Refine
Department, Institute
Document Type
- Conference Object (19)
- Part of a Book (5)
- Article (2)
- Report (1)
Year of publication
Keywords
- Robot sensing systems (3)
- Mobile robots (2)
- path planning (2)
- AAL-Technik (1)
- Accuracy (1)
- Algorithm design and analysis (1)
- Altenpflege (1)
- Art (1)
- Benchmarking (1)
- Best practice algorithms (1)
In Artificial Intelligence, numerous learning paradigms have been developed over the past decades. In most cases of embodied and situated agents, the learning goal for the artificial agent is to „map“ or classify the environment and the objects therein [1, 2], in order to improve navigation or the execution of some other domain-specific task. Dynamic environments and changing tasks still pose a major challenge for robotic learning in real-world domains. In order to intelligently adapt its task strategies, the agent needs cognitive abilities to more deeply understand its environment and the effects of its actions. In order to approach this challenge within an open-ended learning loop, the XPERO project (http://www.xpero.org) explores the paradigm of Learning by Experimentation to increase the robot's conceptual world knowledge autonomously. In this setting, tasks which are selected by an actionselection mechanism are interrupted by a learning loop in those cases where the robot identifies learning as necessary for solving a task or for explaining observations. It is important to note that our approach targets unsupervised learning, since there is no oracle available to the agent, nor does it have access to a reward function providing direct feedback on the quality of its learned model, as e.g. in reinforcement learning approaches. In the following sections we present our framework for integrating autonomous robotic experimentation into such a learning loop. In section 1 we explain the different modules for stimulation and design of experiments and their interaction. In section 2 we describe our implementation of these modules and how we applied them to a real world scenario to gather target-oriented data for learning conceptual knowledge. There we also indicate how the goaloriented data generation enables machine learning algorithms to revise the failed prediction model.
This paper presents a kinodynamic multi-step motion planner that aims to investigate the influence and mutual dependencies of different kinematic robot configurations and 3D environments. The planner is able to plan the motion for a defined robot that can move along known docking points placed at walls and ceiling. This planner is based on the available multi-step planner of Bretl [1] and adds extended kinodynamic awareness into the planning process. The robot kinematic and dynamic properties, as well as the environment, can be easily changed and thus allows an investigation on the influence of different designs on the movement capabilities. The focus of the planner is the usage as a design tool that allows to compare different robot properties and their resulting influence on the movement capabilities as well the maximum joint forces/torques for desired movements.
In this paper we present work towards the benchmarking of mobile manipulation algorithms. We review the current state-of-the art in mobile manipulation and analyze the most prominent algorithms concerning common structures and sub-components. We propose and implement harmonized interfaces for those components, building upon existing software frameworks and libraries. The foundation on the same subcomponents makes it possible to evaluate mobile manipulation planning algorithms in a systematic way. In particular it enables us to investigate on the influence of different combinations of sub-components for the overall planning task, for which we present experiments in simulation.
This paper presents a novel approach for representing and maintaining a shared 3D world model for robotic applications. This approach is based on the scene graph concept which has been adapted to the requirements of the robotic domain. A key feature is the temporal and centralized sharing of all available 3D data in the leaves of the graph structure. The approach enables tracking of dynamic objects, incorporates uncertainty and allows for annotations by semantic tags. A demonstration is given for a perception application that exploits the temporal sharing of 3D data. A Region of Interest (ROI) is extracted from the stored scene data in order to accelerate processing cycle times.
This paper presents the development of Modelica model for the youBot manipulator. Whereas other robotic simulations focus on the robot interaction with its environment, Modelica allows the modeling of the manipulator controllers and motors. The model was developed with a Modelica library for the manipulator’s components which provides modularity, reusability and abstraction. A comparison test with the actual system has been performed to ensure the model accuracy. The test shows promising result and provides possible future work. The Modelica model of the youBot manipulator is freely available.
With the increasing average age of the population in many developed countries, afflictions like cardiovascular diseases have also increased. Exercising has a proven therapeutic effect on the cardiovascular system and can counteract this development. To avoid overstrain, determining an optimal training dose is crucial. In previous research, heart rate has been shown to be a good measure for cardiovascular behavior. Hence, prediction of the heart rate from work load information is an essential part in models used for training control. Most heart-rate-based models are described in the context of specific scenarios, and have been evaluated on unique datasets only. In this paper, we conduct a joint evaluation of existing approaches to model the cardiovascular system under a certain strain, and compare their predictive performance. For this purpose, we investigated some analytical models as well as some machine learning approaches in two scenarios: prediction over a certain time horizon into the future, and estimation of the relation between work load and heart rate over a whole training session.
