Erik Schaffernicht

Multi-modal sensor fusion using a probabilistic aggregation scheme for people detection and tracking

Efficient and robust techniques for people detection and tracking are basic prerequisites when dealing with Human‐Robot Interaction (HRI) in real-world scenarios. In this paper, we introduce a new approach for the integration of several sensor modalities and present a multi-modal, probability-based people detection and tracking system and its application using the different sensory systems of our mobile interaction robot HOROS. These include a laser range-finder, a sonar system, and a fisheye-based omni-directional camera. For each of these sensory systems, separate and specific Gaussian probability distributions are generated to model the belief in observing one or several persons. These probability distributions are further merged into a robot-centered map by means of a flexible probabilistic aggregation scheme based on Covariance Intersection (CI). The main advantages of this approach are the simple extensibility by the integration of further sensory channels, even with different update frequencies, and the usability in real-world HRI tasks. Finally, the first promising experimental results achieved for people detection and tracking in a real-world environment (our institute building) are presented. c 2006 Elsevier B.V. All rights reserved.

Combining Non Selective Gas Sensors on a Mobile Robot for Identification and Mapping of Multiple Chemical Compounds

In this paper, we address the task of gas distribution modeling in scenarios where multiple heterogeneous compounds are present. Gas distribution modeling is particularly useful in emission monitoring applications where spatial representations of the gaseous patches can be used to identify emission hot spots. In realistic environments, the presence of multiple chemicals is expected and therefore, gas discrimination has to be incorporated in the modeling process. The approach presented in this work addresses the task of gas distribution modeling by combining different non selective gas sensors. Gas discrimination is addressed with an open sampling system, composed by an array of metal oxide sensors and a probabilistic algorithm tailored to uncontrolled environments. For each of the identified compounds, the mapping algorithm generates a calibrated gas distribution model using the classification uncertainty and the concentration readings acquired with a photo ionization detector. The meta parameters of the proposed modeling algorithm are automatically learned from the data. The approach was validated with a gas sensitive robot patrolling outdoor and indoor scenarios, where two different chemicals were released simultaneously. The experimental results show that the generated multi compound maps can be used to accurately predict the location of emitting gas sources.

On estimating mutual information for feature selection

Mutual Information (MI) is a powerful concept from information theory used in many application fields. For practical tasks it is often necessary to estimate the Mutual Information from available data. We compare state of the art methods for estimating MI from continuous data, focusing on the usefulness for the feature selection task. Our results suggest that many methods are practically relevant for feature selection tasks regardless of their theoretic limitations or benefits.

Robot assisted gas tomography — Localizing methane leaks in outdoor environments

In this paper we present an inspection robot to produce gas distribution maps and localize gas sources in large outdoor environments. The robot is equipped with a 3D laser range finder and a remote gas sensor that returns integral concentration measurements. We apply principles of tomography to create a spatial gas distribution model from integral gas concentration measurements. The gas distribution algorithm is framed as a convex optimization problem and it models the mean distribution and the fluctuations of gases. This is important since gas dispersion is not an static phenomenon and furthermore, areas of high fluctuation can be correlated with the location of an emitting source. We use a compact surface representation created from the measurements of the 3D laser range finder with a state of the art mapping algorithm to get a very accurate localization and estimation of the path of the laser beams. In addition, a conic model for the beam of the remote gas sensor is introduced. We observe a substantial improvement in the gas source localization capabilities over previous state-of-the-art in our evaluation carried out in an open field environment.

Weighted mutual information for feature selection

In this paper, we apply weighted Mutual Information for effective feature selection. The presented hybrid filter wrapper approach resembles the well known AdaBoost algorithm by focusing on those samples that are not classified or approximated correctly using the selected features. Redundancies and bias of the employed learning machine are handled implicitly by our approach. In experiments, we compare the weighted Mutual Information algorithm with other basic approaches for feature subset selection that use similar selection criteria. The efficiency and effectiveness of our method are demonstrated by the obtained results.

