Contributions published at Department of Informatics (Burkhard Stiller)

The behaviour of Bitcoin owners is reflected in the structure and the number of bitcoin transactions encoded in the Blockchain. Likewise, the behaviour of Bitcoin traders is reflected in the formation of bullish and bearish trends in the crypto market. In light of these observations, we wonder if human behaviour underlies some relationship between the Blockchain and the crypto market. To address this question, we map the Blockchain to a spin-lattice problem, whose configurations form ordered and disordered patterns, representing the behaviour of Bitcoin owners. This novel approach allows us to obtain time series suitable to detect a causal relationship between the dynamics of the Blockchain and market trends of the Bitcoin and to find that disordered patterns in the Blockchain precede Bitcoin panic selling. Our results suggest that human behaviour underlying Blockchain evolution and the crypto market brings out a fascinating connection between disorder and panic in Bitcoin dynamics.

While there has been significant study of both individuals and teams of knowledge workers, research has focused largely on one or the other, with less focus on the interaction between the two. In this paper, we explore the tensions between the individual and their team, focusing on the choices an individual makes towards their own productivity versus their team's productivity. We developed a technology probe with a team nudge that fosters recurring reflection and prompts individuals to consider how their team helps them to be productive. We examined its impact through a longitudinal field study with 48 participants. We chose to undertake this study with software development teams as they are examples of knowledge workers who collaborate on a shared set of tasks with specific goals. Our exploration took place with hybrid development teams, which have increasingly become the norm. Our analysis of a total of 8338 hourly self-reports and 1389 daily diary entries found that the team nudge increased participants' productivity ratings and team awareness, led to participants spending more time on their own tasks, reshaped their perceptions of themselves and their team, yet, in general, did not increase team cohesion or affect well-being.

Digital marketing has transformed referral marketing, revealing limitations in traditional centralized systems such as trust, transparency, and efficiency, however, the potential advantages of decentralized systems remain underexplored. This paper investigates the feasibility of a high-volume, decentralized referral system. The approach assesses smart contract prototypes for cost-effectiveness and performance in high-user engagement scenarios in different EVM-compatible blockchains and referral strategies, such as multilevel referrals. Findings confirm the technical viability as a blueprint for designing and implementing similar systems, highlighting challenges in real-world deployments, such as Sybil attacks, and the interplay between technical and economical design factors.

Many employees prefer to work from home, yet struggle to squeeze their office into an already fully-utilized space. Virtual Reality (VR) seemingly offered a solution with its ability to transform even modest physical spaces into spacious, productive virtual offices, but hardware challenges---such as low resolution---have prevented this from becoming a reality. Now that hardware issues are being overcome, we are able to investigate the suitability of VR for daily work. To do so, we (1) studied the physical space that users typically dedicate to home offices and (2) conducted an exploratory study of users working in VR for one week. For (1) we used digital ethnography to study 430 self-published images of software developer workstations in the home, confirming that developers faced myriad space challenges. We used speculative design to re-envision these as VR workstations, eliminating many challenges. For (2) we asked 10 developers to work in their own home using VR for about two hours each day for four workdays, and then interviewed them. We found that working in VR improved focus and made mundane tasks more enjoyable. While some subjects reported issues---annoyances with the fit, weight, and umbilical cord of the headset---the vast majority of these issues seem to be addressable. Together, these studies show VR technology has the potential to address many key problems with home workstations, and, with continued improvements, may become an integral part of creating an effective workstation in the home.

The use of drones (or Unmanned Aerial Vehicles) in urban areas has emerged rapidly in the last decade, and continues to expand at an accelerating pace. Alongside the emergent uses of high-impact technology in both public and private sectors, political debates about the potential risks and challenges have arisen, encompassing diverse perspectives and attitudes about the ethical, legal, social, and regulatory implications of introducing and integrating new technology in society. This scoping review offers an assessment of the societal acceptance factors of urban drones discussed in the current academic literature. We used a hybrid approach including quantitative landscape mapping and qualitative content analysis of the selected articles to inductively develop a typology of acceptance factors associated with urban use of drones. This review illuminates areas that have been the focus of attention within the current body of knowledge (e.g., visual and noise pollution of drones), sketches the evolution of the relevant discussions over time (e.g., a focus on the safety of the drone technology toward safety of the cargo it carries and security of the data it collects), and points to areas that have received less considerations (e.g., media appropriation and social group influence). It can, thus, help situate the topic of societal acceptance of urban drones in specific contexts, and orient future research on promoting value sensitive innovation in society more broadly.