Currently, a multitude of studies, whitepapers and roadmaps claim to predict the future of robotics and automation. Unfortunately, all of them are to some extent limited in their scope, by design and/or are biased by some particular interests. In most cases this is not even by intention, but the fact, that the euRobotics Multiannual Roadmap and Strategic Research Agenda only focus on the European robotics community limits it to a certain field of view. The same applies for the U.S. Robotics Roadmap and other documents. In this paper, we suggest the methodology for a Delphi-Study – a guided, unbiased and holistic approach based on neutral interviews with internationally well renowned experts. The necessity for a Delphi-Study has already been laid out by the IEEE RAS Industries Activities Board in the IEEE RAM Magazine [6].
Open ended robotic discovery aims at enabling robots to autonomously design and execute sophisticated experiments for gaining conceptual insight about real world. Such experiments are planned activities rather than innate motor commands and thus each single experiment results in a multivariate time series. In such a scenario, reducing the number of features in order to allow a symbolic learner to build a correct conceptual model of underlying phenomena is a fundamental task. Only few feature selection approaches deal with finding relevant features in multivariate time series, which is just what the robot receives through its sensors. In this paper, we present results of applicability of a range of feature selection and time series analysis approaches on a novel real world scenario for autonomous robotic discovery. We found that even sophisticated representations and state of the art techniques, which perform very well on other benchmarks, do not show significant results in context of open ended discovery.
To achieve autonomous mobile robot navigation, one of fundamental requirements is localization technique. It is difficult for any mobile robot to perform assigned task or sequence of tasks autonomously without accurate position knowledge relative to its environment. There are many existing approaches to accomplish localization in mobile robots but some of the existing mobile robot localization techniques can not be used in large indoor environments because of installation and environmental requirements. Many of existing indoor localization solutions costs significantly more as compared to the solution evaluated in this paper. In this paper a novel passive landmark based indoor position estimation technique is investigated for its robustness, accuracy, repeatability and reliability in large public environment. Use of passive artificial landmarks minimize the disturbance to the user, make the installation process easy, scalable and reduces the total system cost. A comprehensive testing arena is developed at Computer Science Department, University of Applied Sciences Bonn-Rhein-Sieg. This technique has been tested here under different environmental conditions i.e different lighting conditions, different sensor angles, different orientations and different ceiling heights. These tests are designed to subject this technique to possible unexpected and expected conditions that could arise in any large indoor public environment e.g supermarkets, museums etc. This makes our testing and experimental evaluation highly valuable before integrating this technique with mobile service robots at a large scale. This paper experimentally evaluates reliability, robustness and accuracy of this technique over other techniques.
In this paper we focus on the task of automatically and autonomously initiating experimentation and learning based on the recognition of prediction failure. We present a mechanism that utilizes conceptual knowledge to predict the outcome of robot actions, observes their execution and indicates when discrepancies occur. We show how this mechanism was applied to a robot that learns using the paradigm of learning by experimentation, and present first results obtained from this implementation.
An approach to motion planning among moving obstacles is presented, whereby obstacles are modeled as intelligent decision-making agents. The decision-making processes of the obstacles are assumed to be similar to that of the mobile robot. A probabilistic extension to the velocity obstacle approach is used as a means for navigation and modeling uncertainty about the moving obstacles’ decisions.
The positive influence of physical activity for people at all life stages is well known. Exercising has a proven therapeutic effect on the cardiovascular system and can counteract the increase of cardiovascular diseases in our aging society. An easy and good measure of the cardiovascular feedback is the heart rate. Being able to model and predict the response of a subject’s heart rate on work load input allows the development of more advanced smart devices and analytic tools. These tools can monitor and control the subject’s activity and thus avoid overstrain which would eliminate the positive effect on the cardiovascular system. Current heart rate models were developed for a specific scenario and evaluated on unique data sets only. Additionally, most of these models were tested in indoor environments, e.g. on treadmills and bicycle ergometers. However, many people prefer to do sports in outdoors environments and use their smart phone to record their training data. In this paper, we present an evaluation of existing heart rate models and compare their prediction performance for indoor as well as for outdoor running exercises. For this purpose, we investigate analytical models as well as machine learning approaches in two training sets: one indoor exercise set recorded on a treadmill and one outdoor exercise set recorded by a smart phone.
