Refine
H-BRS Bibliography
- yes (86)
Departments, institutes and facilities
- Fachbereich Informatik (86) (remove)
Document Type
- Conference Object (34)
- Article (20)
- Preprint (12)
- Report (7)
- Research Data (3)
- Doctoral Thesis (3)
- Book (monograph, edited volume) (2)
- Contribution to a Periodical (2)
- Master's Thesis (2)
- Working Paper (1)
Year of publication
- 2020 (86) (remove)
Keywords
- Quality diversity (3)
- Usable Security (3)
- post-buckling (3)
- ARIMA (2)
- Autoencoder (2)
- Automatic Short Answer Grading (2)
- Bayesian optimization (2)
- Black-Box Optimization (2)
- Deep Learning (2)
- Domestic Robots (2)
Graph drawing with spring embedders employs a V x V computation phase over the graph's vertex set to compute repulsive forces. Here, the efficacy of forces diminishes with distance: a vertex can effectively only influence other vertices in a certain radius around its position. Therefore, the algorithm lends itself to an implementation using search data structures to reduce the runtime complexity. NVIDIA RT cores implement hierarchical tree traversal in hardware. We show how to map the problem of finding graph layouts with force-directed methods to a ray tracing problem that can subsequently be implemented with dedicated ray tracing hardware. With that, we observe speedups of 4x to 13x over a CUDA software implementation.
This paper introduces FaceHaptics, a novel haptic display based on a robot arm attached to a head-mounted virtual reality display. It provides localized, multi-directional and movable haptic cues in the form of wind, warmth, moving and single-point touch events and water spray to dedicated parts of the face not covered by the head-mounted display.The easily extensible system, however, can principally mount any type of compact haptic actuator or object. User study 1 showed that users appreciate the directional resolution of cues, and can judge wind direction well, especially when they move their head and wind direction is adjusted dynamically to compensate for head rotations. Study 2 showed that adding FaceHaptics cues to a VR walkthrough can significantly improve user experience, presence, and emotional responses.
Risk-based Authentication (RBA) is an adaptive security measure that improves the security of password-based authentication by protecting against credential stuffing, password guessing, or phishing attacks. RBA monitors extra features during login and requests for an additional authentication step if the observed feature values deviate from the usual ones in the login history. In state-of-the-art RBA re-authentication deployments, users receive an email with a numerical code in its body, which must be entered on the online service. Although this procedure has a major impact on RBA's time exposure and usability, these aspects were not studied so far.
We introduce two RBA re-authentication variants supplementing the de facto standard with a link-based and another code-based approach. Then, we present the results of a between-group study (N=592) to evaluate these three approaches. Our observations show with significant results that there is potential to speed up the RBA re-authentication process without reducing neither its security properties nor its security perception. The link-based re-authentication via "magic links", however, makes users significantly more anxious than the code-based approaches when perceived for the first time. Our evaluations underline the fact that RBA re-authentication is not a uniform procedure. We summarize our findings and provide recommendations.
Risk-based Authentication (RBA) is an adaptive security measure to strengthen password-based authentication. RBA monitors additional features during login, and when observed feature values differ significantly from previously seen ones, users have to provide additional authentication factors such as a verification code. RBA has the potential to offer more usable authentication, but the usability and the security perceptions of RBA are not studied well.
We present the results of a between-group lab study (n=65) to evaluate usability and security perceptions of two RBA variants, one 2FA variant, and password-only authentication. Our study shows with significant results that RBA is considered to be more usable than the studied 2FA variants, while it is perceived as more secure than password-only authentication in general and comparably secure to 2FA in a variety of application types. We also observed RBA usability problems and provide recommendations for mitigation. Our contribution provides a first deeper understanding of the users' perception of RBA and helps to improve RBA implementations for a broader user acceptance.
Human and robot tasks in household environments include actions such as carrying an object, cleaning a surface, etc. These tasks are performed by means of dexterous manipulation, and for humans, they are straightforward to accomplish. Moreover, humans perform these actions with reasonable accuracy and precision but with much less energy and stress on the actuators (muscles) than the robots do. The high agility in controlling their forces and motions is actually due to "laziness", i.e. humans exploit the existing natural forces and constraints to execute the tasks.
