Optimisation and Sequential Decision-Making

My work in optimisation studies how to compute decisions and estimates under constraints and uncertainty. It includes convex relaxations for localisation, nonconvex optimisation for facial animation, and learning from pairwise comparisons. Statistical estimation and system identification supply models for control.

In sequential decision-making, I study algorithms that act on observations revealed over time. This includes constrained online learning and reinforcement learning for orbital systems, task offloading and distributed energy markets. Related work uses predictive models to support operational decisions in transport and industrial production.

CLASP establishes bounds on regret and cumulative squared violation under its stated assumptions. ASTRA combines an ADMM-based solver with online learning for satellite-network decisions. OrbitZoo and PeersimGym provide environments for evaluating reinforcement learning.

Agentic decision-making

My research on agentic decision-making asks when algorithmic guarantees survive substitutions selected by an AI agent during operation. A decision architecture comprises the operator, model class, planning horizon, planning objective and subgoal structure. A substitution replaces one component of that decision architecture.

This line of work examines how computational approximations and changes to the model class affect regret and cumulative squared violation, and under what conditions bounds for individual substitutions extend to a sequence. The physical loss and constraints remain fixed during evaluation.

Funded projects

Projects active during 2021 or later, including completed projects.

• TaRDIS (2023–2026). Task leader. Decentralised learning and task orchestration for edge systems, with use cases including satellite swarms and distributed marketplaces. Reference 101093006.
• AI4Europe (2022–2025). Team member. Shared AI research infrastructure for experimentation and reproducibility through the AI-on-demand platform. Reference 101070000.
• Sustainable Stone by Portugal (2022–2025). Team member. Natural-stone valorisation and digital production, with related research on a data-centric learning pipeline under weak supervision. Reference C644943391-00000051/40.
• Artificial Intelligence Fights Space Debris (2022–2026). Local PI. Collision-risk prediction and machine learning for space traffic management, in collaboration with Neuraspace. Reference C626449889-0046305.
• BIGMATH (2018–2022). National principal investigator. Industrial doctoral research in optimisation, statistics and linear algebra, including facial data analysis. The University of Milan coordinated the consortium. Reference 812912.

Publications since 2021

• Satellite Safe Margin: Fast Solutions for Conjunction Analysis. Journal of Guidance, Control, and Dynamics, 2026.
• ASTRA: ADMM-Accelerated Topology Reconfiguration for Dynamic Satellite Constellations. NeurIPS 2026.
• CLASP: Online learning algorithms for Convex Losses And Squared Penalties. Proceedings of the 43rd International Conference on Machine Learning, 2026.
• A Multi-Token Coordinate Descent Method for Semi-Decentralized Vertical Federated Learning. IEEE Transactions on Signal Processing, 2026.
• AdamX: Cosine similarity meets gradient descent. Accepted at EPIA 2026.
• Multi-Agent Reinforcement Learning for Procurement Auctions Based Distributed Energy Market Governance. Accepted at EPIA 2026.
• Towards a data-centric KDD pipeline for natural-stone valorization under weak supervision. Accepted at EPIA 2026.
• Error- and Prediction-Driven Motor Learning in the Cortico-Cerebellar Loop. Accepted at EPIA 2026.
• Generalizing Trilateration: Approximate Maximum Likelihood Estimator for Initial Orbit Determination in Low Earth Orbit. IEEE Transactions on Aerospace and Electronic Systems, 2025.
• Ranking with Confidence for Large Scale Comparison Data. Proceedings of the 2025 SIAM International Conference on Data Mining (SDM).
• OrbitZoo: Real Orbital Systems Challenges for Reinforcement Learning. NeurIPS 2025 poster.
• Communication-Efficient Vertical Federated Learning via Compressed Error Feedback. IEEE Transactions on Signal Processing, 2025.
• Refined Inverse Rigging: A Balanced Approach to High-fidelity Blendshape Animation. SIGGRAPH Asia 2024 Conference Papers.
• Maximum likelihood localization of a network of moving agents from ranges, bearings and velocity measurements. Signal Processing, 2024.
• One-Shot Initial Orbit Determination in Low-Earth Orbit. Accepted at the IEEE Aerospace Conference, 2024.
• PeersimGym: An Environment for Solving the Task Offloading Problem with Reinforcement Learning. Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track, 2024.
• Distributed Solution of the Blendshape Rig Inversion Problem. SIGGRAPH Asia 2023 Technical Communications.
• Probabilistic Registration for Gaussian Process Three-Dimensional Shape Modelling in the Presence of Extensive Missing Data. SIAM Journal on Mathematics of Data Science, 2023.
• A majorization-minimization-based method for nonconvex inverse rig problems in facial animation: algorithm derivation. Optimization Letters, 2023.
• Decision Support Models for Predicting and Explaining Airport Passenger Connectivity From Data. IEEE Transactions on Intelligent Transportation Systems, 2022.
• Robust Localization With Bounded Noise: Creating a Superset of the Possible Target Positions via Linear-Fractional Representations. IEEE Transactions on Signal Processing, 2022.
• A Multi-Token Coordinate Descent Method for Vertical Federated Learning. NeurIPS 2022 Workshop on Federated Learning.
• STRONG: Synchronous and asynchronous robust network localization, under non-Gaussian noise. Signal Processing, 2021.
• Clustering of the Blendshape Facial Model. 2021 29th European Signal Processing Conference (EUSIPCO).
• ISEE.U: Distributed online estimation and control for improved target localization accuracy. 2021.

Preprints and related manuscripts since 2021

• FAuNO: Semi-Asynchronous Federated Reinforcement Learning Framework for Task Offloading in Edge Systems. arXiv preprint, 2025.
• A Proof of Strong Consistency of Maximum Likelihood Estimator for Independent Non-Identically Distributed Data. arXiv preprint, 2025.
• Finding Real-World Orbital Motion Laws from Data. arXiv preprint, 2023.
• Probability of Collision of satellites and space debris for short-term encounters: Rederivation and fast-to-compute upper and lower bounds. arXiv preprint, 2023.
• High-fidelity Interpretable Inverse Rig: An Accurate and Sparse Solution Optimizing the Quartic Blendshape Model. arXiv preprint, 2023.
• Accurate and Interpretable Solution of the Inverse Rig for Realistic Blendshape Models with Quadratic Corrective Terms. arXiv preprint, 2023.
• ISEE.U: Distributed online active target localization with unpredictable targets. arXiv preprint, 2022.
• Decentralized EM to Learn Gaussian Mixtures from Datasets Distributed by Features. arXiv preprint, 2022.
• A Cluster-Based Trip Prediction Graph Neural Network Model for Bike Sharing Systems. arXiv preprint, 2022.
• COSMIC: fast closed-form identification from large-scale data for LTV systems. arXiv preprint, 2021.
• Accurate, Interpretable, and Fast Animation: An Iterative, Sparse, and Nonconvex Approach. arXiv preprint, 2021.