Institution profile

University of Southampton

Academic institutioneurope · gb
Official website
Research library446linked papers
Opportunities0open roles
Selected work

Representative Papers

A Kullback–Leibler divergence method for input–system–state identification

Jul 01, 2023Journal of Sound and Vibration

In Kalman filtering, high estimation uncertainty and difficulty in jointly identifying inputs, system dynamics, and hidden states arise from inaccurate initial parameter guesses and model–structure mismatches. To address this, we propose a unified variational identification framework grounded in the Kullback–Leibler (KL) divergence. Our method integrates probabilistic graphical models with variational inference, jointly modeling input signals, unknown system dynamics, and latent states as an information-minimization optimization problem—thereby reducing reliance on strong prior structural assumptions. Crucially, KL divergence serves as a unifying objective that simultaneously drives the co-estimation of inputs, dynamics, and states, enhancing both accuracy and robustness under model mismatch and noise corruption—particularly in nonlinear systems. Experimental results on synthetic and real-world systems demonstrate significant reductions in modeling error and estimation uncertainty compared to conventional approaches.

9 citationsRead paper

Rydberg Atomic Quantum Receivers for Classical Wireless Communications and Sensing: Their Models and Performance

Dec 07, 2024arXiv.org

The absence of a rigorous system-level model and quantitative performance characterization for Rydberg atom quantum receivers (RAQRs) impedes their integration into classical wireless communication and sensing systems. Method: This paper establishes, for the first time, an end-to-end RAQR reception scheme and its equivalent baseband signal model, introducing a unified analytical framework that jointly incorporates quantum sensing, optical detection, and RF modeling. Contribution/Results: We derive fundamental theoretical gain bounds for RAQRs relative to classical RF receivers, demonstrating ≥27 dB and ≥40 dB improvements in received signal-to-noise ratio (SNR) under photon shot-noise-limited and standard quantum-limited conditions, respectively. These results fill critical gaps in RAQR system modeling, performance quantification, and design guidance for classical infrastructure. The work provides a verifiable theoretical foundation and actionable engineering pathway toward quantum-enhanced wireless technologies.

4 citationsRead paper

Model-Based Reward Shaping for Adversarial Inverse Reinforcement Learning in Stochastic Environments

Oct 04, 2024arXiv.org

To address the theoretical failure and performance degradation of Adversarial Inverse Reinforcement Learning (AIRL) in stochastic environments, this paper proposes a model-augmented AIRL framework that explicitly incorporates a dynamics model into the reward shaping process—yielding, for the first time, a model-driven reward design for AIRL in stochastic settings with rigorous theoretical guarantees. Our core contributions are: (1) a novel theoretical analysis framework establishing bounds on both reward estimation error and policy performance gap; and (2) an algorithm that jointly optimizes transition model estimation and adversarial training. Experiments on MuJoCo demonstrate that our method significantly outperforms existing baselines in stochastic environments, maintains competitive performance in deterministic ones, and achieves substantially improved sample efficiency.

2 citationsRead paper

A Generalised Framework for Property-Driven Machine Learning

May 01, 2025

Neural networks often fail to satisfy formal properties required in safety-critical applications. Method: This paper proposes an attribute-driven unified training framework that jointly integrates geometric constraints from adversarial training (generalized hyperrectangular input domains) and semantic constraints encoded via differentiable first-order logic, thereby translating arbitrary formal properties into differentiable loss terms. The framework jointly optimizes a property-weighted loss function and neural network controllers. Contribution/Results: It enables concurrent assurance of robustness and correctness under flexible, domain-specific regional specifications across diverse fields (e.g., control systems, NLP). Evaluated on a neural controller for unmanned aerial vehicles, the framework achieves significant improvement in formal property satisfaction rates. Open-sourced and fully reproducible, it supports a broad range of canonical formal properties, demonstrating strong generalizability and plug-and-play usability.

1 citations1 influentialRead paper

Constraint Learning for Non-confluent Proof Search

Mar 05, 2026International Conference on Theorem Proving with Analytic Tableaux and Related Methods

This work addresses the inefficiency of non-confluent tableau calculi—such as connection tableaux—in proof search, where frequent backtracking severely hampers performance, and naive backtracking restrictions compromise completeness. For the first time, the authors integrate a constraint learning mechanism into the classical first-order connection calculus, substantially reducing backtracking while preserving theoretical completeness. The core innovation lies in a novel, iteratively refinable constraint language that dynamically prunes and optimizes the search space. Experimental results demonstrate a significant improvement in practical proof efficiency. Moreover, the proposed framework offers a general and transferable approach to backtracking control, applicable to other non-confluent tableau systems.

1 citationsRead paper
Recent publications

Latest Papers

Utility-Level-Dependent Ambiguity

Sep 10, 2026

本文探讨了决策情境下模糊性敏感选择的变化,通过引入依赖于确定性等价值水平的模糊权重方法来解决这一问题。

0 citationsRead paper