Robotics and Artificial Intelligent Systems

RUGGED addresses a critical gap in autonomous ground robotics: reliable operation in harsh, unstructured outdoor environments such as agriculture, forestry, and wildfire scenarios. While perception, localization, and navigation have advanced rapidly in controlled settings, little is known about how robotic systems and sensors degrade under dust, smoke, debris, extreme vibrations, and high temperatures. RUGGED tackles this problem head-on by jointly addressing hardware robustness and software resilience.The project follows a strongly experimental, multidisciplinary approach combining electrical and mechanical engineering with robotics and AI. It begins with a systematic assessment of how environmental stressors affect sensors, electronics, and perception pipelines, using both laboratory tests and real-world field trials with forestry machinery. Based on this evidence, RUGGED develops mitigation strategies at two levels: hardware solutions, such as optimized sensor placement, vibration damping, and thermal management; and software solutions, including robust multisensory perception, localization, and mapping.On the software side, the project advances real-time sensor fusion using complementary modalities (LiDAR, RADAR, depth, multispectral and thermal cameras), metric-semantic mapping, traversability analysis, and multimodal localization. These methods are developed and validated using noise-rich virtual environments and real datasets, within a ROS-based framework. The project concludes with large-scale field validation in progressively challenging scenarios and the release of open multisensory datasets. RUGGED ultimately delivers practical guidelines, validated technologies, and data resources to enable dependable semi-autonomous robots in extreme outdoor operations.

More than 2 million new cases of colorectal cancer are diagnosed every year around the globe. The entire global population, living anywhere on earth, above the age of 45 50 can be potentially affected by this challenge. Colorectal cancer remained the second leading cause of cancer-related deaths, accounting for 12 to 14% of all cancers recorded in Europe in 2012, and contributes U$14 billion to annual healthcare costs in the United States alone. In Asia, incidence rates range from 49.3 in Japan, 24.7 in South Korea, and 35.1 in Singapore, and they are equally high in many African and South American countries, turning this into a global challenge. Current scientific breakthroughs enable the development of non-invasive, high-quality imaging, energy-efficient, and miniaturized electronic devices that can travel inside the gastrointestinal tract using natural body cavities. Applying this technology to the screening of large groups, once they reach the age of 45 50, can significantly lower the number of new cases in an advanced stage of progression that are diagnosed every year. 

Conventional computing, implementing for instance Convolutional Neural Networks, transfer learning, and ensemble learning, struggle with the computational load of analysing billions of hours of track images used. By using quantum computing, it could be possible to solve some complex optimisation and pattern recognition tasks more efficiently than classical computers. Quantum algorithms such as Quantum Neural Networks and Quantum Support Vector Machines could enhance the efficiency and accuracy of image analysis, and quantum features space could provide more nuanced insights from the data, potentially improving detection rates.

The NewGenVIP project aims to develop and optimize, from an ecodesign perspective, a new range of industrial mixers for Vacuum Infusion Processes (VIP), integrating real-time monitoring and decision support through the application of artificial intelligence powered by historical data and new instrumentation developed within the project. Among the planned innovations are the ability to operate with multiple infusion channels simultaneously, an active degassing system, an IoT interface, and tools that enable comprehensive tracking of operating conditions. The goal is to reduce the occurrence of failures, improve efficiency, safety, and sustainability in industrial processes within the composites sector, increase the quality of final products, and lower operational costs.

The WOOSU project aims to develop a solution that promotes the optimization of the operation, sustainability and circularity of wastewater treatment plants (WWTP). Its results lead to a reduction in operational costs, the sector’s ecological footprint with the reduction of greenhouse gas emissions and local water reuse methodologies. The project will also value other surpluses from the operation, sludge, biogas and heat, reintegrating them into the production chain and closing the resource use cycle, through new models of multi-aspect industrial symbiosis. Process optimization leads to improved effluent quality, taking into account energy consumption and variations in equipment operating times. The project will innovate by incorporating Life Cycle Analysis into the decision-making process for treatment actions, developing complementary tools to support the ISO 59020:2024 standard, aligning with EU best practices and future industry procedures. The inclusion of sustainability and circularity components in ETEs allows for management that reduces resource waste and maximizes the reuse of treated water, nutrients included in biofertilizers and biogas. This results in a reduction in greenhouse gas emissions, operational costs, and the adoption of concrete circular economy practices. These advances reduce the immediate local environmental impact of treatment operations, promote more efficient use of natural resources, ensure the economic sustainability of WWTPs and contribute to public health and environmental preservation. The project strives for scientific advances and development of technologies in various areas, namely process monitoring, creation of biochemical models and symbiotic models, sustainability assessment, development of advanced control systems and industrial interface tools with the operator. All technologies will be tested in a real environment, using facilities contracted with one of the co-promoters. A great effort will be made to promote the dissemination of the project and its results among the population, entities managing similar equipment, government decision-making bodies and environmental systems, thus also reinforcing the social sustainability component.

