Abstract

The rise of robotic body augmentation brings forth new developments that will transform robotics, human-machine interaction, and wearable electronics. Extra robotic limbs, although building upon restorative technologies, bring their own set of challenges in achieving effective bidirectional human-machine collaboration. The questions are whether people can adjust and learn to use a new robotic limb and whether this is achievable without limiting their other physical capabilities. In realizing successful robotic body augmentation, it's crucial to make sure that introducing an extra (artificial) limb doesn't compromise the functions of a natural (biological) limb. This thesis presents research on robotic body augmentation via extra robotic limbs, merging the definition of theoretical foundations with empirical investigations on the adaptability of the human body and brain to advanced technological integrations. Central to this work is the concept of the 'Neural Resource Allocation Problem', defined and discussed in the introduction of this thesis. It addresses the challenges of integrating augmentative devices with the human body without compromising natural functionalities. Such conceptualization is crucial to ensure that augmentation technologies effectively expand user's capacities rather than simply rerouting resources and replacing an existing function with a different, new one.

Based on this theoretical groundwork, I then proposed operational guidelines and detailed the development and characterization of an ad-hoc human-machine interface based on gaze and diaphragmatic respiration for extra robotic arms. The validation carried out on a virtual extra arm thanks to the neuro-robotic platform engineered for this work and the subsequent testing with an extra robotic arm proved the proposed human-machine interface to be effective and non-intrusive, substantiating the proposed methodology. The in-depth analysis of how users adapt to a toe-controlled robotic thumb that concludes the empirical work reported in this thesis is once again rooted in the conceptual framework detailed at the beginning of the thesis. It offered a window into necessary trade-offs, long term effects and the neural adaptations involved with significant and generalisable augmented-hand motor learning. This thesis contributes to the improvement of targeted human machine interfaces design for extra robotic limbs. The non-intrusive biosignals identified have the potential to be further explored and be applied for the control of degrees of freedom of more sophisticated robotic arms to enable more advanced augmentation. This thesis also contributes to a deeper understanding of the consequences of semi-intensive use of robotic body augmentation at behavioural and neural level.

Publications

Leal Pinheiro, D., Dominijanni, G., Maenza, F. P., Dirat, H., Shokur, S., & Micera, S. (2025). Exploring Skill Generalization with an Extra Robotic Arm for Motor Augmentation. Advanced Intelligent Systems, 2500086. https://doi.org/10.1002/aisy.202500086

Clode, D., Dowdall, L., da Silva, E., Selén, K., Cowie, D., Dominijanni, G., & Makin, T. R. (2024). Evaluating initial usability of a hand augmentation device across a large and diverse sample. Science robotics, 9(90), eadk5183. DOI: 10.1126/scirobotics.adk5183

Dominijanni, G., Pinheiro, D. L., Pollina, L., Orset, B., Gini, M., Anselmino, E., ... & Micera, S. (2023). Human motor augmentation with an extra robotic arm without functional interference. Science Robotics, 8(85), eadh1438. DOI: 10.1126/scirobotics.adh1438

Ozgur, A. G., Wessel, M. J., Olsen, J. K., Cadic-Melchior, A. G., Zufferey, V., Johal, W., ... & Hummel, F. C. (2022). The effect of gamified robot-enhanced training on motor performance in chronic stroke survivors. Heliyon, 8(11). DOI: 10.1016/j.heliyon.2022.e11764 

Dominijanni, G., Shokur, S., Salvietti, G., Buehler, S., Palmerini, E., Rossi, S., ... & Micera, S. (2021). The neural resource allocation problem when enhancing human bodies with extra robotic limbs. Nature Machine Intelligence, 3(10), 850-860. https://doi.org/10.1038/s42256-021-00398-9

Abstract

Embargo

Publications

Dynamics of the judgment of tactile stimulus intensity under stable and volatile conditions. https://hdl.handle.net/20.500.14242/168635

Darani, Z. Y., Hachen, I., & Diamond, M. E. (2023). Dynamics of the judgment of tactile stimulus intensity. Neuromorphic Computing and Engineering, 3(1), 014014.

Abstract

Developments in robotic tactile sensing, including the transduction of physical forces analogous to biological mechanoreception, have significantly enhanced robots' abilities to grasp, manipulate objects, and explore their environments. The integration of touch has driven advancements in perception and memory models capable of processing and actively storing tactile information. Emulating biological memory systems holds great promise for advancing autonomous systems by enabling robots to store and utilise memories from experiences, thereby performing complex tasks using contextual sensorimotor information.

