arXiv subjects

Jerome Sieber

Publications and source records attributed to Jerome Sieber.

17 recordsLinked to original sources

arXiv2026

Eigenvalues as a Metric for Memory Dynamics in Sequence Models

While softmax attention drives state-of-the-art performance in sequence modeling, its quadratic complexity motivates linear alternatives such as state space models (SSMs). Structural differences between the two model classes, however, hinder direct comparisons of their memory dynamics, creating the need for a common metric to analyze, interpret, and improve their information processing capabilities. Inspired by recent advances in SSM performance driven by eigenvalue-guided insights, we leverage the dynamical systems framework to bring attention models into a unified analytical framework with SSMs. This allows us to perform a structured analysis, which investigates the applicability of an eigenvalue-spectrum memory dynamics metric to attention models. To this end, we first conduct an extensive empirical study across diverse attention-based models and SSMs on a range of benchmarks. We show that, for both model classes, eigenvalues influence key aspects of memory and long-range dependency modeling, revealing spectral signatures that align with task requirements. Building on these findings, we show how spectral signatures can motivate architectural modifications, how they can be guided through the training process, and how they can provide information about feature importance. The results thereby enable and emphasize the role of eigenvalue analysis as a principled metric for interpreting, explaining, and ultimately improving the capabilities of sequence models.

cs.LG↗

arXiv2026

Stochastic MPC with Online-optimized Policies and Closed-loop Guarantees

This paper proposes a stochastic model predictive control method for linear systems affected by additive Gaussian disturbances that optimizes over disturbance feedback matrices online. Closed-loop satisfaction of probabilistic constraints and recursive feasibility of the underlying convex optimization problem is guaranteed. Optimization over feedback policies online increases performance and reduces conservatism compared to fixed-feedback approaches. The central mechanism is a finitely determined maximal admissible set for probabilistic constraints, together with the reconditioning of the predicted probabilistic constraints on the current knowledge at every time step. The proposed method's applicability is demonstrated on a building temperature control example.

eess.SY↗

arXiv2026

A Predictive Law for On-Policy Self-Distillation From World Feedback

Moving beyond simple scalar rewards toward richer world feedback is a natural path to more scalable RL post-training. On-policy self-distillation (OPSD) is a promising recent approach that uses arbitrary feedback as learning signal, yet its reliability compared to established methods, such as GRPO, remains unclear. We identify a strikingly consistent linear correlation between the initial student-self-teacher performance gap and the final performance improvement in OPSD. This relationship holds across context types and model families, providing a powerful predictive law for anticipating the outcome of an OPSD configuration without running the full training procedure. Interestingly, we show that this linear predictability holds with model scale, suggesting a potential basis for new empirical scaling laws on larger models with stronger in-context learning capabilities. In essence, our findings show that OPSD performance can be predicted and tuned before training, offering a principled way to incorporate world feedback as a first-class component of the post-training pipeline.

cs.LG↗

arXiv2026

Design Principles for Sequence Models via Coefficient Dynamics

Deep sequence models, ranging from Transformers and State Space Models (SSMs) to more recent approaches such as gated linear RNNs, fundamentally compute outputs as linear combinations of past value vectors. To draw insights and systematically compare such architectures, we develop a unified framework that makes this output operation explicit, by casting the linear combination coefficients as the outputs of autonomous linear dynamical systems driven by impulse inputs. This viewpoint, in spirit substantially different from approaches focusing on connecting linear RNNs with linear attention, reveals a common mathematical theme across diverse architectures and crucially captures softmax attention, on top of RNNs, SSMs, and related models. In contrast to new model proposals that are commonly evaluated on benchmarks, we derive design principles linking architectural choices to model properties. Thereby identifying tradeoffs between expressivity and efficient implementation, geometric constraints on input selectivity, and stability conditions for numerically stable training and information retention. By connecting several insights and observations from recent literature, the framework both explains empirical successes of recent designs and provides guiding principles for systematically designing new sequence model architectures.

