-
Robust Uncertainty Estimation for Classification of Maritime Objects
Authors:
Jonathan Becktor,
Frederik Scholler,
Evangelos Boukas,
Lazaros Nalpantidis
Abstract:
We explore the use of uncertainty estimation in the maritime domain, showing the efficacy on toy datasets (CIFAR10) and proving it on an in-house dataset, SHIPS. We present a method joining the intra-class uncertainty achieved using Monte Carlo Dropout, with recent discoveries in the field of outlier detection, to gain more holistic uncertainty measures. We explore the relationship between the int…
▽ More
We explore the use of uncertainty estimation in the maritime domain, showing the efficacy on toy datasets (CIFAR10) and proving it on an in-house dataset, SHIPS. We present a method joining the intra-class uncertainty achieved using Monte Carlo Dropout, with recent discoveries in the field of outlier detection, to gain more holistic uncertainty measures. We explore the relationship between the introduced uncertainty measures and examine how well they work on CIFAR10 and in a real-life setting. Our work improves the FPR95 by 8% compared to the current highest-performing work when the models are trained without out-of-distribution data. We increase the performance by 77% compared to a vanilla implementation of the Wide ResNet. We release the SHIPS dataset and show the effectiveness of our method by improving the FPR95 by 44.2% with respect to the baseline. Our approach is model agnostic, easy to implement, and often does not require model retraining.
△ Less
Submitted 3 July, 2023;
originally announced July 2023.
-
Self-Supervised Traversability Prediction by Learning to Reconstruct Safe Terrain
Authors:
Robin Schmid,
Deegan Atha,
Frederik Schöller,
Sharmita Dey,
Seyed Fakoorian,
Kyohei Otsu,
Barry Ridge,
Marko Bjelonic,
Lorenz Wellhausen,
Marco Hutter,
Ali-akbar Agha-mohammadi
Abstract:
Navigating off-road with a fast autonomous vehicle depends on a robust perception system that differentiates traversable from non-traversable terrain. Typically, this depends on a semantic understanding which is based on supervised learning from images annotated by a human expert. This requires a significant investment in human time, assumes correct expert classification, and small details can lea…
▽ More
Navigating off-road with a fast autonomous vehicle depends on a robust perception system that differentiates traversable from non-traversable terrain. Typically, this depends on a semantic understanding which is based on supervised learning from images annotated by a human expert. This requires a significant investment in human time, assumes correct expert classification, and small details can lead to misclassification. To address these challenges, we propose a method for predicting high- and low-risk terrains from only past vehicle experience in a self-supervised fashion. First, we develop a tool that projects the vehicle trajectory into the front camera image. Second, occlusions in the 3D representation of the terrain are filtered out. Third, an autoencoder trained on masked vehicle trajectory regions identifies low- and high-risk terrains based on the reconstruction error. We evaluated our approach with two models and different bottleneck sizes with two different training and testing sites with a fourwheeled off-road vehicle. Comparison with two independent test sets of semantic labels from similar terrain as training sites demonstrates the ability to separate the ground as low-risk and the vegetation as high-risk with 81.1% and 85.1% accuracy.
△ Less
Submitted 2 August, 2022;
originally announced August 2022.
-
Early Recall, Late Precision: Multi-Robot Semantic Object Map** under Operational Constraints in Perceptually-Degraded Environments
Authors:
Xianmei Lei,
Taeyeon Kim,
Nicolas Marchal,
Daniel Pastor,
Barry Ridge,
Frederik Schöller,
Edward Terry,
Fernando Chavez,
Thomas Touma,
Kyohei Otsu,
Ali Agha
Abstract:
Semantic object map** in uncertain, perceptually degraded environments during long-range multi-robot autonomous exploration tasks such as search-and-rescue is important and challenging. During such missions, high recall is desirable to avoid missing true target objects and high precision is also critical to avoid wasting valuable operational time on false positives. Given recent advancements in…
▽ More
Semantic object map** in uncertain, perceptually degraded environments during long-range multi-robot autonomous exploration tasks such as search-and-rescue is important and challenging. During such missions, high recall is desirable to avoid missing true target objects and high precision is also critical to avoid wasting valuable operational time on false positives. Given recent advancements in visual perception algorithms, the former is largely solvable autonomously, but the latter is difficult to address without the supervision of a human operator. However, operational constraints such as mission time, computational requirements, mesh network bandwidth and so on, can make the operator's task infeasible unless properly managed. We propose the Early Recall, Late Precision (EaRLaP) semantic object map** pipeline to solve this problem. EaRLaP was used by Team CoSTAR in DARPA Subterranean Challenge, where it successfully detected all the artifacts encountered by the team of robots. We will discuss these results and performance of the EaRLaP on various datasets.