Robots need a representation of their environment to reason about and to interact with it. Different 3D perception and modeling approaches exist to create such a representation, but they are not yet easily comparable. This work tries to identify best practice algorithms in the domain of 3D perception and modeling with a focus on environment reconstruction for robotic applications. The goal is to have a collection of refactored algorithms that are easily measurable and comparable. The realization follows a methodology consisting of five steps. After a survey of relevant algorithms and libraries, common representations for the core data-types Cartesian point, Cartesian point cloud and triangle mesh are identified for use in harmonized interfaces. Atomic algorithms are encapsulated into four software components: the Octree component, the Iterative Closest Point component, the k-Nearest Neighbors search component and the Delaunay triangulation component. A sample experiment demonstrates how the component structure can be used to deduce best practice.
Domestic Robotics
(2016)
In this work a graph-based, semantic mapping approach for indoor robotics applications is presented, which is extending OpenStreetMap (OSM) with robotic-specific, semantic, topological, and geometrical information. Models for common indoor structures (such as walls, doors, corridors, elevators, etc.) are introduced. The architectural principles support composition with additional domain and application specific knowledge. As an example, a model for an area is introduced and it is explained how this can be used in navigation. A key advantages of the proposed graph-based map representation is that it allows seamless transitions between maps, e.g., indoor and outdoor maps by exploiting the hierarchical structure of the graphs. Finally, the compatibility of the approach with existing, grid-based motion planning algorithms is shown.
Stress is necessary for optimal performance and functioning in daily life. However, when stress exceeds person-specific coping levels, then it begins to negatively impact health and productivity. An automatic stress monitoring system that tracks stress levels based on physical and physiological parameters, can assist the user in maintaining stress within healthy limits. In order to build such a system, we need to develop and test various algorithms on a reference dataset consisting of multimodal stress responses. Such a reference dataset should fulfil requirements derived from results and practices of clinical and empirical research. This paper proposes a set of such requirements to support the establishment of a reference dataset for multimodal human stress detection. The requirements cover person-dependent and technical aspects such as selection of sample population, choice of stress stimuli, inclusion of multiple stress modalities, selection of annotation methods, and selection of data acquisition devices. Existing publicly available stress datasets were evaluated based on criteria derived from the proposed requirements. It was found that none of these datasets completely fulfilled the requirements. Therefore, efforts should be made in the future to establish a reference dataset, satisfying the specified requirements, in order to ensure comparability and reliability of results.
Robots generate large amounts of data which need to be stored in a meaningful way such that they can be used and interpreted later. Such data can be written into log files, but these files lack the querying features and scaling capabilities of modern databases - especially when dealing with multi-robot systems, where the trade-off between availability and consistency has to be resolved. However, there is a plethora of existing databases, each with its own set of features, but none designed with robotic use cases in mind. This work presents three main contributions: (a) structures for benchmarking scenarios with a focus on networked multi-robot architectures, (b) an extensible workbench for benchmarking databases for different scenarios that makes use of Docker containers and (c) a comparison of existing databases given a set of multi-robot use cases to showcase the usage of the framework. The comparison gives indications for choosing an appropriate database.
Robot deployment in realistic environments is challenging despite the fact that robots can be quite skilled at a large number of isolated tasks. One reason for this is that robots are rarely equipped with powerful introspection capabilities, which means that they cannot always deal with failures in an acceptable manner; in addition, manual diagnosis is often a tedious task that requires technicians to have a considerable set of robotics skills. In this paper, we discuss our ongoing efforts to address some of these problems. In particular, we (i) present our early efforts at developing a robotic black box and consider some factors that complicate its design, (ii) explain our component and system monitoring concept, and (iii) describe the necessity for remote monitoring and experimentation as well as our initial attempts at performing those. Our preliminary work opens a range of promising directions for making robots more usable and reliable in practice.
Robot deployment in realistic dynamic environments is a challenging problem despite the fact that robots can be quite skilled at a large number of isolated tasks. One reason for this is that robots are rarely equipped with powerful introspection capabilities, which means that they cannot always deal with failures in a reasonable manner; in addition, manual diagnosis is often a tedious task that requires technicians to have a considerable set of robotics skills.