The above-mentioned properties of the human lazy strategy motivate us to relax the problem of controlling robot motions and forces, and solve it with the help of the environment. Therefore, in this work, we developed a lazy control strategy, i.e. task specification models and control architectures that relax several aspects of robot control by exploiting prior knowledge about the task and environment. The developed control strategy is realized in four different robotics use cases. In this work, the Popov-Vereshchagin hybrid dynamics solver is used as one of the building blocks in the proposed control architectures. An extension of the solver’s interface with the artificial Cartesian force and feed-forward joint torque task-drivers is proposed in this thesis.
To validate the proposed lazy control approach, an experimental evaluation was performed in a simulation environment and on a real robot platform.
Dieses Buch bietet einen leicht verständlichen Einstieg in die Thematik des Data Minings und der Prädiktiven Analyseverfahren. Als Methodensammlung gedacht, bietet es zu jedem Verfahren zunächst eine kurze Darstellung der Theorie und erklärt die zum Verständnis notwendigen Formeln. Es folgt jeweils eine Illustration der Verfahren mit Hilfe von Beispielen, die mit dem Programmpaket R erarbeitet werden.
Zum Abschluss wird eine einfache Möglichkeit präsentiert, mit der die Performancewerte verschiedener Verfahren mit statistischen Mitteln verglichen werden können. Zum Einsatz kommen hierbei geeignete Grafiken und Konfidenzintervalle.
Das Buch verzichtet nicht auf Theorie, es präsentiert jedoch so wenig Theorie wie möglich, aber so viel wie nötig und ist somit optimal für Studium und Selbststudium geeignet.
The ongoing digitisation in everyday working life means that ever larger amounts of personal data of employees are processed by their employers. This development is particularly problematic with regard to employee data protection and the right to informational self-determination. We strive for the use of company Privacy Dashboards as a means to compensate for missing transparency and control. For conceptual design we use among other things the method of mental models. We present the methodology and first results of our research. We highlight the opportunities that such an approach offers for the user-centred development of Privacy Dashboards.
This paper presents groupware to study group behavior while conducting a creative task on large, high-resolution displays. Moreover, we present the results of a between-subjects study. In the study, 12 groups with two participants each prototyped a 2D level on a 7m x 2.5m large, high-resolution display using tablet-PCs for interaction. Six groups underwent a condition where group members had equal roles and interaction possibilities. Another six groups worked in a condition where group members had different roles: level designer and 2D artist. The results revealed that in the different roles condition, the participants worked significantly more tightly and created more assets. We could also detect some shortcomings for that configuration. We discuss the gained insights regarding system configuration, groupware interfaces, and groups behavior.
Gone But Not Forgotten: Evaluating Performance and Scalability of Real-Time Mesoscopic Agents
(2020)
Evaluation of a Multi-Layer 2.5D display in comparison to conventional 3D stereoscopic glasses
(2020)
In this paper we propose and evaluate a custom-build projection-based multilayer 2.5D display, consisting of three layers of images, and compare performance to a stereoscopic 3D display. Stereoscopic vision can increase the involvement and enhance game experience, however may induce possible side effects, e.g. motion sickness and simulator sickness. To overcome the disadvantage of multiple discrete depths, in our system perspective rendering and head-tracking is used. A study was performed to evaluate this display with 20 participants playing custom-designed games. The results indicated that the multi-layer display caused fewer side effects than the stereoscopic display and provided good usability. The participants also stated a better or equal spatial perception, while the cognitive load stayed the same.
Modern Monte-Carlo-based rendering systems still suffer from the computational complexity involved in the generation of noise-free images, making it challenging to synthesize interactive previews. We present a framework suited for rendering such previews of static scenes using a caching technique that builds upon a linkless octree. Our approach allows for memory-efficient storage and constant-time lookup to cache diffuse illumination at multiple hitpoints along the traced paths. Non-diffuse surfaces are dealt with in a hybrid way in order to reconstruct view-dependent illumination while maintaining interactive frame rates. By evaluating the visual fidelity against ground truth sequences and by benchmarking, we show that our approach compares well to low-noise path-traced results, but with a greatly reduced computational complexity, allowing for interactive frame rates. This way, our caching technique provides a useful tool for global illumination previews and multi-view rendering.
Modern Monte-Carlo-based rendering systems still suffer from the computational complexity involved in the generation of noise-free images, making it challenging to synthesize interactive previews. We present a framework suited for rendering such previews of static scenes using a caching technique that builds upon a linkless octree. Our approach allows for memory-efficient storage and constant-time lookup to cache diffuse illumination at multiple hitpoints along the traced paths. Non-diffuse surfaces are dealt with in a hybrid way in order to reconstruct view-dependent illumination while maintaining interactive frame rates. By evaluating the visual fidelity against ground truth sequences and by benchmarking, we show that our approach compares well to low-noise path traced results, but with a greatly reduced computational complexity allowing for interactive frame rates. This way, our caching technique provides a useful tool for global illumination previews and multi-view rendering.
