Wolfram Burgard

University of Technology Nuremberg

In today's chemical production plants, human field operators perform frequent checks on the plant's integrity to guarantee high safety standards, and thus are possibly the first to encounter dangerous operating conditions. To alleviate their tasks of failure detection and monitoring by audio, visual, and olfactory perceptions, we present a robotic...

Simultaneous localization and mapping (SLAM) in slowly varying scenes is important for long-term robot task completion. Failing to detect scene changes may lead to inaccurate maps and, ultimately, lost robots. Classical SLAM algorithms assume static scenes, and recent works take dynamics into account, but require scene changes to be observed in con...

Geometric regularity, which leverages data symmetry, has been successfully incorporated into deep learning architectures such as CNNs, RNNs, GNNs, and Transformers. While this concept has been widely applied in robotics to address the curse of dimensionality when learning from high-dimensional data, the inherent reflectional and rotational symmetry...

Lane graph estimation is a long-standing problem in the context of autonomous driving. Previous works aimed at solving this problem by relying on large-scale, hand-annotated lane graphs, introducing a data bottleneck for training models to solve this task. To overcome this limitation, we propose to use the motion patterns of traffic participants as...

In policy learning for robotic manipulation, sample efficiency is of paramount importance. Thus, learning and extracting more compact representations from camera observations is a promising avenue. However, current methods often assume full observability of the scene and struggle with scale invariance. In many tasks and settings, this assumption do...

Reliable scene understanding is indispensable for modern autonomous systems. Current learning-based methods typically try to maximize their performance based on segmentation metrics that only consider the quality of the segmentation. However, for the safe operation of a system in the real world it is crucial to consider the uncertainty in the predi...

The safe deployment of autonomous vehicles relies on their ability to effectively react to environmental changes. This can require maneuvering on varying surfaces which is still a difficult problem, especially for slippery terrains. To address this issue we propose a new approach that learns a surface-aware dynamics model by conditioning it on a la...

Visual odometry is a fundamental task for many applications on mobile devices and robotic platforms. Since such applications are oftentimes not limited to predefined target domains and learning-based vision systems are known to generalize poorly to unseen environments, methods for continual adaptation during inference time are of significant intere...

Operating a robot in the open world requires a high level of robustness with respect to previously unseen environments. Optimally, the robot is able to adapt by itself to new conditions without human supervision, e.g., automatically adjusting its perception system to changing lighting conditions. In this work, we address the task of continual learn...

Early stopping based on the validation set performance is a popular approach to find the right balance between under- and overfitting in the context of supervised learning. However, in reinforcement learning, even for supervised sub-problems such as world model learning, early stopping is not applicable as the dataset is continually evolving. As a...

Robots operating in the open world encounter various different environments that can substantially differ from each other. This domain gap also poses a challenge for Simultaneous Localization and Mapping (SLAM) being one of the fundamental tasks for navigation. In particular, learning-based SLAM methods are known to generalize poorly to unseen envi...

Uncertainty estimation is crucial in safety-critical settings such as automated driving as it provides valuable information for several downstream tasks including high-level decision-making and path planning. In this work, we propose EvCenterNet, a novel uncertainty-aware 2D object detection framework utilizing evidential learning to directly estim...

A key component of graph-based SLAM systems is the ability to detect loop closures in a trajectory to reduce the drift accumulated over time from the odometry. Most LiDAR-based methods achieve this goal by using only the geometric information, disregarding the semantics of the scene. In this work, we introduce PADLoC for joint loop closure detectio...

Lane graph estimation is an essential and highly challenging task in automated driving and HD map learning. Existing methods using either onboard or aerial imagery struggle with complex lane topologies, out-of-distribution scenarios, or significant occlusions in the image space. Moreover, merging overlapping lane graphs to obtain consistent large-s...

Bird's-Eye-View (BEV) semantic maps have become an essential component of automated driving pipelines due to the rich representation they provide for decision-making tasks. However, existing approaches for generating these maps still follow a fully supervised training paradigm and hence rely on large amounts of annotated BEV data. In this work, we...