The PharmaRobot project aims to improve hospital efficiency through an autonomous robot that distributes urgent medications quickly and accurately, reducing waiting times and minimizing human error. The proposed solution will enable more effective resource management, freeing up professionals for critical tasks and ensuring greater safety and quality in healthcare.

The operation of Biomass Valorization Plants (CVB) is complex due to the great variability of factors related to biomass, such as humidity, types and granulometry, as well as due to the condition of the equipment, quantity of pollutants produced, etc. The uncertainty regarding operational factors implies a high need for human intervention in balancing the various process parameters. This model of constant human supervision of industrial processes, transversal to different production sectors, places the operator with a preponderant role in the performance and efficiency of the processes, with invariably large associated inefficiencies. Taking advantage of its recognition in the industrial process control solutions market, OnControl, in cooperation with Universidade de Coimbra, thus advances this project with the aim of introducing a transformation of the usual scenario of control and management of biomass burning processes, through the development of an advanced control solution to optimize the operation of biomass boilers, applicable to the CVB market or other industries with biomass boilers. The ultimate objective of this technology is to increase the amount of energy per volume of biomass consumed and reduce the maintenance cost of process equipment. From a holistic approach, advanced control technology will be developed and adapted, through advanced computing and artificial intelligence methods that allow the digitalization of the production process, in a digital transformation logic, which allows achieving performance gains and new formats assistance/decision support for operators. Performance gains will be achieved by automatically guiding the process using computing algorithms and supporting the decision to execute tasks by operators and redirecting them to activities with greater added value, contributing to the green transition by promoting more efficient processes and sustainable.

Autism spectrum disorders (ASD) affect the way a person perceives emotions and socialises with others in a flexible way, often leading to difficulties with interaction and communication. As a result, most scientific efforts focus on interventions aimed at improving these two aspects. Drugs have been developed which, although somewhat successful in helping to control some of these manifestations, do not address the core symptoms. Behavioural therapies have also been introduced, which have yet to prove fully effective. They are very expensive and not suitable for intensive rehabilitation. In the present project, they will conduct a phase IIb clinical trial using a brain-computer interface that incorporates educational games. The ultimate goal is to develop an inexpensive, portable, easy-to-use device that can be used anywhere, enabling people with ASD to improve their social and emotional cognition.

The ForestSphere project aims to develop a Digital Twin (DT) for forest management, recognizing their socio-economic and ecological importance.Forests contribute significantly to the EU's GDP and serve as critical carbon sinks. However, wildfires pose a substantial threat to forests, leading to environmental damage, loss of life, and economic losses.

Despite advancements in DT technology in various sectors, current literature has limited exploration of its application in forest management, particularly regarding wildfire prediction.

The ForestSphere project intends to bridge this research gap by creating a Forest DT that combines multimodal and multi-scale data from various sensors to enhance forest health monitoring and wildfire risk assessment. The project's key objectives include:

• Developing an open, scalable, and modular Digital Twin architecture.
• Enabling precise, high-resolution forest mapping using advanced sensing techniques.
• Implementing data processing for dynamic updates of the DT.
• Utilizing AI for predictive wildfire analytics.
• Designing user-centric tools for decision support to assist firefighting teams and policymakers.
• Conducting pilot cases to demonstrate the Forest DT's applications in forest and wildfire management.Challenges include managing high computation costs, ensuring real-time data synchronization, and effectively integrating diverse data sources.

The project follows a structured three-phase methodology: Build, Optimize and Deploy, and the project promises significant socio-economic benefits by providing real-time data to authorities, aiding infrastructure management, and supporting civil protection efforts during wildfires.

