Jonas Frey
YOU?
Author Swipe
View article: Offline vs. Online Learning in Model-based RL: Lessons for Data Collection Strategies
Offline vs. Online Learning in Model-based RL: Lessons for Data Collection Strategies Open
Data collection is crucial for learning robust world models in model-based reinforcement learning. The most prevalent strategies are to actively collect trajectories by interacting with the environment during online training or training on…
View article: Boxi: Design Decisions in the Context of Algorithmic Performance for Robotics
Boxi: Design Decisions in the Context of Algorithmic Performance for Robotics Open
Achieving robust autonomy in mobile robots operating in complex and unstructured environments requires a multimodal sensor suite capable of capturing diverse and complementary information. However, designing such a sensor suite involves mu…
View article: Zero-Shot Offline Imitation Learning via Optimal Transport
Zero-Shot Offline Imitation Learning via Optimal Transport Open
Zero-shot imitation learning algorithms hold the promise of reproducing unseen behavior from as little as a single demonstration at test time. Existing practical approaches view the expert demonstration as a sequence of goals, enabling imi…
View article: Autonomous Forest Inventory with Legged Robots: System Design and Field Deployment
Autonomous Forest Inventory with Legged Robots: System Design and Field Deployment Open
We present a solution for autonomous forest inventory with a legged robotic platform. Compared to their wheeled and aerial counterparts, legged platforms offer an attractive balance of endurance and low soil impact for forest applications.…
View article: Uniform Preorders and Partial Combinatory Algebras
Uniform Preorders and Partial Combinatory Algebras Open
Uniform preorders are a class of combinatory representations of Set-indexed preorders that generalize Pieter Hofstra's basic relational objects. An indexed preorder is representable by a uniform preorder if and only if it has as generic pr…
View article: RoadRunner -- Learning Traversability Estimation for Autonomous Off-road Driving
RoadRunner -- Learning Traversability Estimation for Autonomous Off-road Driving Open
Autonomous navigation at high speeds in off-road environments necessitates robots to comprehensively understand their surroundings using onboard sensing only. The extreme conditions posed by the off-road setting can cause degraded camera i…
View article: Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-to-End
Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-to-End Open
Autonomous robots must navigate reliably in unknown environments even under compromised exteroceptive perception, or perception failures. Such failures often occur when harsh environments lead to degraded sensing, or when the perception al…
View article: MEM: Multi-Modal Elevation Mapping for Robotics and Learning
MEM: Multi-Modal Elevation Mapping for Robotics and Learning Open
Elevation maps are commonly used to represent the environment of mobile robots and are instrumental for locomotion and navigation tasks. However, pure geometric information is insufficient for many field applications that require appearanc…
View article: Learning Risk-Aware Quadrupedal Locomotion using Distributional Reinforcement Learning
Learning Risk-Aware Quadrupedal Locomotion using Distributional Reinforcement Learning Open
Deployment in hazardous environments requires robots to understand the risks associated with their actions and movements to prevent accidents. Despite its importance, these risks are not explicitly modeled by currently deployed locomotion …
View article: Duality for Clans: an Extension of Gabriel-Ulmer Duality
Duality for Clans: an Extension of Gabriel-Ulmer Duality Open
Clans are representations of generalized algebraic theories that contain more information than the finite-limit categories associated to the locally finitely presentable categories of models via Gabriel-Ulmer duality. Extending Gabriel-Ulm…
View article: Fast Traversability Estimation for Wild Visual Navigation
Fast Traversability Estimation for Wild Visual Navigation Open
Fig. 1: Wild Visual Navigation (WVN) learns to predict traversability from images via online self-supervised learning.Starting from a randomly initialized traversability estimation network without prior assumptions about the environment (a…
View article: SMUG Planner: A Safe Multi-Goal Planner for Mobile Robots in Challenging Environments
SMUG Planner: A Safe Multi-Goal Planner for Mobile Robots in Challenging Environments Open
Robotic exploration or monitoring missions require mobile robots to autonomously and safely navigate between multiple target locations in potentially challenging environments. Currently, this type of multi-goal mission often relies on huma…
View article: Seeing Through the Grass: Semantic Pointcloud Filter for Support Surface Learning
Seeing Through the Grass: Semantic Pointcloud Filter for Support Surface Learning Open
Mobile ground robots require perceiving and understanding their surrounding support surface to move around autonomously and safely. The support surface is commonly estimated based on exteroceptive depth measurements, e.g., from LiDARs. How…
View article: Constructing coproducts in locally Cartesian closed $\infty$-categories
Constructing coproducts in locally Cartesian closed $\infty$-categories Open
We prove that every locally Cartesian closed oo-category with a subobject classifier has a strict initial object and disjoint and universal binary coproducts.