Lane graph estimation is a long-standing problem in the context of autonomous driving. Previous works aimed at solving this problem by relying on large-scale, hand-annotated lane graphs, introducing a data bottleneck for training models to solve this task. To overcome this limitation, we propose to use the motion patterns of traffic participants as...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

Grounding language to the visual observations of a navigating agent can be performed using off-the-shelf visual-language models pretrained on Internet-scale data (e.g., image captions). While this is useful for matching images to natural language descriptions of object goals, it remains disjoint from the process of mapping the environment, so that...

Modern autonomous systems often rely on LiDAR scanners, in particular for autonomous driving scenarios. In this context, reliable scene understanding is indispensable. Current learning-based methods typically try to achieve maximum performance for this task, while neglecting a proper estimation of the associated uncertainties. In this work, we intr...

Recent works have shown that Large Language Models (LLMs) can be applied to ground natural language to a wide variety of robot skills. However, in practice, learning multi-task, language-conditioned robotic skills typically requires large-scale data collection and frequent human intervention to reset the environment or help correcting the current p...

A long-standing goal in robotics is to build robots that can perform a wide range of daily tasks from perceptions obtained with their onboard sensors and specified only via natural language. While recently substantial advances have been achieved in language-driven robotics by leveraging end-to-end learning from pixels, there is no clear and well-un...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In this chapter, the law scholar Jan von Hein analyses and evaluates the European Parliament’s proposal on a civil liability regime for artificial intelligence against the background of the already existing European regulatory framework on private international law, in particular the Rome I and II Regulations. The draft regulation (DR) proposed by...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In this chapter, the philosopher Mathias Risse reflects on the medium and long-term prospects and challenges democracy faces from AI. Comparing the political nature of AI systems with traffic infrastructure, the author points out AI’s potential to greatly strengthen democracy, but only with the right efforts. The chapter starts with a critical exam...

The philosopher Wilfried Hinsch focuses on statistical discrimination by means of computational profiling. He defines statistical profiling as an estimate of what individuals will do by considering the group of people they can be assigned to. The author explores which criteria of fairness and justice are appropriate for the assessment of computatio...

In this chapter, the ethics and international law scholar Silja Voeneky and the mathematician Thorsten Schmidt propose a new adaptive regulation scheme for AI-driven products and services. To this end, the authors examine different regulatory regimes, including the European Medical Devices Regulation (MDR), and the proposed AI Act by the European C...

To answer the question of what responsible AI means, the authors, Jaan Tallinn and Richard Ngo, propose a framework for the deployment of AI which focuses on two concepts: delegation and supervision. The framework aims towards building ‘delegate AIs’ which lack goals of their own but can perform any task delegated to them. However, AIs trained with...

In this chapter, the law scholar Ralf Poscher sets out to show how AI challenges the traditional understanding of the right to data protection and presents an outline of an alternative conception that better deals with emerging AI technologies. Firstly, Poscher explains how the traditional conceptualisation of data protection as an independent fund...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

The chapter aims to serve as a conceptual sketch for the intricacies involved in autonomous algorithmic collusion, including the notion of concerted practices for cases that would otherwise elude the cartel prohibition. Stefan Thomas, a law scholar, starts by assessing how algorithms can influence competition in markets before dealing with the trad...

In this chapter, the philosophers Oliver Mueller and Boris Essmann address AI-supported neurotechnology, especially Brain–Computer Interfaces (BCIs) that may in the future supplement and restore functioning in agency-limited individuals or even augment or enhance capacities for natural agency. The authors propose a normative framework for the evalu...

In this chapter, Fruzsina Molnár-Gábor and Johanne Giesecke consider specific aspects of how the application of AI-based systems in medical contexts may be guided under international standards. They sketch the relevant international frameworks for the governance of medical AI. Among the frameworks that exist, the World Medical Association’s activit...

This chapter by the philosopher Johanna Thoma focuses on the ‘moral proxy problem’, which arises when an autonomous artificial agent makes a decision as a proxy for a human agent, without it being clear for whom specifically it does so. Thoma recognises that, in general, there are broadly two categories of agents an artificial agent can be a proxy...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In this chapter, the law scholar Christoph Krönke focuses on the legal challenges faced by healthcare AI Alter Egos, especially in the European Union. Firstly, the author outlines the functionalities of AI Alter Egos in the healthcare sector. Based on this, he explores the applicable legal framework as AI Alter Egos have two main functions: collect...

