Sebastian Trimpe
YOU?
Author Swipe
View article: RockNet: Distributed Learning on Ultra-Low-Power Devices
RockNet: Distributed Learning on Ultra-Low-Power Devices Open
As Machine Learning (ML) becomes integral to Cyber-Physical Systems (CPS), there is growing interest in shifting training from traditional cloud-based to on-device processing (TinyML), for example, due to privacy and latency concerns. Howe…
View article: Concept Extraction for Time Series with ECLAD-ts
Concept Extraction for Time Series with ECLAD-ts Open
Convolutional neural networks (CNNs) for time series classification (TSC) are being increasingly used in applications ranging from quality prediction to medical diagnosis. The black box nature of these models makes understanding their pred…
View article: DMPC-Swarm: distributed model predictive control on nano UAV swarms
DMPC-Swarm: distributed model predictive control on nano UAV swarms Open
Swarms of unmanned aerial vehicles (UAVs) are increasingly becoming vital to our society, undertaking tasks such as search and rescue, surveillance and delivery. A special variant of Distributed Model Predictive Control (DMPC) has emerged …
View article: Bayesian optimization for extreme high-speed laser material deposition
Bayesian optimization for extreme high-speed laser material deposition Open
Extreme high-speed laser material deposition, also known by the acronym EHLA, enables metallic coatings of different thicknesses at deposition speeds of up to several hundred meters per minute and deposition rates of several kilograms per …
View article: Robust screening of atrial fibrillation with distribution classification
Robust screening of atrial fibrillation with distribution classification Open
View article: MPX: Mixed Precision Training for JAX
MPX: Mixed Precision Training for JAX Open
Mixed-precision training has emerged as an indispensable tool for enhancing the efficiency of neural network training in recent years. Concurrently, JAX has grown in popularity as a versatile machine learning toolbox. However, it currently…
View article: Kernel conditional tests from learning-theoretic bounds
Kernel conditional tests from learning-theoretic bounds Open
We propose a framework for hypothesis testing on conditional probability distributions, which we then use to construct statistical tests of functionals of conditional distributions. These tests identify the inputs where the functionals dif…
View article: Event-triggered Robust Model Predictive Control under Hard Computation Resource Constraints
Event-triggered Robust Model Predictive Control under Hard Computation Resource Constraints Open
Model predictive control (MPC) is capable of controlling nonlinear systems with guaranteed constraint satisfaction and stability. However, MPC requires solving optimization problems online periodically, which often exceeds the local system…
View article: Concept Extraction for Time Series with ECLAD-ts
Concept Extraction for Time Series with ECLAD-ts Open
Convolutional neural networks (CNNs) for time series classification (TSC) are being increasingly used in applications ranging from quality prediction to medical diagnosis. The black box nature of these models makes understanding their pred…
View article: Diffusion-Based Approximate MPC: Fast and Consistent Imitation of Multi-Modal Action Distributions
Diffusion-Based Approximate MPC: Fast and Consistent Imitation of Multi-Modal Action Distributions Open
Approximating model predictive control (MPC) using imitation learning (IL) allows for fast control without solving expensive optimization problems online. However, methods that use neural networks in a simple L2-regression setup fail to ap…
View article: The Mini Wheelbot: A Testbed for Learning-based Balancing, Flips, and Articulated Driving
The Mini Wheelbot: A Testbed for Learning-based Balancing, Flips, and Articulated Driving Open
The Mini Wheelbot is a balancing, reaction wheel unicycle robot designed as a testbed for learning-based control. It is an unstable system with highly nonlinear yaw dynamics, non-holonomic driving, and discrete contact switches in a small,…
View article: On Rollouts in Model-Based Reinforcement Learning
On Rollouts in Model-Based Reinforcement Learning Open
Model-based reinforcement learning (MBRL) seeks to enhance data efficiency by learning a model of the environment and generating synthetic rollouts from it. However, accumulated model errors during these rollouts can distort the data distr…
View article: Safety in safe Bayesian optimization and its ramifications for control
Safety in safe Bayesian optimization and its ramifications for control Open
A recurring and important task in control engineering is parameter tuning under constraints, which conceptually amounts to optimization of a blackbox function accessible only through noisy evaluations. For example, in control practice para…
View article: Lipschitz Safe Bayesian Optimization for Automotive Control
Lipschitz Safe Bayesian Optimization for Automotive Control Open
Controller tuning is a labor-intensive process that requires human intervention and expert knowledge. Bayesian optimization has been applied successfully in different fields to automate this process. However, when tuning on hardware, such …
View article: Early Stopping Bayesian Optimization for Controller Tuning
Early Stopping Bayesian Optimization for Controller Tuning Open