View article: Fast Traversability Estimation for Wild Visual Navigation
Fast Traversability Estimation for Wild Visual Navigation Open
Natural environments such as forests and grasslands are challenging for robotic navigation because of the false perception of rigid obstacles from high grass, twigs, or bushes. In this work, we propose Wild Visual Navigation (WVN), an onli…
View article: Unsupervised Continual Semantic Adaptation through Neural Rendering
Unsupervised Continual Semantic Adaptation through Neural Rendering Open
An increasing amount of applications rely on data-driven models that are deployed for perception tasks across a sequence of scenes. Due to the mismatch between training and deployment data, adapting the model on the new scenes is often cru…
View article: Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments
Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments Open
Despite the progress in legged robotic locomotion, autonomous navigation in unknown environments remains an open problem. Ideally, the navigation system utilizes the full potential of the robots’ locomotion capabilities while operating wit…
View article: Versatile Skill Control via Self-supervised Adversarial Imitation of Unlabeled Mixed Motions
Versatile Skill Control via Self-supervised Adversarial Imitation of Unlabeled Mixed Motions Open
Learning diverse skills is one of the main challenges in robotics. To this end, imitation learning approaches have achieved impressive results. These methods require explicitly labeled datasets or assume consistent skill execution to enabl…
View article: Learning Agile Skills via Adversarial Imitation of Rough Partial Demonstrations
Learning Agile Skills via Adversarial Imitation of Rough Partial Demonstrations Open
Learning agile skills is one of the main challenges in robotics. To this end, reinforcement learning approaches have achieved impressive results. These methods require explicit task information in terms of a reward function or an expert th…
View article: Categories of partial equivalence relations as localizations
Categories of partial equivalence relations as localizations Open
We construct a category of fibrant objects C〈P〉 in the sense of K. Brown from any indexed frame (a kind of indexed poset generalizing triposes) P, and show that its homotopy category is the Barr-exact category C[P] of partial equivalence r…
View article: Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments
Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments Open
Despite the progress in legged robotic locomotion, autonomous navigation in unknown environments remains an open problem. Ideally, the navigation system utilizes the full potential of the robots' locomotion capabilities while operating wit…
View article: Continual Adaptation of Semantic Segmentation using Complementary 2D-3D Data Representations
Continual Adaptation of Semantic Segmentation using Complementary 2D-3D Data Representations Open
Semantic segmentation networks are usually pre-trained once and not updated during deployment. As a consequence, misclassifications commonly occur if the distribution of the training data deviates from the one encountered during the robot'…
View article: Constructing Coproducts in locally Cartesian closed $\infty$-Categories
Constructing Coproducts in locally Cartesian closed $\infty$-Categories Open
We prove that every locally Cartesian closed $\infty$-category with subobject classifier has a strict initial object and disjoint and universal binary coproducts.
View article: Triposes as a generalization of localic geometric morphisms
Triposes as a generalization of localic geometric morphisms Open
In Hyland et al. (1980), Hyland, Johnstone and Pitts introduced the notion of tripos for the purpose of organizing the construction of realizability toposes in a way that generalizes the construction of localic toposes from complete Heytin…
View article: Triposes as a Generalization of Localic Geometric Morphisms
Triposes as a Generalization of Localic Geometric Morphisms Open
We discuss how triposes may be understood as generalizations of localic geometric morphisms.
View article: Categories of partial equivalence relations as localizations
Categories of partial equivalence relations as localizations Open
We construct a category of fibrant objects $\mathbb{C}\langle P\rangle$ in the sense of K. Brown from any indexed frame (a kind of indexed poset generalizing triposes) $P$, and show that its homotopy category is the Barr-exact category $\m…
View article: Impredicative Encodings of (Higher) Inductive Types
Impredicative Encodings of (Higher) Inductive Types Open
Postulating an impredicative universe in dependent type theory allows System F style encodings of finitary inductive types, but these fail to satisfy the relevant η-equalities and consequently do not admit dependent eliminators. To recover…
View article: Impredicative Encodings of (Higher) Inductive Types
Impredicative Encodings of (Higher) Inductive Types Open
Postulating an impredicative universe in dependent type theory allows System F style encodings of finitary inductive types, but these fail to satisfy the relevant η-equalities and consequently do not admit dependent eliminators. To recover…
View article: Classical Realizability in the CPS Target Language
Classical Realizability in the CPS Target Language Open
Motivated by considerations about Krivine's classical realizability, we introduce a term calculus for an intuitionistic logic with record types, which we call the CPS target language. We give a reformulation of the constructions of classic…