cs.LG↗

arXiv2025

Computationally Efficient System Level Tube-MPC for Uncertain Systems

Tube-based model predictive control (MPC) is one of the principal robust control techniques for constrained linear systems affected by additive disturbances. While tube-based methods with online-computed tubes have been successfully applied to systems with additive disturbances, their application to systems affected by additional model uncertainties is challenging. This paper proposes a tube-based MPC method - named filter-based system level tube-MPC (SLTMPC) - which overapproximates both types of uncertainties with an online optimized disturbance set, while simultaneously computing the tube controller online. For the first time, we provide rigorous closed-loop guarantees for receding horizon control of such a MPC method. These guarantees are obtained by virtue of a new terminal controller design and an online optimized terminal set. To reduce the computational complexity of the proposed method, we additionally introduce an asynchronous computation scheme that separates the optimization of the tube controller and the nominal trajectory. Finally, we provide a comprehensive numerical evaluation of the proposed methods to demonstrate their effectiveness.

eess.SY↗

arXiv2025

Lambda-Skip Connections: the architectural component that prevents Rank Collapse

Rank collapse, a phenomenon where embedding vectors in sequence models rapidly converge to a uniform token or equilibrium state, has recently gained attention in the deep learning literature. This phenomenon leads to reduced expressivity and potential training instabilities due to vanishing gradients. Empirical evidence suggests that architectural components like skip connections, LayerNorm, and MultiLayer Perceptrons (MLPs) play critical roles in mitigating rank collapse. While this issue is well-documented for transformers, alternative sequence models, such as State Space Models (SSMs), which have recently gained prominence, have not been thoroughly examined for similar vulnerabilities. This paper extends the theory of rank collapse from transformers to SSMs using a unifying framework that captures both architectures. We study how a parametrized version of the classic skip connection component, which we call \emph{lambda-skip connections}, provides guarantees for rank collapse prevention. Through analytical results, we present a sufficient condition to guarantee prevention of rank collapse across all the aforementioned architectures. We also study the necessity of this condition via ablation studies and analytical examples. To our knowledge, this is the first study that provides a general guarantee to prevent rank collapse, and that investigates rank collapse in the context of SSMs, offering valuable understanding for both theoreticians and practitioners. Finally, we validate our findings with experiments demonstrating the crucial role of architectural components such as skip connections and gating mechanisms in preventing rank collapse.

cs.LG↗

arXiv2024

Understanding the differences in Foundation Models: Attention, State Space Models, and Recurrent Neural Networks

Softmax attention is the principle backbone of foundation models for various artificial intelligence applications, yet its quadratic complexity in sequence length can limit its inference throughput in long-context settings. To address this challenge, alternative architectures such as linear attention, State Space Models (SSMs), and Recurrent Neural Networks (RNNs) have been considered as more efficient alternatives. While connections between these approaches exist, such models are commonly developed in isolation and there is a lack of theoretical understanding of the shared principles underpinning these architectures and their subtle differences, greatly influencing performance and scalability. In this paper, we introduce the Dynamical Systems Framework (DSF), which allows a principled investigation of all these architectures in a common representation. Our framework facilitates rigorous comparisons, providing new insights on the distinctive characteristics of each model class. For instance, we compare linear attention and selective SSMs, detailing their differences and conditions under which both are equivalent. We also provide principled comparisons between softmax attention and other model classes, discussing the theoretical conditions under which softmax attention can be approximated. Additionally, we substantiate these new insights with empirical validations and mathematical arguments. This shows the DSF's potential to guide the systematic development of future more efficient and scalable foundation models.

cs.LG↗

arXiv2024

State Space Models as Foundation Models: A Control Theoretic Overview

In recent years, there has been a growing interest in integrating linear state-space models (SSM) in deep neural network architectures of foundation models. This is exemplified by the recent success of Mamba, showing better performance than the state-of-the-art Transformer architectures in language tasks. Foundation models, like e.g. GPT-4, aim to encode sequential data into a latent space in order to learn a compressed representation of the data. The same goal has been pursued by control theorists using SSMs to efficiently model dynamical systems. Therefore, SSMs can be naturally connected to deep sequence modeling, offering the opportunity to create synergies between the corresponding research areas. This paper is intended as a gentle introduction to SSM-based architectures for control theorists and summarizes the latest research developments. It provides a systematic review of the most successful SSM proposals and highlights their main features from a control theoretic perspective. Additionally, we present a comparative analysis of these models, evaluating their performance on a standardized benchmark designed for assessing a model's efficiency at learning long sequences.

eess.SY↗

arXiv2023

Chronos and CRS: Design of a miniature car-like robot and a software framework for single and multi-agent robotics and control