Whom to talk to? Estimating user interest from movement trajectories

Correctly identifying people who are interested in an interaction with a mobile robot is an essential task for a smart Human-Robot Interaction. In this paper an approach is presented for selecting suitable trajectory features in a task specific manner from a huge amount of different forms of possible representations. Different sub-sampling techniques are proposed to generate trajectory sequences from which features are extracted. The trajectory data was generated in real world experiments that include extensive user interviews to acquire information about user behaviors and intentions. Using those feature vectors in a classification method enables the robot to estimate the users interaction interest. For generating low-dimensional feature vectors, a common method, the Principle Component Analysis, is applied. The selection and combination of useful features out of a set of possible features is carried out by an information theoretic approach based on the Mutual Information and Joint Mutual Information with respect to the users interaction interest. The introduced procedure is evaluated with neural classifiers, which are trained with the extracted features of the trajectories and the user behavior gained by observation as well as user interviewing. The results achieved indicate that an estimation of the users interaction interest using trajectory information is feasible.

An efficient search strategy for feature selection using Chow-Liu trees

Within the taxonomy of feature extraction methods, recently the Wrapper approaches lost some popularity due to the associated computational burden, compared to Embedded or Filter methods. The dominating factor in terms of computational costs is the number of adaption cycles used to train the black box classifier or function approximator, e.g. a Multi Layer Perceptron. To keep a wrapper approach feasible, the number of adaption cycles has to be minimized, without increasing the risk of missing important feature subset combinations. We propose a search strategy, that exploits the interesting properties of Chow-Liu trees to reduce the number of considered subsets significantly. Our approach restricts the candidate set of possible new features in a forward selection step to children from certain tree nodes. We compare our algorithm with some basic and well known approaches for feature subset selection. The results obtained demonstrate the efficiency and effectiveness of our method.

A probabilistic multimodal sensor aggregation scheme applied for a mobile robot

Dealing with methods of human-robot interaction and using a real mobile robot, stable methods for people detection and tracking are fundamental features of such a system and require information from different sensory. In this paper, we discuss a new approach for integrating several sensor modalities and we present a multimodal people detection and tracking system and its application using the different sensory systems of our mobile interaction robot Horos working in a real office environment. These include a laser-range-finder, a sonar system, and a fisheye-based omnidirectional camera. For each of these sensory information, a separate Gaussian probability distribution is generated to model the belief of the observation of a person. These probability distributions are further combined using a flexible probabilistic aggregation scheme. The main advantages of this approach are a simple integration of further sensory channels, even with different update frequencies and the usability in real-world environments. Finally, promising experimental results achieved in a real office environment will be presented.

Global Coverage Measurement Planning Strategies for Mobile Robots Equipped with a Remote Gas Sensor

The problem of gas detection is relevant to many real-world applications, such as leak detection in industrial settings and landfill monitoring. In this paper, we address the problem of gas detection in large areas with a mobile robotic platform equipped with a remote gas sensor. We propose an algorithm that leverages a novel method based on convex relaxation for quickly solving sensor placement problems, and for generating an efficient exploration plan for the robot. To demonstrate the applicability of our method to real-world environments, we performed a large number of experimental trials, both on randomly generated maps and on the map of a real environment. Our approach proves to be highly efficient in terms of computational requirements and to provide nearly-optimal solutions.

A novel approach for gas discrimination in natural environments with Open Sampling Systems

This work presents a gas discrimination approach for Open Sampling Systems (OSS), composed of non-specific metal oxide sensors only. In an OSS, as used on robots or in sensor networks, the sensors are exposed to the dynamics of the environment and thus, most of the data corresponds to highly diluted samples while high concentrations are sparse. In addition, a positive correlation between class separability and concentration level can be observed. The proposed approach computes the class posteriors by coupling the pairwise probabilities between the compounds to a confidence model based on an estimation of the concentration. In this way a rejection posterior, analogous to the detection limit of the human nose, is learned. Evaluation was conducted in indoor and outdoor sites, with an OSS equipped robot, in the presence of two gases. The results show that the proposed approach achieves a high classification performance with a low sensitivity to the selection of meta parameters.