△ Less
Submitted 20 June, 2022;
originally announced June 2022.
-
RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion
Authors:
Joaquin Royo-Miquel,
Silvia Tolu,
Frederik E. T. Schöller,
Roberto Galeazzi
Abstract:
The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with li…
▽ More
The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with limited loss in performance.
△ Less
Submitted 28 September, 2021; v1 submitted 10 June, 2021;
originally announced June 2021.
-
Trust as Extended Control: Active Inference and User Feedback During Human-Robot Collaboration
Authors:
Felix Schoeller,
Mark Miller,
Roy Salomon,
Karl J. Friston
Abstract:
To interact seamlessly with robots, users must infer the causes of a robot's behavior and be confident about that inference. Hence, trust is a necessary condition for human-robot collaboration (HRC). Despite its crucial role, it is largely unknown how trust emerges, develops, and supports human interactions with nonhuman artefacts. Here, we review the literature on trust, human-robot interaction,…
▽ More
To interact seamlessly with robots, users must infer the causes of a robot's behavior and be confident about that inference. Hence, trust is a necessary condition for human-robot collaboration (HRC). Despite its crucial role, it is largely unknown how trust emerges, develops, and supports human interactions with nonhuman artefacts. Here, we review the literature on trust, human-robot interaction, human-robot collaboration, and human interaction at large. Early models of trust suggest that trust entails a trade-off between benevolence and competence, while studies of human-to-human interaction emphasize the role of shared behavior and mutual knowledge in the gradual building of trust. We then introduce a model of trust as an agent's best explanation for reliable sensory exchange with an extended motor plant or partner. This model is based on the cognitive neuroscience of active inference and suggests that, in the context of HRC, trust can be cast in terms of virtual control over an artificial agent. In this setting, interactive feedback becomes a necessary component of the trustor's perception-action cycle. The resulting model has important implications for understanding human-robot interaction and collaboration, as it allows the traditional determinants of human trust to be defined in terms of active inference, information exchange and empowerment. Furthermore, this model suggests that boredom and surprise may be used as markers for under and over-reliance on the system. Finally, we examine the role of shared behavior in the genesis of trust, especially in the context of dyadic collaboration, suggesting important consequences for the acceptability and design of human-robot collaborative systems.
△ Less
Submitted 22 April, 2021;
originally announced April 2021.
-
Symmetries and Asymptotically Flat Space
Authors:
Friedrich Schöller
Abstract:
The construction of a theory of quantum gravity is an outstanding problem that can benefit from better understanding the laws of nature that are expected to hold in regimes currently inaccessible to experiment. Such fundamental laws can be found by considering the classical counterparts of a quantum theory. For example, conservation laws in a quantum theory often stem from conservation laws of the…
▽ More
The construction of a theory of quantum gravity is an outstanding problem that can benefit from better understanding the laws of nature that are expected to hold in regimes currently inaccessible to experiment. Such fundamental laws can be found by considering the classical counterparts of a quantum theory. For example, conservation laws in a quantum theory often stem from conservation laws of the corresponding classical theory. In order to construct such laws, this thesis is concerned with the interplay between symmetries and conservation laws of classical field theories and their application to asymptotically flat spacetimes.
This work begins with an explanation of symmetries in field theories with a focus on variational symmetries and their associated conservation laws. Boundary conditions for general relativity are then formulated on three-dimensional asymptotically flat spacetimes at null infinity using the method of conformal completion. Conserved quantities related to asymptotic symmetry transformations are derived and their properties are studied. This is done in a manifestly coordinate independent manner. In a separate step a coordinate system is introduced, such that the results can be compared to existing literature. Next, asymptotically flat spacetimes which contain both future as well as past null infinity are considered. Asymptotic symmetries occurring at these disjoint regions of three-dimensional asymptotically flat spacetimes are linked and the corresponding conserved quantities are matched. Finally, it is shown how asymptotic symmetries lead to the notion of distinct Minkowski spaces that can be differentiated by conserved quantities.
