Outstanding results

In the research area of Advanced Electronic Materials and Sensors, in 2022, Professor Tavakoli was awarded the Prestigious European Research Council Consolidator Grant with the value of 2.8 million Euros. Liquid3D proposes bioinspired electronics and machines that are soft, resilient, self-healing, shape-morphing, and fully recyclable. The project has developed (and is continuing to develop) a series of game changer Liquid Metal based composites that can be 3D printed to constitute functional cells of such soft machines. This includes printed batteries, actuators, and sensors that can be printed side by side. This provides an excellent design freedom to scientists for manufacturing complex living electronics, while guaranteeing that any possible product coming from these inventions will be Resilient, Repairable, and Recyclable. Liquid3D foresees to develop fundamental understanding, and mathematical modelling of biphasic systems, and develops novel room temperature printable composites with sensing/acting/energy storage properties, and methods for recycling them. It is as well investigating novel forms of implementing truly 3D electronics, with distributed functional cells. Liquid3D intends to fundamentally rethink the concept of electronics, as we know today. From rigid and brittle to soft, resilient and repairable; from polluting to recyclable; from battery dependent to self-powered; from 2D to truly 3D. It proposes a radically new way of making greener electronics. Liquid3D aims to establish a world leading center on recyclable and green electronics.

Brain-computer interfaces (BCI) face a significant usability challenge. Their low reliability, combined with the high mental workload required for their control, necessitates the exploration of new approaches for their effective use. The Human-centered Mobile Robotics (HcMR) team has developed innovative methods combined into a single framework to steer a robotic wheelchair with an electroencephalography P300-based BCI. The system integrates self-paced control, allowing users to send BCI commands at their own pace, with a collaborative controller that combines sparse BCI commands (user intentions) with navigation information, generating safe and smooth trajectories in complex environments. Additionally, the system incorporates a dynamic time window to detect BCI commands, allowing it to adapt to user attention shifts and fatigue. Altogether, this framework has enabled wheelchair control with nearly 100% accuracy and low mental workload, achieving unprecedented results. Experimental validation included able-bodied participants and participants with severe motor disabilities (cerebral palsy, spinal cord injury, agenesis of the four members, limb-girdle muscular dystrophy, and Duchene muscular dystrophy). This achievement represents a milestone in BCI development, and a significant advancement in its use as Assistive Technology. One of the latest publications related to this robotic wheelchair controlled with a BCI received the prestigious 2022 Andrew P. Sage Best Transactions Paper Award.

Arthroscopy is a modality of orthopedic surgery in which instruments and endoscopic camera (the arthroscope) are inserted into the articular cavity through small incisions (the surgical ports). Arthroscopy is highly beneficial for the patient but clinical execution is difficult to accomplish because of indirect visualization and limited maneuverability inside the joint. This is a scenario where surgical assistive technologies can have strong impact in improving clinical outcome and disseminating the benefits of arthroscopy by increasing the number of adopters. The ISR-UC created the first effective system for accomplishing navigated arthroscopy that combines real-time video processing for accurate 3D measurements on the anatomy, with augmented reality for overlaying meaningful guidance information in images. It is the first of the kind not requiring additional intra-operative sensing modalities that preclude the application in arthroscopy. Moreover, the improved usability, higher metric accuracy, and avoidance of additional capital equipment make video-based navigation also appealing for open orthopedic surgery. The research effort conducted to several high profile publications and patents that were licensed in exclusive to a spin-off company backed by Venture Capital called Perceive3D (P3D). In early 2021 P3D was discretely acquired by an incumbent in orthopedics with almost 20000 workers worldwide. In the follow-up of the acquisition this incumbent established an advanced research unit in the city with a collaboration protocol with the ISR-UC. The research topics being tackled include medical image segmentation, visual 3D registration of anatomic tissues, 6D pose estimation of instruments, and markerless surgical navigation.

ID documents such as passports and national ID cards are used as physical (now increasingly digital) portable documents with a face portrait that assures the citizen is the genuine owner. Additionally, Facial Recognition Systems can now help institutions and companies authenticate citizens, including by scanning ID documents, for instance in airports. However, the portrait is one of the most attacked security elements, which creates the need for new scientific and technical solutions to protect authentication systems against attacks. The Computer Vision Lab of ISR-UC has worked since 2019, in partnership with the Portuguese Mint and Official Printing Office, in projects to develop facial recognition systems resilient to attacks. One of the biggest threats to passports is morphing, where the printed face is the result of blending two or more similar persons, allowing both to cross a border with the same document. Detection and blocking of morphing attacks was identified by the European Commission as a priority for the EU Entry/Exit System (EES). In this context, we developed several algorithms to detect morphing in face images and submitted them to the NIST MORPH benchmark. This benchmark is the de facto evaluation of morphing algorithms from academia, research centers and industry. Our august/2022 submission achieved Top-1 in more than half of the sub-datasets and second place in the others, enabling ISR-UC to continue improving morphing attack detection in subsequent publications and applications.

An innovative dynamic identification model for robot manipulators has been developed. The work addresses physical feasibility of robot dynamics identification using Linear Matrix Inequalities (LMIs). Inertia tensor inequalities (namely positive definiteness) have been extensively used among other physical constraints to check consistency of identification methods. Recently, an extra inequality associated to inertia tensor eigenvalues was included to check physical consistency, showing previous methods were incomplete. We showed that this extra inequality incorporates positive definiteness while being more restrictive, and included it in the LMI framework to obtain fully physical estimates. Its relevance was verified through real 7-DOF robot manipulator experiments, described in the paper Inertia Tensor Properties in Robot Dynamics Identification: A Linear Matrix Inequality Approach, IEEE/ASME Transactions on Mechatronics, Vol. 24, N. 1, pp. 406-411, 2019. Top international researchers validated this model and confirmed its performance. In 2021, a comparative study by Italian authors also showed the method as state-of-the-art in this area. Additionally, this method is used by the groups of Prof. Jean-Jacques Slotine (MIT) and Prof. Bruno Siciliano.