Modern Monte-Carlo-based rendering systems still suffer from the computational complexity involved in the generation of noise-free images, making it challenging to synthesize interactive previews. We present a framework suited for rendering such previews of static scenes using a caching technique that builds upon a linkless octree. Our approach allows for memory-efficient storage and constant-time lookup to cache diffuse illumination at multiple hitpoints along the traced paths. Non-diffuse surfaces are dealt with in a hybrid way in order to reconstruct view-dependent illumination while maintaining interactive frame rates. By evaluating the visual fidelity against ground truth sequences and by benchmarking, we show that our approach compares well to low-noise path traced results, but with a greatly reduced computational complexity allowing for interactive frame rates. This way, our caching technique provides a useful tool for global illumination previews and multi-view rendering.
Evaluation and Optimization of IEEE802.11 multi-hop Backhaul Networks with Directional Antennas
(2020)
A major problem for rural areas is the inaccessibility to affordable broadband Internet connections. In these areas distances are large, and digging a cable into the ground is extremely expensive, considering the small number of potential customers at the end of that cable. This leads to a digital divide, where urban areas enjoy a high-quality service at low cost, while rural areas suffer from the reverse.
This work is dedicated to an alternative technical approach aiming to reduce the cost for Internet Service Provider in rural areas: WiFi-based Long Distance networks. A set of significant contributions of technology related aspects of WiFi-based Long Distance networks is described in three different fields: Propagation on long distance Wi-Fi links, MAC-layer scheduling and Interference modeling and Channel Assignment with directional antennas.
For each field, the author composes and discusses the state-of-the-art. Afterwards, the author derives research questions and tackles several open issues to develop these kinds of networks further towards a suitable technology for the backhaul segment.
The ability to finely segment different instances of various objects in an environment forms a critical tool in the perception tool-box of any autonomous agent. Traditionally instance segmentation is treated as a multi-label pixel-wise classification problem. This formulation has resulted in networks that are capable of producing high-quality instance masks but are extremely slow for real-world usage, especially on platforms with limited computational capabilities. This thesis investigates an alternate regression-based formulation of instance segmentation to achieve a good trade-off between mask precision and run-time. Particularly the instance masks are parameterized and a CNN is trained to regress to these parameters, analogous to bounding box regression performed by an object detection network.
In this investigation, the instance segmentation masks in the Cityscape dataset are approximated using irregular octagons and an existing object detector network (i.e., SqueezeDet) is modified to regresses to the parameters of these octagonal approximations. The resulting network is referred to as SqueezeDetOcta. At the image boundaries, object instances are only partially visible. Due to the convolutional nature of most object detection networks, special handling of the boundary adhering object instances is warranted. However, the current object detection techniques seem to be unaffected by this and handle all the object instances alike. To this end, this work proposes selectively learning only partial, untainted parameters of the bounding box approximation of the boundary adhering object instances. Anchor-based object detection networks like SqueezeDet and YOLOv2 have a discrepancy between the ground-truth encoding/decoding scheme and the coordinate space used for clustering, to generate the prior anchor shapes. To resolve this disagreement, this work proposes clustering in a space defined by two coordinate axes representing the natural log transformations of the width and height of the ground-truth bounding boxes.
When both SqueezeDet and SqueezeDetOcta were trained from scratch, SqueezeDetOcta lagged behind the SqueezeDet network by a massive ≈ 6.19 mAP. Further analysis revealed that the sparsity of the annotated data was the reason for this lackluster performance of the SqueezeDetOcta network. To mitigate this issue transfer-learning was used to fine-tune the SqueezeDetOcta network starting from the trained weights of the SqueezeDet network. When all the layers of the SqueezeDetOcta were fine-tuned, it outperformed the SqueezeDet network paired with logarithmically extracted anchors by ≈ 0.77 mAP. In addition to this, the forward pass latencies of both SqueezeDet and SqueezeDetOcta are close to ≈ 19ms. Boundary adhesion considerations, during training, resulted in an improvement of ≈ 2.62 mAP of the baseline SqueezeDet network. A SqueezeDet network paired with logarithmically extracted anchors improved the performance of the baseline SqueezeDet network by ≈ 1.85 mAP.