From both an educational and research point of view, experiments on hardware are a key aspect of robotics and control. In the last decade, many open-source hardware and software frameworks for wheeled robots have been presented, mainly in the form of unicycles and car-like robots, with the goal of making robotics accessible to a wider audience and to support control systems development. Unicycles are usually small and inexpensive, and therefore facilitate experiments in a larger fleet, but they are not suited for high-speed motion. Car-like robots are more agile, but they are usually larger and more expensive, thus requiring more resources in terms of space and money. In order to bridge this gap, we present Chronos, a new car-like 1/28th scale robot with customized open-source electronics, and CRS, an open-source software framework for control and robotics. The CRS software framework includes the implementation of various state-of-the-art algorithms for control, estimation, and multi-agent coordination. With this work, we aim to provide easier access to hardware and reduce the engineering time needed to start new educational and research projects.

cs.RO↗

arXiv2023

Robust Optimal Control for Nonlinear Systems with Parametric Uncertainties via System Level Synthesis

This paper addresses the problem of optimally controlling nonlinear systems with norm-bounded disturbances and parametric uncertainties while robustly satisfying constraints. The proposed approach jointly optimizes a nominal nonlinear trajectory and an error feedback, requiring minimal offline design effort and offering low conservatism. This is achieved by decomposing the affine-in-the-parameter uncertain nonlinear system into a nominal $\textit{nonlinear}$ system and an uncertain linear time-varying system. Using this decomposition, we can apply established tools from system level synthesis to $\textit{convexly}$ over-bound all uncertainties in the nonlinear optimization problem. Moreover, it enables tight joint optimization of the linearization error bounds, parametric uncertainties bounds, nonlinear trajectory, and error feedback. With this novel controller parameterization, we can formulate a convex constraint to ensure robust performance guarantees for the nonlinear system. The presented method is relevant for numerous applications related to trajectory optimization, e.g., in robotics and aerospace engineering. We demonstrate the performance of the approach and its low conservatism through the simulation example of a post-capture satellite stabilization.

eess.SY↗

arXiv2023

Approximate Predictive Control Barrier Functions using Neural Networks: A Computationally Cheap and Permissive Safety Filter

A predictive control barrier function (PCBF) based safety filter is a modular framework to verify safety of a control input by predicting a future trajectory. The approach relies on the solution of two optimization problems, first computing the minimal state constraint violation given the current state in the form of slacks on the constraint, and then computing the minimal deviation from a proposed input given the previously computed minimal slacks. This paper presents an approximation procedure that uses a neural network to approximate the optimal value function of the first optimization problem, which defines a control barrier function (CBF). By including this explicit approximation in a CBF-based safety filter formulation, the online computation becomes independent of the prediction horizon. It is shown that this approximation guarantees convergence to a neighborhood of the feasible set of the PCBF safety filter problem with zero constraint violation. The convergence result relies on a novel class $\mathcal{K}$ lower bound on the PCBF decrease and depends on the approximation error of the neural network. Lastly, we demonstrate our approach in simulation for an autonomous driving example and show that the proposed approximation leads to a significant decrease in computation time compared to the original approach.

eess.SY↗

arXiv2023

Asynchronous Computation of Tube-based Model Predictive Control

Tube-based model predictive control (MPC) methods bound deviations from a nominal trajectory due to uncertainties in order to ensure constraint satisfaction. While techniques that compute the tubes online reduce conservativeness and increase performance, they suffer from high and potentially prohibitive computational complexity. This paper presents an asynchronous computation mechanism for system level tube-MPC (SLTMPC), a recently proposed tube-based MPC method which optimizes over both the nominal trajectory and the tubes. Computations are split into a primary and a secondary process, computing the nominal trajectory and the tubes, respectively. This enables running the primary process at a high frequency and moving the computationally complex tube computations to the secondary process. We show that the secondary process can continuously update the tubes, while retaining recursive feasibility of the primary process.