Bluetooth Low Energy (BLE) personal trackers are affordable devices misused to track nonconsensual individuals. Due to the increased misuse, Apple implemented two detection applications. However, the Android application is limited to user-initiated scans with a fixed detection algorithm. This paper focuses on reducing the misuse of malicious trackers by examining current solutions, potential generic detection approaches, and improving tracker detection times. HomeScout expands detection to the Tile and Samsung Galaxy SmartTag+, and examines the misuse potential of all BLE-enabled devices. HomeScout can reliably detect devices tracking the user as quickly as 1 minute once in motion by optimizing the parameters. The optimal parameter setting for distance is 200 m due to its high recall rate, for occurrence is 2, and for time is 1 minute. Furthermore, HomeScout applies the tracking algorithm to all BLE-enabled devices.

This thesis introduces the design of KroneDB, a system for compressing time series data using the Kronecker decomposition, while allowing for efficient evaluation of relational queries including selection, projection, join, and aggregates (SPJA). KroneDB allows for a tunable trade-off between compression ratio and approximation error, while exploiting periodic patterns within the data to improve the compression. The compressed data can be queried directly without prior decompression while reducing the runtime of most queries. Updates can be applied directly to the compressed data and naturally enable value imputation and outlier detection in the updating process. By embedding our approach into the Functional Aggregate Queries (FAQ) framework, we show that it can be applied to a wide range of fundamental problems.

Humans can perform parkour by traversing obstacles in a highly dynamic fashion requiring precise eye-muscle coordination and movement. Getting robots to do the same task requires overcoming similar challenges. Classically, this is done by independently engineering perception, actuation, and control systems to very low tolerances. This restricts them to tightly controlled settings such as a predetermined obstacle course in labs. In contrast, humans are able to learn parkour through practice without significantly changing their underlying biology. In this paper, we take a similar approach to developing robot parkour on a small low-cost robot with imprecise actuation and a single front-facing depth camera for perception which is low-frequency, jittery, and prone to artifacts. We show how a single neural net policy operating directly from a camera image, trained in simulation with large-scale RL, can overcome imprecise sensing and actuation to output highly precise control behavior end-to-end. We show our robot can perform a high jump on obstacles 2x its height, long jump across gaps 2x its length, do a handstand and run across tilted ramps, and generalize to novel obstacle courses with different physical properties. Parkour videos at https://extreme-parkour.github.io/.

During the last decades, unmanned drones have been deployed to a broad range of use cases. With the emergence and continuous development of artificial intelligence, applications which involve autonomous drones have attracted the attention of numerous fields of research. Currently, autonomous multi drone systems are studied and developed across fields like robotic, computer science, and environmental science. However, while the technical capabilities of such autonomous systems are constantly increasing, their impact on accountability has gone unnoticed. This seems surprising since autonomous drones nowadays have the capability to decide independently from any human intervention. This study integrates different perspectives on accountability to answer the question: How do autonomous drones’ applications impact accountability? Analysing two firefighting use cases, this study reveals three pitfalls to accountability in the application of autonomous drones, namely: a) autonomous drones are not perceived as being autonomous, b) the drones are not perceived as trustworthy and official actors, and lastly, c) the interaction between the drones and participants is leading to accountability-relevant problems. The study concludes with a set of proposed solutions to mitigate the identified pitfalls and, thus, enhance accountability in the studied firefighting use cases. The proposed solutions also serve as a starting point for further research on this topic and to verify or discover further enhancements to accountability aspects in the context of autonomous systems.

This thesis draws a vision on how to autonomously drive an open water search and rescue (SAR) operation based on drones. On the basis of a real-life SAR process at lake Thunersee, the paper aims to provide benefits and drawbacks of an autonomous drone system. Existing scientific concepts as well as systems in practice are presented within this work. Through expert interviews of SAR specialists, the existing process is analyzed. In a next step an autonomous drone entity is incorporated into the SAR process and presented with the six layers of collaboration model. Further a new thinkLet, focusing the search activity within the SAR process, is derived. Subsequently, a cloud-based solution is developed which allows to autonomously assign drones to a search area. This application serves as a foundation for further functional extensions. Some of these enhancements are described at the end of the thesis and assembled into a vision towards autonomous drone assistance in open water SAR operations. The thesis concludes that the inclusion of autonomous drone entities as teammates, rather than just as a tool, strengthen SAR operations and poses new opportunities.

