Luís F. S. Marques

Curriculum Vitae Google Scholar GitHub Twitter

I’m advised by Dmitry Berenson! I am interested in the algorithmic foundations of decision-making under uncertainty, with a focus on settings where: (i) available models are not sufficiently accurate nor calibrated, (ii) data is sparse or corrupted, and (iii) few assumptions can be made about the data-generating process. My goal is to construct systems capable of resilient, provably safe operation in low-structure environments.

Some time ago in a rainy place far, far away, I obtained an M.Eng. in Aeronautical Engineering @ Imperial College London. There, I collaborated with Panagiotis Angeloudis on safety for learned autonomous vehicles policies, and with Yiannis Demiris on modeling multi-material food manipulation interactions for assistive feeding.

Open to collaborations! Get in touch: lmarques [at] umich.edu

News

Jun 2026	Presenting “Lies We Can Trust” as an Oral Spotlight @ Geometry in the Age of Data‑Driven Robotics Workshop, ICRA
Apr 2026	“Local Conformal Calibration of Dynamics Uncertainty from Semantic Images” has been accepted to WAFR 2026!
Jan 2026	“Lies We Can Trust: Quantifying Action Uncertainty with Inaccurate Stochastic Dynamics through Conformalized Nonholonomic Lie groups” has been accepted to RA-L!
Nov 2025	Grateful to have been recognized with an Outstanding Reviewer Award at ICMI 2025.
Feb 2025	Happy to receive a Rackham Graduate Student Research Grant (university-wide) to help support hardware experiments.

Selected Publications

Lies We Can Trust: Quantifying Action Uncertainty with Inaccurate Stochastic Dynamics through Conformalized Nonholonomic Lie groups

L. Marques, M. Ghaffari, and D. Berenson.

In IEEE Robotics and Automation Letters (RA-L), 2026.

Abstract Paper

Key Takeaways: A) post-hoc calibration layer on top of Lie-algebraic Gaussian uncertainty estimators → makes approximate InEKF covariances provably calibrated. B) Extended conformal guarantees from Euclidean configurations to SE(2) robots & improved volume-efficiency.

We propose Conformal Lie-group Action Prediction Sets (CLAPS), a symmetry-aware conformal prediction-based algorithm that constructs, for a given action, a set guaranteed to contain the resulting system configuration at a user-defined probability. Our assurance holds under both aleatoric and epistemic uncertainty, non-asymptotically, and does not require strong assumptions about the true system dynamics, the uncertainty sources, or the quality of the approximate dynamics model. Typically, uncertainty quantification is tackled by making strong assumptions about the error distribution or magnitude, or by relying on uncalibrated uncertainty estimates — i.e., with no link to frequentist probabilities — which are insufficient for safe control. Recently, conformal prediction has emerged as a statistical framework capable of providing distribution-free probabilistic guarantees on test-time prediction accuracy. While current conformal methods treat robots as Euclidean points, many systems have non-Euclidean configurations, e.g., some mobile robots have SE(2). In this work, we rigorously analyze configuration errors using Lie groups, extending previous Euclidean space theoretical guarantees to SE(2). Our experiments on a simulated Jetbot, and on a real MBot, suggest that by considering the configuration space’s structure, our symmetry-informed nonconformity score leads to more volume-efficient prediction regions that represent the underlying uncertainty better than existing approaches.
Quantifying Aleatoric and Epistemic Dynamics Uncertainty via Local Conformal Calibration

L. Marques, and D. Berenson.

In 16th International Workshop on the Algorithmic Foundations of Robotics (WAFR), 2024.

Abstract Paper

Key Takeaways: A) Transformed approximate linear-Gaussian uncertainty estimates into provably calibrated → can build probabilistic safe plans when OOD. B) Moved from global to state-action-dependent calibration → motion plans are steerable towards regions of low dynamics uncertainty.

Whether learned, simulated, or analytical, approximations of a robot’s dynamics can be inaccurate when encountering novel environments. Many approaches have been proposed to quantify the aleatoric uncertainty of such methods, i.e. uncertainty resulting from stochasticity, however these estimates alone are not enough to properly estimate the uncertainty of a model in a novel environment, where the actual dynamics can change. Such changes can induce epistemic uncertainty, i.e. uncertainty due to a lack of information/data. Accounting for both epistemic and aleatoric dynamics uncertainty in a theoretically-grounded way remains an open problem. We introduce Local Uncertainty Conformal Calibration (LUCCa), a conformal prediction-based approach that calibrates the aleatoric uncertainty estimates provided by dynamics models to generate probabilistically-valid prediction regions of the system’s state. We account for both epistemic and aleatoric uncertainty non-asymptotically, without strong assumptions about the form of the true dynamics or how it changes. The calibration is performed locally in the state-action space, leading to uncertainty estimates that are useful for planning. We validate our method by constructing probabilistically-safe plans for a double-integrator under significant changes in dynamics.