In summary, this work demonstrates that if given sufficient fine instance annotated data, an existing object detection network can be modified to predict much finer approximations (i.e., irregular octagons) of the instance annotations, whilst having the same forward pass latency as that of the bounding box predicting network. The results justify the merits of logarithmically extracted anchors to boost the performance of any anchor-based object detection network. The results also showed that the special handling of image boundary adhering object instances produces more performant object detectors.
Compliant manipulation is a crucial skill for robots when they are supposed to act as helping hands in everyday household tasks. Still, nowadays, those skills are hand-crafted by experts which frequently requires labor-intensive, manual parameter tuning. Moreover, some tasks are too complex to be specified fully using a task specification. Learning these skills, by contrast, requires a high number of costly and potentially unsafe interactions with the environment. We present a compliant manipulation approach using reinforcement learning guided by the Task Frame Formalism, a task specification method. This allows us to specify the easy to model knowledge about a task while the robot learns the unmodeled components by reinforcement learning. We evaluate the approach by performing a compliant manipulation task with a KUKA LWR 4+ manipulator. The robot was able to learn force control policies directly on the robot without using any simulation.
With the digital transformation, software systems have become an integral part of our society and economy. In every part of our life, software systems are increasingly utilized to, e.g., simplify housework or to optimize business processes. All these applications are connected to the Internet, which already includes millions of software services consumed by billions of people. Applications which process such a magnitude of users and data traffic requires to be highly scalable and are therefore denoted as Ultra Large Scale (ULS) systems. Roy Fielding has defined one of the first approaches which allows designing modern ULS software systems. In his doctoral thesis, Fielding introduced the architectural style Representational State Transfer (REST) which builds the theoretical foundation of the web. At present, the web is considered as the world's largest ULS system. Due to a large number of users and the significance of software for society and the economy, the security of ULS systems is another crucial quality factor besides high scalability.
Grasp verification is advantageous for autonomous manipulation robots as they provide the feedback required for higher level planning components about successful task completion. However, a major obstacle in doing grasp verification is sensor selection. In this paper, we propose a vision based grasp verification system using machine vision cameras, with the verification problem formulated as an image classification task. Machine vision cameras consist of a camera and a processing unit capable of on-board deep learning inference. The inference in these low-power hardware are done near the data source, reducing the robot's dependence on a centralized server, leading to reduced latency, and improved reliability. Machine vision cameras provide the deep learning inference capabilities using different neural accelerators. Although, it is not clear from the documentation of these cameras what is the effect of these neural accelerators on performance metrics such as latency and throughput. To systematically benchmark these machine vision cameras, we propose a parameterized model generator that generates end to end models of Convolutional Neural Networks(CNN). Using these generated models we benchmark latency and throughput of two machine vision cameras, JeVois A33 and Sipeed Maix Bit. Our experiments demonstrate that the selected machine vision camera and the deep learning models can robustly verify grasp with 97% per frame accuracy.
Object detectors have improved considerably in the last years by using advanced CNN architectures. However, many detector hyper-parameters are generally manually tuned, or they are used with values set by the detector authors. Automatic Hyper-parameter optimization has not been explored in improving CNN-based object detectors hyper-parameters. In this work, we propose the use of Black-box optimization methods to tune the prior/default box scales in Faster R-CNN and SSD, using Bayesian Optimization, SMAC, and CMA-ES. We show that by tuning the input image size and prior box anchor scale on Faster R-CNN mAP increases by 2% on PASCAL VOC 2007, and by 3% with SSD. On the COCO dataset with SSD there are mAP improvement in the medium and large objects, but mAP decreases by 1% in small objects. We also perform a regression analysis to find the significant hyper-parameters to tune.
Object detectors have improved considerably in the last years by using advanced Convolutional Neural Networks (CNNs) architectures. However, many detector hyper-parameters are not generally tuned, and they are used with values set by the detector authors. Blackbox optimization methods have gained more attention in recent years because of its ability to optimize the hyper-parameters of various machine learning algorithms and deep learning models. However, these methods are not explored in improving CNN-based object detector's hyper-parameters. In this research work, we propose the use of blackbox optimization methods such as Gaussian Process based Bayesian Optimization (BOGP), Sequential Model-based Algorithm Configuration (SMAC), and Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to tune the hyper-parameters in Faster R-CNN and Single Shot MultiBox Detector (SSD). In Faster R-CNN, tuning the input image size, prior box anchor scales and ratios using BOGP, SMAC, and CMA-ES has increased the performance around 1.5% in terms of Mean Average Precision (mAP) on PASCAL VOC. Tuning the anchor scales of SSD has increased the mAP by 3% on PASCAL VOC and marine debris datasets. On the COCO dataset with SSD, mAP improvement is observed in the medium and large objects, but mAP decreases by 1% in small objects. The experimental results show that the blackbox optimization methods have proved to increase the mAP performance by optimizing the object detectors. Moreover, it has achieved better results than the hand-tuned configurations in most of the cases.