In this chapter, Mathias Paul explores the topic of AI systems in the financial industry. After outlining different areas of application of AI in the financial sector and different regulatory regimes relevant to robo-finance, the author analyses the risks emerging from AI applications in the financial industry. He argues that AI systems applied in...

The law scholars Weixing Shen and Yun Liu focus on China’s efforts in the field of AI regulation and spell out recent legislative actions. While there is no unified AI law today in China, many provisions from Chinese data protection law are in part applicable to AI systems. The authors particularly analyse the rights and obligations from the Chines...

In this chapter, law and technology scholar Jonathan Zittrain warns of the danger of relying on answers for which we have no explanations. There are benefits to utilising solutions discovered through trial and error rather than rigorous proof: though aspirin was discovered in the late 19th century, it was not until the late 20th century that scient...

In this chapter, Thomas Burri, an international lawyer, examines how general ethical norms on AI diffuse into domestic law directly, without engaging international law. The chapter discusses various ethical AI frameworks and shows how they influenced the European Union Commission’s proposal for an AI Act. It reveals the origins of the EU proposal a...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In this chapter, Philipp Kellmeyer discusses how to protect mental privacy and mental integrity in the interaction of AI-based neurotechnology from the perspective of philosophy, ethics, neuroscience, and psychology. The author argues that mental privacy and integrity are important anthropological goods that need to be protected from unjustified in...

In this chapter, the law scholar Boris Paal identifies a conflict between two objectives pursued by the data protection law, the comprehensive protection of privacy and personal rights and the facilitation of an effective and competitive data economy. Focusing on the European Union’s General Data Protection Regulation (GDPR), the author recognises...

In this chapter, the philosopher Christoph Durt elaborates a novel view on AI and its relation to humans. He contends that AI is neither merely a tool, nor an artificial subject, nor necessarily a simulation of human intelligence. These misconceptions of AI have led to grave misunderstandings of the opportunities and dangers of AI. A more comprehen...

The law scholar Dustin Lewis explores the requirements of international law with regard to the employments of AI-related tools and techniques in armed conflict. The scope of this chapter is not limited to Lethal Autonomous Weapons (AWS) but also encompasses other AI-related tools and techniques related to warfighting, detention, and humanitarian se...

The chapter by the philosopher Catrin Misselhorn provides an overview of the most central debates in artificial morality and machine ethics. Artificial moral agents are AI systems which are able to recognise the morally relevant aspects of a situation and take them into account in their decisions and actions. Misselhorn shows that artificial morali...

This chapter explores the changes that AI brings about in corporate law and corporate governance, especially in terms of the challenges it poses for corporations. The law scholar Jan Lieder argues that whilst there is the potential to enhance the current system, there are also risks of destabilisation. Although algorithms are already being used in...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In this chapter, the philosopher Thomas Metzinger lists five main problem domains related to AI systems. For each problem field, he proposes several measures which should be taken. Firstly, there should be worldwide safety standards concerning the research and development of AI. If not, Metzinger fears a ‘race to the bottom’ in safety standards. Ad...

In the past decade, artificial intelligence (AI) has become a disruptive force around the world, offering enormous potential for innovation but also creating hazards and risks for individuals and the societies in which they live. This volume addresses the most pressing philosophical, ethical, legal, and societal challenges posed by AI. Contributors...

In this chapter, the law scholar Ebrahim Afsah outlines different implications of AI for the area of national security. He argues that while AI overlaps with many challenges to the national security arising from cyberspace, it also creates new risks, including the emergence of a superintelligence in the future, the development of autonomous weapons...

In this chapter, the law scholar Christine Wendehorst analyses the different potential risks posed by AI as part of two main categories, safety risks and fundamental rights risks. Based on this, the author considers why AI challenges existing liability regimes. She spells out the main solutions put forward so far and evaluates them. This chapter hi...

In this chapter the law scholars Haksoo Ko, Sangchul Park, and Yong Lim, analyse the way South Korea has been dealing with the COVID-19 pandemic and its legal consequences. Instead of enforcing strict lockdowns, South Korea imposed several other measures, such as a robust AI-based contact tracing scheme. The chapter provides an overview of the lega...