This work investigates the use of non-linear probabilistic dimensionality reduction techniques to abstract fundamental functional properties of memory, such as compression, pattern separation, and pattern completion. Additionally, a multiview learning approach is proposed as a unified model for memory and perception of tactile properties critical to the execution of sensorimotor tasks. Improvements in the predictive capabilities for geometric and spatial quantities are demonstrated through the application of hierarchical non-parametric probabilistic models.

The results of this work highlight the robustness of probabilistic models in representing memory and perception of tactile properties, even with relatively small datasets. Furthermore, their efficacy in enabling the execution of sensorimotor tasks featuring active touch sensing underscores their potential for advancing robotic tactile perception.

Publications

Missing info

Abstract

Objective: Traditional theories in neuroscience state that tactile afferents present in the glabrous skin of the human hand encode tactile information following a submodality segregation strategy, meaning that each modality (eg. motion, vibration, shape, ... ) is encoded by a different afferent class. Modern theories suggest a submodality convergence instead, in which different afferent classes work together to capture information about the environment through tactile sense. Typically, studies involve electrophysiological recordings of tens of afferents. At the same time, the human hand is filled with around 17.000 afferents. In this thesis, we want to tackle the theoretical gap this poses. Specifically, we aim to address whether the peripheral nervous system relies on population coding to represent tactile information and whether such population coding enables us to disambiguate submodality convergence against the classical segregation. Approach: Understanding the encoding and flow of information in the nervous system is one of the main challenges of modern neuroscience. Neural signals are highly variable and may be non-linear. Moreover, there exist several candidate codes compatible with sensory and behavioral events. For example, they can rely on single cells or populations and also on rate or timing precision. Information-theoretic methods can capture non-linearities while being model independent, statistically robust, and mathematically well-grounded, becoming an ideal candidate to design pipelines for analyzing neural data. Despite information-theoretic methods being powerful for our objective, the vast majority of neural signals we can acquire from living systems makes analyses highly problem-specific. This is so because of the rich variety of biological processes that are involved (continuous, discrete, electrical, chemical, optical, ...). Main results: The first step towards solving the aforementioned challenges was to have a solid methodology we could trust and rely on. Consequently, the first deliverable from this thesis is a toolbox that gathers classical and state-of-the-art information-theoretic approaches and blends them with advanced machine learning tools to process and analyze neural data. Moreover, this toolbox also provides specific guidance on calcium imaging and electrophysiology analyses, encompassing both simulated and experimental data. We then designed an information-theoretic pipeline to analyze large-scale simulations of the tactile afferents that overcomes the current limitations of experimental studies in the field of touch and the peripheral nervous system. We dissected the importance of population coding for the different afferent classes, given their spatiotemporal dynamics. We also demonstrated that different afferent classes encode information simultaneously about very simple features, and that combining classes increases information levels, adding support to the submodality convergence theory. Significance: Fundamental knowledge about touch is essential both to design human-like robots exhibiting naturalistic exploration behavior and prostheses that can properly integrate and provide their user with relevant and useful information to interact with their environment. Demonstrating that the peripheral nervous system relies on heterogeneous population coding can change the designing paradigm of artificial systems, both in terms of which sensors to choose and which algorithms to use, especially in neuromorphic implementations.

Publications

Corniani, G., Casal, M. A., Panzeri, S., & Saal, H. P. (2022). Population coding strategies in human tactile afferents. PLOS Computational Biology, 18(12), e1010763.

Maffulli, R., Casal, M. A., Celotto, M., Zucca, S., Safaai, H., Fellin, T., & Panzeri, S. (2022). NIT: an open-source tool for information theoretic analysis of neural population data. bioRxiv, 2022-12.