Reinforcement learning (RL) algorithms should learn as much as possible about the environment but not the properties of the physics engines that generate the environment. There are multiple algorithms that solve the task in a physics engine based environment but there is no work done so far to understand if the RL algorithms can generalize across physics engines. In this work, we compare the generalization performance of various deep reinforcement learning algorithms on a variety of control tasks. Our results show that MuJoCo is the best engine to transfer the learning to other engines. On the other hand, none of the algorithms generalize when trained on PyBullet. We also found out that various algorithms have a promising generalizability if the effect of random seeds can be minimized on their performance.
When a robotic agent experiences a failure while acting in the world, it should be possible to discover why that failure has occurred, namely to diagnose the failure. In this paper, we argue that the diagnosability of robot actions, at least in a classical sense, is a feature that cannot be taken for granted since it strongly depends on the underlying action representation. We specifically define criteria that determine the diagnosability of robot actions. The diagnosability question is then analysed in the context of a handle manipulation action, such that we discuss two different representations of the action – a composite policy with a learned success model for the action parameters, and a neural network-based monolithic policy – both of which exist on different sides of the diagnosability spectrum. Through this comparison, we conclude that composite actions are more suited to explicit diagnosis, but representations with less prior knowledge are more flexible. This suggests that model learning may provide balance between flexibility and diagnosability; however, data-driven diagnosis methods also need to be enhanced in order to deal with the complexity of modern robots.
Short summary
This dataset accompanies our paper
A. Mitrevski, P. G. Plöger, and G. Lakemeyer, "Representation and Experience-Based Learning of Explainable Models for Robot Action Execution," in Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020.
Contents
There are three zip archives included, each of them a dump of a MongoDB database corresponding to one of the three experiments in the paper:
Grasping a drawer handle (handle_drawer_logs.zip)
Grasping a fridge handle (handle_fridge_logs.zip)
Pulling an object (pull_logs.zip)
All three experiments were performed with a Toyota HSR. Only the data necessary for learning the models used in our experiments are included here.
Usage
After unzipping the archives, each database can be restored with the command
mongorestore [directory_name]
This will create a MongoDB database with the name of the directory (handle_drawer_logs, handle_fridge_logs, and pull_logs).
Code for processing the data and model learning can be found in our <a href="https://github.com/alex-mitrevski/explainable-robot-execution-models">GitHub repository.
Facial emotion recognition is the task to classify human emotions in face images. It is a difficult task due to high aleatoric uncertainty and visual ambiguity. A large part of the literature aims to show progress by increasing accuracy on this task, but this ignores the inherent uncertainty and ambiguity in the task. In this paper we show that Bayesian Neural Networks, as approximated using MC-Dropout, MC-DropConnect, or an Ensemble, are able to model the aleatoric uncertainty in facial emotion recognition, and produce output probabilities that are closer to what a human expects. We also show that calibration metrics show strange behaviors for this task, due to the multiple classes that can be considered correct, which motivates future work. We believe our work will motivate other researchers to move away from Classical and into Bayesian Neural Networks.
Comparing Non-Visual and Visual Guidance Methods for Narrow Field of View Augmented Reality Displays
(2020)
A Comparative Study of Uncertainty Estimation Methods in Deep Learning Based Classification Models
(2020)
Deep learning models produce overconfident predictions even for misclassified data. This work aims to improve the safety guarantees of software-intensive systems that use deep learning based classification models for decision making by performing comparative evaluation of different uncertainty estimation methods to identify possible misclassifications.
In this work, uncertainty estimation methods applicable to deep learning models are reviewed and those which can be seamlessly integrated to existing deployed deep learning architectures are selected for evaluation. The different uncertainty estimation methods, deep ensembles, test-time data augmentation and Monte Carlo dropout with its variants, are empirically evaluated on two standard datasets (CIFAR-10 and CIFAR-100) and two custom classification datasets (optical inspection and RoboCup@Work dataset). A relative ranking between the methods is provided by evaluating the deep learning classifiers on various aspects such as uncertainty quality, classifier performance and calibration. Standard metrics like entropy, cross-entropy, mutual information, and variance, combined with a rank histogram based method to identify uncertain predictions by thresholding on these metrics, are used to evaluate uncertainty quality.
