, Rainer Stiefelhagen
Title : Combined Multi-Level Intention, Activity, and Motion Recognition for a Humanoid Robot
In : Proceedings of the 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2011)
Date : September 2011

Abstract

In this paper, a multi-level approach to intention,
activity, and motion recognition for a humanoid robot is
proposed. Our system processes images from a monocular
camera and combines this information with domain knowledge.
The recognition works on-line and in real-time, it is independent
of the test person, but limited to predefined view-points.
Main contributions of this paper are the extensible, multi-level
modeling of the robot\'s vision system, the efficient activity and
motion recognition, and the asynchronous information fusion
based on generic processing of mid-level recognition results. The
complementarity of the activity and motion recognition renders
the approach robust against misclassifications. Experimental
results on a real-world data set of complex kitchen tasks,
e.g., Prepare Cereals or Lay Table, prove the performance and
robustness of the multi-level recognition approach.

Proceedings of the 14th International Conference on Information Fusion (Fusion 2011), Chicago, Illinois, USA, July, 2011.

Author : Peter Krauthausen, Patrick Ruoff, Uwe D. Hanebeck
Title : Sparse Mixture Conditional Density Estimation by Superficial Regularization
In : Proceedings of the 14th International Conference on Information Fusion (Fusion 2011)
Date : July 2011

Abstract

In this paper, the estimation of conditional densities between continuous random variables from noisy samples is considered.
The conditional densities are modeled as heteroscedastic Gaussian mixture densities allowing for closed-form solution of Bayesian inference with full-densities.
The main contributions of this paper are an improved generalization quality of the estimates by the introduction of a superficial regularizer, the consideration
of model uncertainty relative to local data densities by means of adaptive covariances, and the proposition of an efficient distance-based estimation algorithm.
This algorithm corresponds to an iterative nested optimization scheme, optimizing hyper-parameters, component placement, and mixture weights.
The obtained solutions are sparse, smooth, and generalize well as benchmark experiments, e.g., in nonlinear filtering show.

Proceedings of the 14th International Conference on Information Fusion (Fusion 2011), Chicago, Illinois, USA, July, 2011.

Author : Patrick Ruoff, Peter Krauthausen,, Uwe D. Hanebeck
Title : Progressive Correction for Deterministic Dirac Mixture Approximations
In : Proceedings of the 14th International Conference on Information Fusion (Fusion 2011)
Date : July 2011

Abstract

Since the advent of Monte-Carlo particle filtering, particle representations of densities have become increasingly popular due to their flexibility and implicit adaptive resolution.
In this paper, an algorithm for the multiplication of a systematic Dirac mixture (DM) approximation with a continuous likelihood function is presented,
which applies a progressive correction scheme, in order to avoid the particle degeneration problem.
The preservation of sample regularity and therefore, representation quality of the underlying smooth density, is ensured by including
a new measure of smoothness for Dirac mixtures, the DM energy, into the distance measure.
A comparison to common correction schemes in Monte-Carlo methods reveals large improvements especially in cases of small overlap between the likelihood and prior density,
as well as for multi-modal likelihoods.

Proceedings of the 2011 American Control Conference (ACC 2011), San Francisco, California, USA, June, 2011.

Author : Peter Krauthausen, Masoud Roschani, Uwe D. Hanebeck
Title : Incorporating Prior Knowledge into Nonparametric Conditional Density Estimation
In : Proceedings of the 2011 American Control Conference (ACC 2011)
Date : June 2011

Abstract

In this paper, the problem of sparse nonparametric conditional density estimation based on samples and prior knowledge is addressed.
The prior knowledge may be restricted to parts of the state space and given as generative models in form of mean-function constraints
or as probabilistic models in the form of Gaussian mixtures.
The key idea is the introduction of additional constraints and a modified kernel function into the conditional density estimation problem.
This approach to using prior knowledge is a generic solution applicable to all nonparametric conditional density estimation approaches phrased as constrained optimization problems.
The quality of the estimates, their sparseness, and the achievable improvements by using prior knowledge are shown in experiments for both Support-Vector Machine-based and integral distance-based conditional density estimation.