△ Less
Submitted 16 March, 2020;
originally announced March 2020.
-
Methods for suspensions of passive and active filaments
Authors:
S. F. Schoeller,
A. K. Townsend,
T. A. Westwood,
E. E. Keaveny
Abstract:
Flexible filaments and fibres are essential components of important complex fluids that appear in many biological and industrial settings. Direct simulations of these systems that capture the motion and deformation of many immersed filaments in suspension remain a formidable computational challenge due to the complex, coupled fluid--structure interactions of all filaments, the numerical stiffness…
▽ More
Flexible filaments and fibres are essential components of important complex fluids that appear in many biological and industrial settings. Direct simulations of these systems that capture the motion and deformation of many immersed filaments in suspension remain a formidable computational challenge due to the complex, coupled fluid--structure interactions of all filaments, the numerical stiffness associated with filament bending, and the various constraints that must be maintained as the filaments deform. In this paper, we address these challenges by describing filament kinematics using quaternions to resolve both bending and twisting, applying implicit time-integration to alleviate numerical stiffness, and using quasi-Newton methods to obtain solutions to the resulting system of nonlinear equations. In particular, we employ geometric time integration to ensure that the quaternions remain unit as the filaments move. We also show that our framework can be used with a variety of models and methods, including matrix-free fast methods, that resolve low Reynolds number hydrodynamic interactions. We provide a series of tests and example simulations to demonstrate the performance and possible applications of our method. Finally, we provide a link to a MATLAB/Octave implementation of our framework that can be used to learn more about our approach and as a tool for filament simulation.
△ Less
Submitted 24 September, 2020; v1 submitted 29 March, 2019;
originally announced March 2019.
-
All Weight Systems for Calabi-Yau Fourfolds from Reflexive Polyhedra
Authors:
Friedrich Schöller,
Harald Skarke
Abstract:
For any given dimension $d$, all reflexive $d$-polytopes can be found (in principle) as subpolytopes of a number of maximal polyhedra that are defined in terms of $(d+1)$-tuples of integers (weights), or combinations of $k$-tuples of weights with $k<d+1$. We present the results of a complete classification of sextuples of weights pertaining to the construction of all reflexive polytopes in five di…
▽ More
For any given dimension $d$, all reflexive $d$-polytopes can be found (in principle) as subpolytopes of a number of maximal polyhedra that are defined in terms of $(d+1)$-tuples of integers (weights), or combinations of $k$-tuples of weights with $k<d+1$. We present the results of a complete classification of sextuples of weights pertaining to the construction of all reflexive polytopes in five dimensions. We find 322 383 760 930 such weight systems. 185 269 499 015 of them give rise directly to reflexive polytopes and thereby to mirror pairs of Calabi-Yau fourfolds. These lead to 532 600 483 distinct sets of Hodge numbers.
△ Less
Submitted 7 August, 2018;
originally announced August 2018.
-
From flagellar undulations to collective motion: predicting the dynamics of sperm suspensions
Authors:
S. F. Schoeller,
E. E. Keaveny
Abstract:
Swimming cells and microorganisms are as diverse in their collective dynamics as they are in their individual shapes and propulsion mechanisms. Even for sperm cells, which have a stereotyped shape consisting of a cell body connected to a flexible flagellum, a wide range of collective dynamics is observed spanning from the formation of tightly packed groups to the display of larger-scale, turbulenc…
▽ More
Swimming cells and microorganisms are as diverse in their collective dynamics as they are in their individual shapes and propulsion mechanisms. Even for sperm cells, which have a stereotyped shape consisting of a cell body connected to a flexible flagellum, a wide range of collective dynamics is observed spanning from the formation of tightly packed groups to the display of larger-scale, turbulence-like motion. Using a detailed mathematical model that resolves flagellum dynamics, we perform simulations of sperm suspensions containing up to 1000 cells and explore the connection between individual and collective dynamics. We find that depending on the level of variation in individual dynamics from one swimmer to another, the sperm exhibit either a strong tendency to aggregate, or the suspension exhibits large-scale swirling. Hydrodynamic interactions govern the formation and evolution of both states. In addition, a quantitative analysis of the states reveals that the flows generated at the time-scale of flagellum undulations contribute significantly to the overall energy in the surrounding fluid, highlighting the importance of resolving these flows.