The results indicate that Monte Carlo dropout combined with test-time data augmentation outperforms all other methods by identifying more than 95% of the misclassifications and representing uncertainty in the highest number of samples in the test set. It also yields a better classifier performance and calibration in terms of higher accuracy and lower Expected Calibration Error (ECE), respectively. A python based uncertainty estimation library for training and real-time uncertainty estimation of deep learning based classification models is also developed.
Bei der sechsten Ausgabe des wissenschaftlichen Workshops ”Usable Security und Privacy” auf der Mensch und Computer 2020 werden wie in den vergangenen Jahren aktuelle Forschungs- und Praxisbeiträge präsentiert und anschließend mit allen Teilnehmenden diskutiert. Drei Beiträge befassen sich dieses Jahr mit dem Thema Privatsphäre, einer mit dem Thema Sicherheit. Mit dem Workshop wird ein etabliertes Forum fortgeführt und weiterentwickelt, in dem sich Expert*innen aus unterschiedlichen Domänen, z. B. dem Usability- und Security-Engineering, transdisziplinär austauschen können.
Efficient and comprehensive assessment of students knowledge is an imperative task in any learning process. Short answer grading is one of the most successful methods in assessing the knowledge of students. Many supervised learning and deep learning approaches have been used to automate the task of short answer grading in the past. We investigate why assistive grading with active learning would be the next logical step in this task as there is no absolute ground truth answer for any question and the task is very subjective in nature. We present a fast and easy method to harness the power of active learning and natural language processing in assisting the task of grading short answer questions. A webbased GUI is designed and implemented to incorporate an interactive short answer grading system. The experiments show that active learning saves the time and effort of graders in assessment and reaches the performance of supervised learning with less amount of graded answers for training.
Quantum mechanical theories are used to search and optimized the conformations of proposed small molecule candidates for treatment of SARS-CoV-2. These candidate compounds are taken from what is reported in the news and in other pre-peer-reviewed literature (e.g. ChemRxiv, bioRxiv). The goal herein is to provided predicted structures and relative conformational stabilities for selected drug and ligand candidates, in the hopes that other research groups can make use of them for developing a treatment.
This work provides a short but technical introduction to the main building blocks of a blockchain. It argues that a blockchain is not a revolutionary technology but rather a clever combination of three fields: cryptography, decentralization and game theory. In addition, it summaries the differences between a public, private and federate blockchain model and the two prominent consensus mechanism Proof-of-Work (POW) and Proof-of-Stake (POS).
Deep learning models are extensively used in various safety critical applications. Hence these models along with being accurate need to be highly reliable. One way of achieving this is by quantifying uncertainty. Bayesian methods for UQ have been extensively studied for Deep Learning models applied on images but have been less explored for 3D modalities such as point clouds often used for Robots and Autonomous Systems. In this work, we evaluate three uncertainty quantification methods namely Deep Ensembles, MC-Dropout and MC-DropConnect on the DarkNet21Seg 3D semantic segmentation model and comprehensively analyze the impact of various parameters such as number of models in ensembles or forward passes, and drop probability values, on task performance and uncertainty estimate quality. We find that Deep Ensembles outperforms other methods in both performance and uncertainty metrics. Deep ensembles outperform other methods by a margin of 2.4% in terms of mIOU, 1.3% in terms of accuracy, while providing reliable uncertainty for decision making.
Telepresence robots allow people to participate in remote spaces, yet they can be difficult to manoeuvre with people and obstacles around. We designed a haptic-feedback system called “FeetBack," which users place their feet in when driving a telepresence robot. When the robot approaches people or obstacles, haptic proximity and collision feedback are provided on the respective sides of the feet, helping inform users about events that are hard to notice through the robot’s camera views. We conducted two studies: one to explore the usage of FeetBack in virtual environments, another focused on real environments.We found that FeetBack can increase spatial presence in simple virtual environments. Users valued the feedback to adjust their behaviour in both types of environments, though it was sometimes too frequent or unneeded for certain situations after a period of time. These results point to the value of foot-based haptic feedback for telepresence robot systems, while also the need to design context-sensitive haptic feedback.