eess.SY↗

arXiv2022

A System Level Approach to Regret Optimal Control

We present an optimisation-based method for synthesising a dynamic regret optimal controller for linear systems with potentially adversarial disturbances and known or adversarial initial conditions. The dynamic regret is defined as the difference between the true incurred cost of the system and the cost which could have optimally been achieved under any input sequence having full knowledge of all future disturbances for a given disturbance energy. This problem formulation can be seen as an alternative to classical $\mathcal{H}_2$- or $\mathcal{H}_\infty$-control. The proposed controller synthesis is based on the system level parametrisation, which allows reformulating the dynamic regret problem as a semi-definite problem. This yields a new framework that allows to consider structured dynamic regret problems, which have not yet been considered in the literature. For known pointwise ellipsoidal bounds on the disturbance, we show that the dynamic regret bound can be improved compared to using only a bounded energy assumption and that the optimal dynamic regret bound differs by at most a factor of $\frac{2}π$ from the computed solution. Furthermore, the proposed framework allows guaranteeing state and input constraint satisfaction.

eess.SY↗

arXiv2021

System Level Disturbance Reachable Sets and their Application to Tube-based MPC

Tube-based model predictive control (MPC) methods leverage tubes to bound deviations from a nominal trajectory due to uncertainties in order to ensure constraint satisfaction. This paper presents a novel tube-based MPC formulation based on system level disturbance reachable sets (SL-DRS), which leverage the affine system level parameterization (SLP). We show that imposing a finite impulse response (FIR) constraint on the affine SLP guarantees containment of all future deviations in a finite sequence of SL-DRS. This allows us to formulate a system level tube-MPC (SLTMPC) method using the SL-DRS as tubes, which enables concurrent optimization of the nominal trajectory and the tubes, while using a positively invariant terminal set. Finally, we show that the SL-DRS tubes can also be computed offline.

eess.SY↗

arXiv2021

A System Level Approach to Tube-based Model Predictive Control

Robust tube-based model predictive control (MPC) methods address constraint satisfaction by leveraging an a priori determined tube controller in the prediction to tighten the constraints. This paper presents a system level tube-MPC (SLTMPC) method derived from the system level parameterization (SLP), which allows optimization over the tube controller online when solving the MPC problem, which can significantly reduce conservativeness. We derive the SLTMPC method by establishing an equivalence relation between a class of robust MPC methods and the SLP. Finally, we show that the SLTMPC formulation naturally arises from an extended SLP formulation and show its merits in a numerical example.

eess.SY↗

arXiv2021

Design, Optimal Guidance and Control of a Low-cost Re-usable Electric Model Rocket

In the last decade, autonomous vertical take-off and landing (VTOL) vehicles have become increasingly important as they lower mission costs thanks to their re-usability. However, their development is complex, rendering even the basic experimental validation of the required advanced guidance and control (G & C) algorithms prohibitively time-consuming and costly. In this paper, we present the design of an inexpensive small-scale VTOL platform that can be built from off-the-shelf components for less than 1000 USD. The vehicle design mimics the first stage of a reusable launcher, making it a perfect test-bed for G & C algorithms. To control the vehicle during ascent and descent, we propose a real-time optimization-based G & C algorithm. The key features are a real-time minimum fuel and free-final-time optimal guidance combined with an offset-free tracking model predictive position controller. The vehicle hardware design and the G & C algorithm are experimentally validated both indoors and outdoor, showing reliable operation in a fully autonomous fashion with all computations done on-board and in real-time.

cs.RO↗

arXiv2020

Distributed Safe Learning using an Invariance-based Safety Framework

In large-scale networks of uncertain dynamical systems, where communication is limited and there is a strong interaction among subsystems, learning local models and control policies offers great potential for designing high-performance controllers. At the same time, the lack of safety guarantees, here considered in the form of constraint satisfaction, prevents the use of data-driven techniques to safety-critical distributed systems. This paper presents a safety framework that guarantees constraint satisfaction for uncertain distributed systems while learning. The framework considers linear systems with coupling in the dynamics and subject to bounded parametric uncertainty, and makes use of robust invariance to guarantee safety. In particular, a robust non-convex invariant set, given by the union of multiple ellipsoidal invariant sets, and a nonlinear backup control law, given by the combination of multiple stabilizing linear feedbacks, are computed offline. In presence of unsafe inputs, the safety framework applies the backup control law, preventing the system to violate the constraints. As the robust invariant set and the backup stabilizing controller are computed offline, the online operations reduce to simple function evaluations, which enables the use of the proposed framework on systems with limited computational resources. The capabilities of the safety framework are illustrated by three numerical examples.

eess.SY↗