Drones have the potential to revolutionize power line inspection by increasing productivity, reducing inspection time, improving data quality, and eliminating the risks for human operators. Current state-of-the-art systems for power line inspection have two shortcomings: (i) control is decoupled from perception and needs accurate information about the location of the power lines and masts; (ii) obstacle avoidance is decoupled from the power line tracking, which results in poor tracking in the vicinity of the power masts, and, consequently, in decreased data quality for visual inspection. In this work, we propose a model predictive controller (MPC) that overcomes these limitations by tightly coupling perception and action. Our controller generates commands that maximize the visibility of the power lines while, at the same time, safely avoiding the power masts. For power line detection, we propose a lightweight learning-based detector that is trained only on synthetic data and is able to transfer zero-shot to real-world power line images. We validate our system in simulation and real-world experiments on a mock-up power line infrastructure. We release our code and datasets to the public.

The development of effective vision-based algorithms has been a significant challenge in achieving autonomous drones, which promise to offer immense potential for many real-world applications. This paper investigates learning deep sensorimotor policies for vision-based drone racing, which is a particularly demanding setting for testing the limits of an algorithm. Our method combines feature representation learning to extract task-relevant feature representations from high-dimensional image inputs with a learning-by-cheating framework to train a deep sensorimotor policy for vision-based drone racing. This approach eliminates the need for globally-consistent state estimation, trajectory planning, and handcrafted control design, allowing the policy to directly infer control commands from raw images, similar to human pilots. We conduct experiments using a realistic simulator and show that our vision-based policy can achieve state-of-the-art racing performance while being robust against unseen visual disturbances. Our study suggests that consistent feature embeddings are essential for achieving robust control performance in the presence of visual disturbances. The key to acquiring consistent feature embeddings is utilizing contrastive learning along with data augmentation.

Today, state-of-the-art deep neural networks that process events first convert them into dense, grid-like input representations before using an off-the-shelf network. However, selecting the appropriate representation for the task traditionally requires training a neural network for each representation and selecting the best one based on the validation score, which is very time-consuming. This work eliminates this bottleneck by selecting representations based on the Gromov-Wasserstein Discrepancy (GWD) between raw events and their representation. It is about 200 times faster to compute than training a neural network and preserves the task performance ranking of event representations across multiple representations, network backbones, datasets, and tasks. Thus finding representations with high task scores is equivalent to finding representations with a low GWD. We use this insight to, for the first time, perform a hyperparameter search on a large family of event representations, revealing new and powerful representations that exceed the state-of-the-art. Our optimized representations outperform existing representations by 1.7 mAP on the 1 Mpx dataset and 0.3 mAP on the Gen1 dataset, two established object detection benchmarks, and reach a 3.8% higher classification score on the mini N-ImageNet benchmark. Moreover, we outperform state-of-the-art by 2.1 mAP on Gen1 and state-of-the-art feed-forward methods by 6.0 mAP on the 1 Mpx datasets. This work opens a new unexplored field of explicit representation optimization for event-based learning.

Event-based cameras are ideal for line-based motion estimation, since they predominantly respond to edges in the scene. However, accurately determining the camera displacement based on events continues to be an open problem. This is because line feature extraction and dynamics estimation are tightly coupled when using event cameras, and no precise model is currently available for describing the complex structures generated by lines in the space-time volume of events. We solve this problem by deriving the correct non-linear parametrization of such manifolds, which we term eventails, and demonstrate its application to eventbased linear motion estimation, with known rotation from an Inertial Measurement Unit. Using this parametrization, we introduce a novel minimal 5-point solver that jointly estimates line parameters and linear camera velocity projections, which can be fused into a single, averaged linear velocity when considering multiple lines. We demonstrate on both synthetic and real data that our solver generates more stable relative motion estimates than other methods while capturing more inliers than clustering based on spatiotemporal planes. In particular, our method consistently achieves a 100% success rate in estimating linear velocity where existing closed-form solvers only achieve between 23% and 70%. The proposed eventails contribute to a better understanding of spatio-temporal event-generated geometries and we thus believe it will become a core building block of future event-based motion estimation algorithms.