Bridging classical computing and quantum computing towards high-performance computing (HPC) systems involves harnessing the strengths of both approaches to tackle complex problems efficiently. Classical computing excels at executing sequential algorithms exploiting data locality and handling vast amounts of data up to some level of parallelism, while quantum computing offers the potential to solve certain problems exponentially faster through quantum parallelism and entanglement. Q-Bet will try to find a methodology to bridge the know-how acquired with 80 years from classical computing to make the leap to quantum computing, exploiting the strengths of each and overcoming the limitations of the other. This hybrid exploration of both technologies will be performed first running perfect qubits with HPC at quantum simulators in some of the fastest machines available in our datacenters at the Universidade de Coimbra, and later on real noisy qubits from IBM (partner of the project) quantum machines. The middleware developed will foster the experimentation of new use cases in the quantum domain, namely in the physics arena by exploiting new potential agents in the form of particles sensing for solar energy conversion cells. We will also advance security problems for networks and communications by developing powerful error- correcting codes for quantum technology. Finally, we will explore bio-inspired computing and AI for performing Bayesian inference and running optimization algorithms in several domains. By integrating classical and quantum computing techniques, HPC systems can leverage the power of quantum algorithms for specific tasks while maintaining the robustness and scalability of classical systems. The exploratory nature of the project will allow integrating on a competitive environment a core team experienced in teaching and researching quantum computing. Adding new post-grads involving more departments and targeting emerging applications will be possible thanks to the active collaboration of IBM Portugal. Q-Bet outlines the exciting potential for hybrid classical-quantum computing architectures in realising the next generation of HPC. Moreover, collaborating across borders with external partners of Q-Bet will allow sharing the costs and resources associated with research and development. This can help mitigate the financial and logistical constraints that small countries or individual institutions may face, enabling a more ambitious and impactful Q-Bet project. Finally, we expect to integrate in this preliminary research project specific recommendations regarding the integration of Quantum Technologies into the work programs of the Chips-Act for semiconductors, and EuroHPC for High- Performance Computing (HPC).

Project led at CINEICC (Universidade de Coimbra) that aims to develop and pilot-test a self management, 8-week Virtual Reality (VR) program delivering Acceptance and Commitment Therapy (ACT) to support pain management and mental health in adults with chronic post-stroke pain. Using a parallel-group randomized controlled feasibility trial, the study compares (i) the VR-ACT intervention, consisting of eight 3D immersive skills-training modules (e.g., attentional and emotional regulation through mindful meditation practices, and fostering valued behavior change), with (ii) a sham VR control comprising eight 2D non-immersive video animations and distraction exercises delivered via a VR headset. The project adopts a mixed-methods evaluation (quantitative and qualitative, with qualitative content analysis supported by NVivo) and includes psychosocial outcomes as well as fMRI-based assessment of triple-network connectivity using Network Based Statistic methods. ISR-UC participates in the project, providing technical support for the development and integration of the VR delivery components/platform.

In 2024, the Embalagem do Futuro project advanced through the coordinated work of three groups: Codification, Sensoring, and Futuring. Codification developed hybrid graphical codes combining public and encrypted private data, enabling offline access by authorized users. Sensoring created a miniaturized, autonomous sensing device with environmental monitoring and wireless communication. Futuring focused on scalable fabrication of stretchable electronics and battery-free sensor systems powered by inductive coupling. Together, these developments pave the way for smarter, more sustainable, and fully integrated packaging solutions.Codification Over the past year, the group focused on adapting advanced graphical codes to the constraints imposed by smart packaging. It was established that the packaging must contain codes compatible with conventional readers, while also enabling authorized users to retrieve the necessary information to unlock encrypted sensor data. The developed solution integrates both public and private data into a single code. During the shipping phase (first mile), a label is applied to the packaging containing two levels of information, with the private part comprising access keys to the sensors embedded in the packaging, as illustrated in Figure 1. Upon reception (last mile), if internet access for authentication is unavailable, the user can rely on a dedicated application capable of reading the private part of the code and locally retrieving the necessary credentials, as shown in Figure 2.One of the key innovations introduced is the ability to access encrypted information entirely offline, restricted only to authorized users — a new paradigm in physical environments with limited connectivity. The solution maintains compatibility with conventional readers for accessing public information, while fully concealing the private data. This hybrid approach allows for a smooth transition to industry adoption and minimizes the need for changes in existing processes.

The project positions itself at the forefront of industrial automation and human-machine coexistence, focusing on the growing influence of robotics over experience and performance. It envisions cognition-adaptation synergy as the natural step leading to both improved productivity and worker well-being.