Leonardo Lamanna

Leonardo Lamanna graduated in bachelor computer engineering on September 2017 at Università degli Studi di Brescia, in Italy. His bachelor’s thesis was about applying an optimization metaheuristic to solve an NP-hard problem. On September 2019, he graduated cum laude in master computer engineering at Università degli Studi di Brescia. His master’s thesis was about integrating Mathematical Programming and Artificial Intelligence techniques to solve integer linear optimization problems.

On April 2023, he obtained a Ph.D. in Information Engineering with Doctor Europaeus mention at the University of Brescia in collaboration with the Fondazione Bruno Kessler (FBK), in Italy. His Ph.D. topic was the integration of data-driven learning and symbolic planning techniques for agents acting in unknown environments.

On November 2023 he won the “Marco Cadoli” award 2023, which is an annual recognition given by the Italian Association for Artificial Intelligence (AIxIA) for the best doctoral thesis in the field of artificial intelligence at the national level.

Currently he is a postdoctoral researcher at the Fondazione Bruno Kessler.

Major research activity:

My major research activity focuses on the integration of acting, learning and planning. The main objective is to build a system that is capable to learn how to plan and act in a dynamic and complex environment. On the learning side, I’m interested in developing algorithms that allow an artificial agent to learn an abstract model of the dynamics of the environment (e.g. an explicit model like a deterministic finite state machine or a model description in a language to express planning domains). In addition to learning the abstract model, I’m interested in learning probabilistic (generative) models that connects the abstract model with the perceptions of the artificial agents.
On the acting and planning side, the artificial agent decides how to act by means of state-of-art planners (e.g. Fastforward). With its own experience, it enriches the planner knowledge, as well as the learned model of the environment.

 

Minor research activity:

The Multidimensional Multiple Choice Knapsack Problem (MMKP) is a complex combinatorial optimization problem for which finding high quality feasible solutions is a very challenging task. Despite several heuristic approaches have been proposed for its solution, many benchmark instances for the MMKP still remain unclosed to optimality. I developed a new variant of the well-known heuristic framework called Kernel Search and applied it to the MMKP, outperforming state-of-the-art results.

Artificial Intelligence and Machine Learning for autonomous agents that learn to plan and operate in unpredictable dynamic environments

1. Lamanna, Leonardo and Mohamadreza, Faridghasemnia and Gerevini, Alfonso Emilio and Saetti, Alessandro and Alessandro, Saffiotti and Serafini, Luciano and Paolo, Traverso and others,
   Learning to Act for Perceiving in Partially Unknown Environments,
   in «»,
   The 32nd International Joint Conference on Artificial Intelligence,
   ,
   2023
   , pp. 5485-
   5493
2. Lamanna, Leonardo; Serafini, Luciano; Faridghasemnia, Mohamadreza; Saffiotti, Alessandro; Saetti, Alessandro; Gerevini, Alfonso; Traverso, Paolo,
   Planning for Learning Object Properties,
   in «»,
   Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence (AAAI-23),
   ,
   2023
   , pp. -
3. Campari, Tommaso and Lamanna, Leonardo and Traverso, Paolo and Serafini, Luciano and Ballan, Lamberto,
   Online Learning of Reusable Abstract Models for Object Goal Navigation,
   in «»,
   2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),
   ,
   2022
   , pp. 14850-
   14859
4. Lamanna, Leonardo; Serafini, Luciano; Saetti, Alessandro; Emilio Gerevini, Alfonso; Traverso, Paolo,
   Online Grounding of Symbolic Planning Domains in Unknown Environments,
   in «»,
   Proceedings of the 19th International Conference on Principles of Knowledge Representation and Reasoning (KR 2022),
   ,
   2022
   , pp. 511-
   521
5. Lamanna, L.; Serafini, L.; Traverso, P.; Gerevini, A.; Saetti, A.,
   On-line Learning of Planning Domains from Sensor Data in PAL: Scaling up to Large State Spaces,
   in «»,
   Proceeding of the Thirty-Fifth AAAI Conference on Artificial Intelligence (AAAI-21),
   ,
   2021
   , pp. 11862-
   11869
6. Lamanna, Leonardo; Saetti, Alessandro; Serafini, Luciano; Emilio Gerevini, Alfonso; Traverso, Paolo,
   Online Learning of Action Models for PDDL Planning,
   in «»,
   Proceedings of KEPS 2021 – Workshop on Knowledge Engineering for Planning and Scheduling @ ICAPS 2021,
   ,
   2021
   , pp. 4112-
   4118