The year 2020 marks the beginning of a new period for CominLabs with its 5-year extension. After the CominLabs 2020 call, seven projects have been selected by the CominLabs steering committee based on the International Advisory Committee evaluation.
The Affinity project will investigate, qualitatively and quantitatively, the influence of “affinity” on the behavior of people suffering from Autism Spectrum Disorder (ASD). The overall goal is to define recommendations adapted to each person with autism according to his/her specific affinity. Affinity is an object, a passion, a field of knowledge, a specified interest on which autist people focus. They develop competences from this strong investment and open up to social link and language. These recommendations will be the results of two complementary works following two different approaches. The first one will produce statistical measurements related to computer vision and data mining frameworks, which will quantitatively evaluate the influence of affinity on the behaviors of ASD people. The second one will be related to psychoanalysis that considers each person as unique, and bases its ethics on the autism singularity by working with elective affinity specific to the autistic mind.
Eye tracking experiments on affinity images and neutral images. The tracking ratio on affinity images is much higher.
The DATERAC project will develop two innovative solutions for microwave reconfigurable circuits and antennas for future communication systems, in the context of both civilian and military applications (5G and beyond, automotive radar, IoT, satellite constellations). The first solution will rely on the electrical control of semi-conductor junctions directly integrated in a silicon substrate through localized doped areas. It will offer great flexibility in terms of size and shape of the junctions and more generally in the design of reconfigurable devices. The second solution will leverage the optical control of the state (amorphous or crystalline) of chalcogenide glasses, leading to an important conductivity change. The final objective will be to combine the two reconfiguration solutions to build complex systems allowing to modify independently and simultaneously several characteristics of microwave components and antennas (central frequency, bandwidth, radiation pattern, polarization).
Chalcogenide glasses: (a) amorphous state (b) crystalline state.
The dnarXiv project will investigate the archiving of information on DNA molecules.
Indeed, DNA offers the advantage of a storage density of 1000 to 10000 times greater than current technologies. The objective is to set up a prototype to demonstrate the feasibility of this new storage possibility based on the latest bio-technologies: enzymatic synthesis (writing information) and sequencing nanopore (reading information). From a numerical point of view, the research axes focus on how to encode information inside DNA molecules and on how to ensure their safety.
When neural nets meet physics.
Neural networks are powerful models used in machine learning, but poorly understood from a theoretical point of view. A recent line of research consists in studying the flow of information through or within these networks from the perspective of dynamical systems and their associated physics. The Dynalearn project will explore how dynamical formulation of learning process can help in understanding better deep neural architectures, as well as proposing new learning paradigms based on the regularization of the information flow. In addition, by leveraging novel neural architectures and available data, DynaLearn will design new data-driven dynamical simulation models, with applications in earth observation and medical imaging.
Aerial robots meet many industrial needs, but their gripping capacity, reduced to a rigid gripper, is not adapted to complex objects. The MAMBO project (MAnipulation with Multiples drones for soft BOdies) proposes a novel solution: a soft gripper composed of a soft body actuated by drones, and capable of withstanding large deformation and thus adapting to any object shape. Lightweight, flexible clamps can be larger and have no impact on the autonomy of drones.
Parkinson’s disease affects more than 160 000 patients in France, and the general public is aware of this disease mainly for its motor symptoms. However, patients with Parkinson’s disease also suffer from so-called “non-motor” disorders, affecting memory or the ability to control impulsive behavior. Currently, no drug is able to relieve these disorders, which motivates the development of new therapeutic approaches to improve the quality of life of patients. The PKSTIM project proposes to test, for the first time, the use of a non-invasive technique of electrical stimulation of the brain to improve “working memory” in patients with Parkinson’s disease. Additionally, using computational neuro-electric modeling, the PKSTIM project will assess novel stimulation signals that could maximize the clinical effectiveness of non-invasive electrical stimulation of the brain. The overarching goal of PKSTIM is therefore to improve the quality of life of patients with Parkinson’s disease and to provide an innovative technology for non-invasive brain stimulation that could also be used in other neurological diseases.
Wireless data traffic is increasing by a factor of 100 every ten years and the trend is even stronger with upcoming 5G networks. Data rate is somehow proportional to the total available bandwidth, which can in turn be larger at higher frequencies. Sub-THz wireless communication is able to provide the bandwidths required by future wireless systems beyond 5G. However, wireless communications in such frequency bands are impaired due to the huge propagation losses, several atmospheric absorption peaks, the relatively low output power provided by commercial devices at room temperature, and waveforms not perfectly adapted to the specific features of the sub-THz channel. The WASSAP project will contribute to a new generation of sub-THz wireless systems. It will investigate a photonic approach (arrays of photo-mixers directly excited by optical fibers) to increase the power radiated by transmitting modules. In addition, optimized waveforms will be proposed to satisfy the requirements of the sub-THz channel and to maximize the spectral and energy efficiencies of the developed hardware.