Sustainable and Process-aware Artificial Intelligence – SPAI

Research Activity

The group’s research goes beyond conventional organizational and production processes/workflows, extending the notion of process to various application domains, such as: programs/plans obtained as a composition of services/functions; complex human/social activities; execution of complex tasks by one or more robots/agents (also in the perspective of Agentic AI); physical processes (e.g., concerning floods, pollution, fires); execution of applications in software systems or computer networks; security/privacy attacks and associated prevention/protection policies.

The group’s activities focus on the study and resolution of various research problems, including:

(i) discovery of descriptive and predictive models of process behavior;

(ii) generative AI methods for simulation/prediction of processes (as “digital twins”) and for solving complex tasks (e.g., agentic workflows);

(iii) construction of dynamic models to support decisions and optimization of activities by human operators or agents;

(iv) identification of potentially dangerous behaviors (anomalies, violation of constraints derived from business rules or recommendations/guidelines, security attacks);

(v) monitoring of critical events and quality/performance indicators defined on event logs (and possibly contextual information);

(vi) interpretation of streaming data and event sequences in terms of activities/events at a higher level of abstraction.

The group members boast solid skills and experience in the following disciplinary fields: Data Mining, Machine Learning, Knowledge Representation & Reasoning, Process Modeling, Workflow Management, Mathematical modelling, Optimization.

The complexity, volume and dynamicity of the analyzed data make these objectives challenging and require the development of efficient and robust solutions with respect to the non-stationary and/or uncertain/incomplete nature of the data, and possibly the ability to fully exploit knowledge and feedback from experts and the environment. Particular attention is also paid by the group to the efficiency, transparency and security of the developed solutions, in the emerging perspective of Green/Sustainable AI.

The group’s research embrace frontier topics in the AI and Data Science sectors, including:

– Neuro-Symbolic AI, Informed Machine/Reinforcement Learning;

– Meta-learning, Continual learning;

– Explainable and Human-in-the-loop AI;

– Data/compute/energy-efficient AI;

– Distributed/Federated AI.

Goals

The overall goal pursued by the group can be summarized as follows: study and development of techniques for the analysis, monitoring and prediction of the behavior of complex (potentially distributed and heterogeneous) processes and systems.

The main focus of these techniques consists of data generated by the execution of activities and/or acquired from environmental sensors, and information/knowledge associated with the processes/systems under study.

In this context, the group aims at defining models, tools and methodologies that enable the extraction of information and knowledge from such data, in order to understand, predict and improve the behavior (in terms of efficiency, safety, resilience, quality, etc.) of the processes that generate them and to support decisions concerning their management.

Application Fields

The results produced by the research group will be applicable in important application contexts, including the following: Industry 4.0/5.0, Smart Cities, Civil protection, Power grids, Autonomous systems, Cybersecurity.

Back To Top