Daniele Dell'Erba

My research focus on theoretical computer science, precisely the field of formal methods. The majority of the effort aim at solving infinite-duration games on graphs which can be used for formal verification (model checking and synthesis of multi-agent systems) and have applications in automata theory, machine learning, and control theory.

In the last years my the primary task has been to find more efficient solution algorithms for games such as parity, mean-payoff, discounted payoff, and discounted stochastic, developing several techniques like Priority Promotion, Quasi-Dominion measures, and Objective Improvement.

I am also investigating the fundamental connections and differences between the types of infinite-duration games, considering reductions that go back in the chain of standard reductions, both at the game level and at the automaton formulation of the acceptance conditions, as well as connecting the algorithmic advances between the algorithms that solve these games.

Currently, I am applying the algorithms for solving games that model competitive optimization settings to the context of AI, in particular to reinforcement learning where rewards are computed on models like Markov Decision Progress. Solving games more complex than MDPs allow to compute optimal policies for multiple agents and to export advancements on efficient algorithms for specialized models to more general frameworks used in AI.