△ Less
Submitted 24 January, 2018;
originally announced January 2018.
-
Distinct Minkowski Spaces from BMS Supertranslations
Authors:
Friedrich Schöller
Abstract:
This work provides a smooth and everywhere well-defined extension of Bondi-Metzner-Sachs (BMS) supertranslations into the bulk of Minkowski space. The supertranslations lead to physically distinct spacetimes, all isometric to Minkowski space. This construction is in contrast to the often used, non-smooth BMS transformations that appear in a gauge-fixed description of the theory.
This work provides a smooth and everywhere well-defined extension of Bondi-Metzner-Sachs (BMS) supertranslations into the bulk of Minkowski space. The supertranslations lead to physically distinct spacetimes, all isometric to Minkowski space. This construction is in contrast to the often used, non-smooth BMS transformations that appear in a gauge-fixed description of the theory.
△ Less
Submitted 18 February, 2018; v1 submitted 7 November, 2017;
originally announced November 2017.
-
Linking Past and Future Null Infinity in Three Dimensions
Authors:
Stefan Prohazka,
Jakob Salzer,
Friedrich Schöller
Abstract:
We provide a map** between past null and future null infinity in three-dimensional flat space, using symmetry considerations. From this we derive a map** between the corresponding asymptotic symmetry groups. By studying the metric at asymptotic regions, we find that the map** is energy preserving and yields an infinite number of conservation laws.
We provide a map** between past null and future null infinity in three-dimensional flat space, using symmetry considerations. From this we derive a map** between the corresponding asymptotic symmetry groups. By studying the metric at asymptotic regions, we find that the map** is energy preserving and yields an infinite number of conservation laws.
△ Less
Submitted 14 April, 2017; v1 submitted 23 January, 2017;
originally announced January 2017.
-
Flat space cosmologies in two dimensions - Phase transitions and asymptotic mass-domination
Authors:
Arjun Bagchi,
Daniel Grumiller,
Jakob Salzer,
Sourav Sarkar,
Friedrich Schöller
Abstract:
We study flat space cosmologies in two dimensions by taking the flat space limit of the Achucarro-Ortiz model. We unravel a phase transition between hot flat space and flat space cosmologies, and derive a new dilaton-dependent counterterm required for the consistency of the Euclidean partition function. Our results generalize to asymptotically mass-dominated 2-dimensional dilaton gravity models, w…
▽ More
We study flat space cosmologies in two dimensions by taking the flat space limit of the Achucarro-Ortiz model. We unravel a phase transition between hot flat space and flat space cosmologies, and derive a new dilaton-dependent counterterm required for the consistency of the Euclidean partition function. Our results generalize to asymptotically mass-dominated 2-dimensional dilaton gravity models, whose thermodynamical properties we discuss. The novel case of asymptotic mass-domination is neither covered by the comprehensive discussion of hep-th/0703230 nor by the more recent generalization to dilaton gravity with confining U(1) charges in 1406.7007.
△ Less
Submitted 22 August, 2014;
originally announced August 2014.
-
Variational principle and 1-point functions in 3-dimensional flat space Einstein gravity
Authors:
Stephane Detournay,
Daniel Grumiller,
Friedrich Scholler,
Joan Simon
Abstract:
We provide a well-defined variational principle for 3-dimensional flat space Einstein gravity by adding one half of the Gibbons-Hawking-York boundary term to the bulk action. We check the 0-point function, recovering consistency with thermodynamics of flat space cosmologies. We then apply our result to calculate the 1-point functions in flat space Einstein gravity for the vacuum and all flat space…
▽ More
We provide a well-defined variational principle for 3-dimensional flat space Einstein gravity by adding one half of the Gibbons-Hawking-York boundary term to the bulk action. We check the 0-point function, recovering consistency with thermodynamics of flat space cosmologies. We then apply our result to calculate the 1-point functions in flat space Einstein gravity for the vacuum and all flat space cosmologies. The results are compatible with the ones for the zero mode charges obtained by canonical analysis.
△ Less
Submitted 26 February, 2014; v1 submitted 15 February, 2014;
originally announced February 2014.