Proceedings of the 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2010), Taipei, Taiwan, October, 2010.

Author : Peter Krauthausen, Uwe D. Hanebeck
Title : A Model-Predictive Switching Approach To Efficient Intention Recognition
In : Proceedings of the 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2010)
Date : October 2010

Abstract

Estimating a user’s intention is central to close
human-robot cooperation. In this paper, the problem of per-
forming intention recognition with tree-structured Dynamic
Bayesian Networks for large environments with many features
is addressed. The proposed approach reduces the computational
complexity of inference O(b^s) for tree-structured measurement
models with an average branching factor b and tree height s
to O((b˜)s), where b << b. The key idea is to switch between a
finite set of reduced system and measurement models in order
to restrict inference to the most important features. A model
predictive approach to online switching between the reduced
models is proposed that exploits an upper bound of the distances
of the reduced models to the full model. The effectiveness of
the proposed algorithm is validated in the intention recognition
for a humanoid robot using a telepresent household scenario.

33rd Annual German Conference on Artificial Intelligence (KI 2010), Karlsruhe, Germany, September, 2010.

Author : Peter Krauthausen, Uwe D. Hanebeck
Title : Situation-Specific Intention Recognition for Human-Robot-Cooperation
In : 33rd Annual German Conference on Artificial Intelligence (KI 2010)
Date : September 2010

Abstract

Recognizing human intentions is part of the decision
process in many technical devices. In order to achieve
natural interaction, the required estimation quality and
the used computation time need to be balanced. This becomes
challenging, if the number of sensors is high and measurement
systems are complex. In this paper, a model predictive approach
to this problem based on online switching of small,
situation-specific Dynamic Bayesian Networks is proposed.
The contributions are an efficient modeling and inference
of situations and a greedy model predictive switching algorithm
maximizing the mutual information of predicted situations. The
achievable accuracy and computational savings are demonstrated
for a household scenario by using an extended range telepresence system.

Proceedings of the 2010 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI 2010), Salt Lake City, Utah, USA, September, 2010.

Author : Peter Krauthausen, Henning Eberhardt, Uwe D. Hanebeck
Title : Multivariate Parametric Density Estimation Based On The Modified Cramér-von Mises Distance
In : Proceedings of the 2010 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI 2010)
Date : September 2010

Abstract

In this paper, a novel distance-based density
estimation method is proposed, which considers the overall
density function in the goodness-of-fit. In detail, the parameters
of Gaussian mixture densities are estimated from samples,
based on the distance of the cumulative distributions over
the entire state space. Due to the ambiguous definition of the
standard multivariate cumulative distribution, the Localized
Cumulative Distribution and a modified Cramér-von Mises
distance measure are employed. A further contribution is the
derivation of a simple-to-implement optimization procedure
for the optimization problem. The proposed approach’s good
performance in estimating arbitrary Gaussian mixture densities
is shown in an experimental comparison to the Expectation
Maximization algorithm for Gaussian mixture densities.

Proceedings of the 2010 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI 2010), Salt Lake City, Utah, USA, September, 2010.

Author : Peter Krauthausen, Uwe D. Hanebeck
Title : Regularized Non-Parametric Multivariate Density and Conditional Density Estimation
In : Proceedings of the 2010 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI 2010)
Date : September 2010

Abstract

In this paper, a distance-based method for both
multivariate non-parametric density and conditional density
estimation is proposed. The contributions are the formulation
of both density estimation problems as weight optimization
problems for Gaussian mixtures centered about samples with
identical parameters. Furthermore, the minimization is based
on the modified Cramér-von Mises distance of the Localized
Cumulative Distributions, removing the ambiguity of the defi-
nition of the multivariate cumulative distribution function. The
minimization problem is amended with a regularization term
penalizing the densities’roughness to avoid overfitting. The
resulting estimation problems for both densities and conditional
densities are shown to be phrasable in the form of readily
implementable quadratic programs. Experimental comparison
against EM, SVR, and GPR based on the log-likelihood and
performance in benchmark recursive filtering applications show
high quality of the densities and good performance at less
computational cost, i.e., the density representations are sparser.

