Selected research themes

Decision policies under distribution shifts

Evaluating and learning new decision policies from past data – whether observational or experimental – gives rise to several research challenges.

  1. “Learning Pareto-Efficient Decisions with Confidence”.
    S. Ek, D. Zachariah, and P. Stoica.
    International Conference on Artificial Intelligence and Statistics, 2022, pp. 9969–9981
  2. “Off-Policy Evaluation with Out-of-Sample Guarantees”.
    S. Ek, D. Zachariah, F. D. Johansson, and P. Stoica.
    Transactions on Machine Learning Research, 2023
  3. “Externally Valid Policy Evaluation from Randomized Trials Using Additional Observational Data”.
    S. Ek and D. Zachariah.
    The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

Data-driven control

In this theme, the aim is to analyze and develop new methods for constructing controllers directly from data. We consider both indirect methods, where a model for the system to be controlled is first estimated, and direct methods.

  1. “On the regularization in DeePC”.
    P. Mattsson and T. B. Schön.
    IFAC-PapersOnLine, vol. 56, no. 2, pp. 625–631, 2023
  2. “On the equivalence of direct and indirect data-driven predictive control approaches”.
    P. Mattsson, F. Bonassi, V. Breschi, and T. B. Schön.
    IEEE Control Systems Letters, 2024
  3. “Entropy-regularized diffusion policy with q-ensembles for offline reinforcement learning”.
    R. Zhang, Z. Luo, J. Sjölund, T. B. Schön, and P. Mattsson.
    Advances in Neural Information Processing Systems (NeurIPS), 2024
  4. “Safe Output Feedback Improvement with Baselines”.
    R. Zhang, P. Mattsson, and D. Zachariah.
    63rd IEEE Conference on Decision and Control (CDC), 2024
  5. “Learning state observers for recurrent neural network models”.
    F. Bonassi, C. Andersson, P. Mattsson, and T. B. Schön.
    63rd IEEE Conference on Decision and Control (CDC), 2024