Manual tuning of performance-critical controller parameters can be tedious and sub-optimal. Bayesian Optimization (BO) is an increasingly popular practical alternative to automatically optimize controller parameters from few experiments. S…
View article: CHEQ-ing the Box: Safe Variable Impedance Learning for Robotic Polishing
CHEQ-ing the Box: Safe Variable Impedance Learning for Robotic Polishing Open
Robotic systems are increasingly employed for industrial automation, with contact-rich tasks like polishing requiring dexterity and compliant behaviour. These tasks are difficult to model, making classical control challenging. Deep reinfor…
View article: Robust direct data-driven control for probabilistic systems
Robust direct data-driven control for probabilistic systems Open
View article: Simulation-Aided Policy Tuning for Black-Box Robot Learning
Simulation-Aided Policy Tuning for Black-Box Robot Learning Open
How can robots learn and adapt to new tasks and situations with little data? Systematic exploration and simulation are crucial tools for efficient robot learning. We present a novel black-box policy search algorithm focused on data-efficie…
View article: Newtonian and Lagrangian Neural Networks: A Comparison Towards Efficient Inverse Dynamics Identification
Newtonian and Lagrangian Neural Networks: A Comparison Towards Efficient Inverse Dynamics Identification Open
View article: Learning Deformable Linear Object Dynamics from a Single Trajectory
Learning Deformable Linear Object Dynamics from a Single Trajectory Open
sponsorship: The work of Shamil Mamedov was supported by RWTH Aachen University, FWO-Vlaanderen through SBO Project ELYSA for Cobot Applications, under Grant S001821 N. (RWTH Aachen University, FWO-Vlaanderen through SBO Project ELYSA for …
View article: Twin-in-the-Loop Observers Tuning via Gradient-Information Bayesian Optimization with Line Search
Twin-in-the-Loop Observers Tuning via Gradient-Information Bayesian Optimization with Line Search Open
View article: CHEQ-ing the Box: Safe Variable Impedance Learning for Robotic Polishing
CHEQ-ing the Box: Safe Variable Impedance Learning for Robotic Polishing Open
View article: Bayesian Optimization via Continual Variational Last Layer Training
Bayesian Optimization via Continual Variational Last Layer Training Open
Gaussian Processes (GPs) are widely seen as the state-of-the-art surrogate models for Bayesian optimization (BO) due to their ability to model uncertainty and their performance on tasks where correlations are easily captured (such as those…
View article: On Foundation Models for Dynamical Systems from Purely Synthetic Data
On Foundation Models for Dynamical Systems from Purely Synthetic Data Open
Foundation models have demonstrated remarkable generalization, data efficiency, and robustness properties across various domains. In this paper, we explore the feasibility of foundation models for applications in the control domain. The su…
View article: Recent kernel methods for interacting particle systems: first numerical results
Recent kernel methods for interacting particle systems: first numerical results Open
Interacting particle systems (IPSs) are a very important class of dynamical systems, arising in different domains like biology, physics, sociology and engineering. In many applications, these systems can be very large, making their simulat…
View article: Feedforward Controllers from Learned Dynamic Local Model Networks with Application to Excavator Assistance Functions
Feedforward Controllers from Learned Dynamic Local Model Networks with Application to Excavator Assistance Functions Open
Complicated first principles modelling and controller synthesis can be prohibitively slow and expensive for high-mix, low-volume products such as hydraulic excavators. Instead, in a data-driven approach, recorded trajectories from the real…
View article: Learning deformable linear object dynamics from a single trajectory
Learning deformable linear object dynamics from a single trajectory Open
The manipulation of deformable linear objects (DLOs) via model-based control requires an accurate and computationally efficient dynamics model. Yet, data-driven DLO dynamics models require large training data sets while their predictions o…
View article: Contextualized Hybrid Ensemble Q-learning: Learning Fast with Control Priors
Contextualized Hybrid Ensemble Q-learning: Learning Fast with Control Priors Open
Combining Reinforcement Learning (RL) with a prior controller can yield the best out of two worlds: RL can solve complex nonlinear problems, while the control prior ensures safer exploration and speeds up training. Prior work largely blend…
View article: Combining Automated Optimisation of Hyperparameters and Reward Shape
Combining Automated Optimisation of Hyperparameters and Reward Shape Open
There has been significant progress in deep reinforcement learning (RL) in recent years. Nevertheless, finding suitable hyperparameter configurations and reward functions remains challenging even for experts, and performance heavily relies…
View article: On the Consistency of Kernel Methods with Dependent Observations
On the Consistency of Kernel Methods with Dependent Observations Open
The consistency of a learning method is usually established under the assumption that the observations are a realization of an independent and identically distributed (i.i.d.) or mixing process. Yet, kernel methods such as support vector m…