In this paper, we introduce a general concept for the analysis of temporal and multivariate data and the system LFPeers that applies this concept to temporal similarity search and results exploration. The conceptual workflow divides the analysis in two phases: a search phase to find the most similar objects to a query object before a time point in the temporal data, and an exploration phase to analyze and contextualize this subset of objects after . LFPeers enables users to search for peers through interactive similarity search and filtering, explore interesting behavior of this peer group, and learn from peers through the assessment of diverging behaviors. We present the conceptual workflow to learn from peers and the LFPeers system with novel interfaces for search and exploration in temporal and multivariate data. An earlier workshop publication for LFPeers included a usage scenario targeting epidemiologists and the public who want to learn from the Covid-19 pandemic and distinguish successful and ineffective measures. In this extended paper, we now show how our concept is generalized and applied by domain experts in two case studies, including a novel case on stocks data. Finally, we reflect on the new state of development and on the insights gained by the experts in the case studies on the search and exploration of temporal data to learn from peers.

The use of drones offers police forces potential gains in efficiency and safety. However, their use may also harm public perception of the police if drones are refused. Therefore, police forces should consider the perception of bystanders and broader society to maximize drones' potential. This article examines the concerns expressed by members of the public during a field trial involving 52 test participants. Analysis of the group interviews suggests that their worries go beyond airspace safety and privacy, broadly dis-cussed in existing literature and regulations. The interpretation of the results indicates that the perceived justice of drone use is a significant factor in acceptance. Leveraging the concept of organizational justice and data collected, we propose a catalogue of guidelines for just operation of drones to supplement the existing policy. We present the organizational justice perspective as a framework to integrate the concerns of the public and bystanders into legal work. Finally, we discuss the relevance of justice for the legitimacy of the police's actions and provide implications for research and practice.

This study explores shifts in political trust during the outbreak of the COVID-19 pandemic in Switzerland, examining the role that media consumption and threat perceptions played in individuals’ trust in politics. We combine panel surveys taken before and during the first nation-wide lockdown with webtracking data of participants' online behaviour to paint a nuanced picture of media effects during the crisis. Our work has several findings. First, political trust, an attitude known for its stability, increased following lockdown. Second, consumption of mainstream news on COVID-19 directly hindered this increase, with those reading more news having lower over-time trust, while the relatively minor alternative news consumption had no direct effect on political trust. Third, threat perceptions a) to health and b) from the policy response to the pandemic, have strong and opposite effects on political trust, with threats to health increasing trust, and threats from the government policy response decreasing it. Lastly, these threat perceptions condition the effect of COVID-19 news consumption on political trust: perceptions of threat had the power to both exacerbate and mute the effect of media consumption on government trust during the pandemic. Notably, we show that the expected negative effect of alternative news on political trust only exists for those who did not think COVID-19 posed a threat to their health, while public service news consumption reduced the negative effect produced by government threat perceptions. The paper therefore advances our understanding of the nuanced nature of media effects, particularly as relates to alternative media, especially during moments of crisis.

Scatterplot sampling has long been an efficient and effective way to resolve the overplotting issues commonly occurring in large-scale scatterplot visualization applications. However, it is challenging to preserve the existence of low-density points or outliers after sampling for a sub-sampling algorithm if, at the same time, faithfully representing the relative data densities is of importance. In this work, we propose to address this issue in a visual-assisted manner. While the whole dataset is sub-sampled, the density of the outliers is modeled and visually integrated into the final scatterplot together with the sub-sampled point data. We showcase the effectiveness of our proposed method in various cases and user studies.

Video retrieval still offers many challenges which can so far only be effectively mediated through interactive, human-in-the-loop retrieval approaches. The vitrivr multimedia retrieval stack offers a broad range of query mechanisms to enable users to perform such interactive retrieval. While these multiple mechanisms offer various options to experienced users, they can be difficult to use for novices. In this paper, we present a minimal user interface geared towards novice users that only exposes a subset of vitrivr’s functionality but simplifies user interaction.

null