In this chapter, political philosopher Alex Leveringhaus asks whether Lethal Autonomous Weapons (AWS) are morally repugnant and whether this entails that they should be prohibited by international law. To this end, Leveringhaus critically surveys three prominent ethical arguments against AWS: firstly, AWS create ‘responsibility gaps’; secondly, tha...

This chapter by the law scholar Antje von Ungern-Sternberg focuses on the legality of discriminatory AI which is increasingly used to assess people (profiling). Intelligent algorithms – which are free of human prejudices and stereotypes – would prevent discriminatory decisions, or so the story goes. However, many studies show that the use of AI can...

A key component of graph-based SLAM systems is the ability to detect loop closures in a trajectory to reduce the drift accumulated over time from the odometry. Most LiDAR-based methods achieve this goal by using only the geometric information, disregarding the semantics of the scene. In this work, we introduce PADLoC, a LiDAR-based loop closure det...

Everyday tasks of long-horizon and comprising a sequence of multiple implicit subtasks still impose a major challenge in offline robot control. While a number of prior methods aimed to address this setting with variants of imitation and offline reinforcement learning, the learned behavior is typically narrow and often struggles to reach configurabl...

For autonomous skill acquisition, robots have to learn about the physical rules governing the 3D world dynamics from their own past experience to predict and reason about plausible future outcomes. To this end, we propose a transformation-based 3D video prediction (T3VIP) approach that explicitly models the 3D motion by decomposing a scene into its...

Robustly classifying ground infrastructure such as roads and street crossings is an essential task for mobile robots operating alongside pedestrians. While many semantic segmentation datasets are available for autonomous vehicles, models trained on such datasets exhibit a large domain gap when deployed on robots operating in pedestrian spaces. Manu...

Setting up robot environments to quickly test newly developed algorithms is still a difficult and time consuming process. This presents a significant hurdle to researchers interested in performing real-world robotic experiments. RobotIO is a python library designed to solve this problem. It focuses on providing common, simple, and well structured p...

In this paper we propose USegScene, a framework for semantically guided unsupervised learning of depth, optical flow and ego-motion estimation for stereo camera images using convolutional neural networks. Our framework leverages semantic information for improved regularization of depth and optical flow maps, multimodal fusion and occlusion filling...

General-purpose robots coexisting with humans in their environment must learn to relate human language to their perceptions and actions to be useful in a range of daily tasks. Moreover, they need to acquire a diverse repertoire of general-purpose skills that allow composing long-horizon tasks by following unconstrained language instructions. In thi...

Robust detection of moving vehicles is a critical task for any autonomously operating outdoor robot or self-driving vehicle. Most modern approaches for solving this task rely on training image-based detectors using large-scale vehicle detection datasets such as nuScenes or the Waymo Open Dataset. Providing manual annotations is an expensive and lab...

Reliable scene understanding is indispensable for modern autonomous systems. Current learning-based methods typically try to maximize their performance based on segmentation metrics that only consider the quality of the segmentation. However, for the safe operation of a system in the real world it is crucial to consider the uncertainty in the predi...

A long-standing goal in robotics is to build robots that can perform a wide range of daily tasks from perceptions obtained with their onboard sensors and specified only via natural language. While recently substantial advances have been achieved in language-driven robotics by leveraging end-to-end learning from pixels, there is no clear and well-un...

Service robots in the future need to execute abstract instructions such as “fetch the milk from the fridge”. To translate such instructions into actionable plans, robots require in-depth background knowledge. With regards to interactions with doors and drawers, robots require articulation models that they can use for state estimation and motion pla...

Learning to solve complex manipulation tasks from visual observations is a dominant challenge for real-world robot learning. Although deep reinforcement learning algorithms have recently demonstrated impressive results in this context, they still require an impractical amount of time-consuming trial-and-error iterations. In this work, we consider t...

While lifelong SLAM addresses the capability of a robot to adapt to changes within a single environment over time, in this paper we introduce the task of continual SLAM. Here, a robot is deployed sequentially in a variety of different environments and has to transfer its knowledge of previously experienced environments to thus far unseen environmen...

Existing Deep Learning (DL) frameworks typically do not provide ready-to-use solutions for robotics, where very specific learning, reasoning, and embodiment problems exist. Their relatively steep learning curve and the different methodologies employed by DL compared to traditional approaches, along with the high complexity of DL models, which often...