Proceedings of the 13th International Conference on Information Fusion (Fusion 2010), Edinburgh, United Kingdom, July, 2010.

Author : Peter Krauthausen, Marco F. Huber, Uwe D. Hanebeck
Title : Support-Vector Conditional Density Estimation for Nonlinear Filtering
In : Proceedings of the 13th International Conference on Information Fusion (Fusion 2010)
Date : July 2010

Abstract

A non-parametric conditional density
estimation algorithm for nonlinear stochastic dynamic
systems is proposed. The contributions are a novel sup-
port vector regression for estimating conditional den-
sities, modeled by Gaussian mixture densities, and an
algorithm based on cross-validation for automatically
determining hyper-parameters for the regression. The
conditional densities are employed with a modified axis-
aligned Gaussian mixture filter. The experimental va-
lidation shows the high quality of the conditional densi-
ties and good accuracy of the proposed filter.

Proceedings of the 2010 American Control Conference (ACC 2010), Baltimore, Maryland, USA, June, 2010.

Author : Peter Krauthausen, Uwe D. Hanebeck
Title : Parameter Learning for Hybrid Bayesian Networks With Gaussian Mixture and Dirac Mixture Conditional Densities
In : Proceedings of the 2010 American Control Conference (ACC 2010)
Date : June 2010

Abstract

In this paper, the first algorithm for learning hybrid Bayesian
Networks with Gaussian mixture and Dirac mixture conditional densities from data
given their structure is presented. The mixture densities to be learned allow for
nonlinear dependencies between the variables and exact closedform inference. For
learning the network’s parameters, an incremental gradient ascent algorithm is derived.
Analytic expressions for the partial derivatives and their combination with messages are
presented. This hybrid approach subsumes the existing approach for purely discrete-valued
networks and is applicable to partially observable networks, too. Its practicability is
demonstrated by a reference example.

Proceedings of the 12th International Conference on Information Fusion (Fusion 2009), Seattle, Washington, USA, July, 2009.

Author : Peter Krauthausen, Uwe D. Hanebeck
Title : Intention Recognition for Partial-Order Plans Using Dynamic Bayesian Networks
In : Proceedings of the 12th International Conference on Information Fusion (Fusion 2009)
Date : July 2009

Abstract

In this paper, a novel probabilistic ap-
proach to intention recognition for partial-order plans
is proposed. The key idea is to exploit independences
between subplans to substantially reduce the state space
sizes in the compiled Dynamic Bayesian Networks.
This makes inference more efficient. The main con-
tributions are the computationally exploitable definition
of subplan structures, the introduction of a novel Lay-
ered Intention Model and a Dynamic Bayesian Net-
work representation with an inference mechanism that
exploits consecutive and concurrent subplans\' indepen-
dences. The presented approach reduces the state space
to the order of the most complex subplan and requires
only minor changes in the standard inference mecha-
nism. The practicability of this approach is demon-
strated by recognizing the process of shelf-assembly.

Peter Krauthausen, Astrid Laubenheimer,
A Comparative Study of Decision Tree Approaches to Multi-Class Support Vector Machines,
Proceedings of Artificial Intelligence and Applications (AIA 2008), Innsbruck, Austria, February, 2008.

Peter Krauthausen, Alexander Kipp, Frank Dellaert,
Exploiting Locality in SLAM by Nested Dissection,
Proceedings of Robotics: Science and Systems (RSS II), Philadelphia, USA, August, 2006.

Frank Dellaert, Alexander Kipp, Peter Krauthausen,
A Multifrontal QR Factorization Approach to Distributed Inference Applied to Multi-Robot Localization and Mapping,
Proceedings of the Twentieth National Conference on Artificial Intelligence (AAAI 2005), Pittsburgh, Pennsylvania, July, 2005.