Artificial Intelligence

Artificial Intelligence (AI) research aims at developing algorithms, models, and technologies that enable machines to operate autonomously, adapting to new situations, and improving over time through experience. AI focuses on creating systems capable of performing tasks that typically require human intelligence, such as problem-solving, learning, reasoning, perception, understanding visual scenes and natural language, and interacting with the environment. Our area spans the following interdisciplinary subfields of AI: Automated reasoning and intelligent systems include knowledge representation, theorem proving, model building, satisfiability modulo theories, temporal reasoning, intelligent agents, multi-agent systems, and search methodologies with emphasis on discrete space search. Machine learning methods and approaches encompass supervised learning, unsupervised learning, reinforcement learning, and deep learning, along with advanced methodologies such as continual learning, self-supervised learning, data generation, domain adaptation, and generalisation. Natural language processing primarily leverages statistical, linguistic, and deep learning techniques for text understanding, language generation, sentiment analysis, and machine translation. Computer vision enables machines to interpret and understand visual information through tasks such as image classification, object detection, image segmentation, 3D reconstruction, biometric and human behaviour recognition, motion analysis, image and video understanding and generation, and scene understanding, which are essential for applications including autonomous vehicles, medical imaging, augmented reality, and surveillance systems. Combining visual and language processing modalities, computer vision and natural language processing present solutions for several multimodal applications, such as cross-modal retrieval, visual question answering, enhanced human-computer interactions, and multimodal content creation. Multimodal systems are not limited to vision and language, but other modalities can be involved for the sake of (scene) understanding or (data) generation. Computer graphics applies geometric and physically-based theories to various tasks, including object modelling, shape analysis, visualisation, texture mapping, rendering, animation, and simulation. It encompasses techniques for creating realistic images, virtual environments, and interactive graphics in applications such as video games, film production, virtual reality, and scientific visualisation. In medical bioinformatics, the studies involve the development of advanced network-based AI algorithms and deep learning algorithms to analyse heterogeneous personal data such as genomes and integrate them, to extract the important features to stratify patients based on disease subtypes, environmental interactions, and personalising treatment responses. In multi-dimensional signal processing, the studied techniques regard advanced filtering, feature extraction and segmentation methods, multiresolution and sparse signal representations, time(space)/scale methods, including wavelets, compressive sensing, large-scale image characterisation and retrieval tools, related to both optical and multimodal images. Social and affective computing involves creating systems that recognise and respond to human emotions and understand social interactions and dynamics. This field leverages data from various sources, such as facial expressions, voice tone, body language, and physiological signals, to interpret users' emotional states and social contexts. The aim is to enhance human-computer interactions by enabling systems to adapt to both individual emotional and psychological conditions and broader social environments. By integrating these insights, social and affective computing seeks to create more intuitive, empathetic, and context-aware technology, improving user experiences in areas such as virtual assistants, social media, customer service, and collaborative platforms. Autonomous agents and multiagent systems: Design and development of autonomous agents (e.g., robotic platforms) that can sense, model, and interact with the environment in which they operate and with other agents (including humans) to carry out complex tasks (e.g., deliver items in a logistic scenario). Main research topics include task assignment, distributed constraint optimization, decentralised optimization, Multi-Agent reinforcement learning, cooperative perception. Reinforcement learning and planning under uncertainty: Design and development of algorithms for learning policies in unknown environments. Specific topics concern safe reinforcement learning (e.g., learning policies in safety-critical environments such as autonomous driving), deep reinforcement learning, model-based reinforcement learning, offline reinforcement learning (e.g., learning policies from pre-acquired data in mobile robotics, Industry 5.0 and environmental sustainability applications), bridging learning and planning, planning under uncertainty in partially observable environments, and reasoning in the face of uncertainty. Explainable and neuro-symbolic planning: Design and development of algorithms that merge symbolic methods - i.e., high-level symbolic (human-readable) representations of problems, logic, and search - with neural network-based techniques to address the weaknesses of each approach, providing a robust and explainable AI capable of reasoning, learning, and cognitive modeling (e.g., learning interpretable policies and allow agents to explain their decisions in autonomous driving and industrial plant control).