Abstract

New demands in artificial intelligence, the increase of data available, and the forecasts of reaching Moore’s law ceiling push algorithms towards the edge for low latency, low power, and highly intelligent devices. Neuromorphic systems mimic brain-like computations to capture the efficiency and adaptive behaviour exhibited in biological systems being promising candidates to lead the new generation of artificial systems. Sensory information is encoded with precise spatio-temporal patterns in the nervous system. Similarly, neuromorphic computation utilises event-based representations in its computational systems, inspiring its use for building artificial cognitive systems. In this work, we build Spiking Neural Networks (SNNs) in the tactile and auditory sensory modalities for classification tasks in the context of full event-based sensory systems. Compared with standard machine learning implementations in GPU, SNNs implemented in neuromorphic hardware output close to 500 times less energy consumption highlighting the strengths of the new hardware paradigm. Also, we explore how to encode spatio-temporal features from sensing devices emphasising the benefits of using full neuromorphic event-based sensors and systems. We further explore improvements in spiking networks based on the Leaky-Integrate-and-Fire (LIF) models. The proposed network’s architecture implements Time Difference Encoders (TDEs), a cell model based on brain-inspired computations. We find promising results with a 92% reduction of synaptic operations and highly interpretable network results against typical current-based recurrent LIF networks. The results of this work contribute towards building neuromorphic sensing systems from sensor devices, algorithms and circuit design to quantify the strengths of the new computational paradigm with promising capabilities for intelligent machines on edge inspired by their biological counterparts.

Publications

Pequeño-Zurro, A., Khacef, L., Panzeri, S., & Chicca, E. (2025). Towards efficient keyword spotting using spike-based time difference encoders. arXiv preprint arXiv:2503.15402.

Panzeri, S., Janotte, E., Pequeño-Zurro, A., Bonato, J., & Bartolozzi, C. (2023). Constraints on the design of neuromorphic circuits set by the properties of neural population codes. Neuromorphic computing and engineering, 3(1), 012001.

Abstract

Heparin and heparan sulfate (HS) glycosaminoglycans (GAGs), essential components of the extracellular matrix (ECM), regulate a vast array of biological processes by modulating protein interactions and cellular signaling. Their structural complexity, driven by diverse sulfation patterns, underpins their broad physiological activities, including cell proliferation, differentiation, and neural development. However, the intrinsic heterogeneity of native HS complicates the elucidation of structure-activity relationships (SAR), hindering efforts to fully harness their therapeutic potential. This challenge underscores the need for structurally defined HS glycomimetics that can replicate the functional diversity of native HS with precise control and optimized therapeutic effects. 2 Here, we present a library of HS glycomimetics, rationally designed using molecular modeling and synthesized through a divergent synthesis strategy that allows sulfate groups to be installed at specific positions along the carbohydrate backbone. Biophysical characterizations unveil that these glycomimetics selectively bind and stabilize growth factors, including fibroblast growth factors (FGF-1, FGF-2) and nerve growth factor (NGF), in a sulfation-dependent manner without inducing anticoagulant activity, which is a critical prerequisite for successful clinical translation in nerve regeneration. The lead glycomimetic has neuritogenic ability because in two neuronal cell models, PC12 and SH-SY5Y, it enhances NGF-mediated neural maturation when immobilized on a surface. Moreover, functional studies in primary rat hippocampal neurons reveal that the lead glycomimetic potentiates FGF-2- mediated neurite outgrowth and spontaneous synaptic activity, effectively translating its molecular interactions into measurable cellular responses. By bridging molecular-level insights with functional bioactivity, this work establishes HS glycomimetics as precision tools for neurotrophic signaling. Their ability to fine-tune growth factor activity offers a versatile platform for regenerative medicine, extending beyond neural regeneration to broader tissue repair applications. These findings advance the development of next-generation carbohydrate-based therapeutics, unlocking new opportunities for precise and targeted regenerative strategies.

Publications

Riva, E. R., Özkan, M., Contreras, E., Pawar, S., Zinno, C., Escarda-Castro, E., ... & Navarro, X. (2024). Beyond the limiting gap length: peripheral nerve regeneration through implantable nerve guidance conduits. Biomaterials science, 12(6), 1371-1404.

Redolfi Riva, E., Özkan, M., Stellacci, F., & Micera, S. (2024). Combining external physical stimuli and nanostructured materials for upregulating pro-regenerative cellular pathways in peripheral nerve repair. Frontiers in Cell and Developmental Biology, 12, 1491260.

Abstract

Humanoid robots have garnered significant attention for their adaptability to dynamic tasks. Despite advancements in vision-based sensing and deep learning, tactile sensing -- crucial for safe human-robot interaction and dexterous manipulation -- remains underutilized. This thesis addresses the challenges of integrating touch, a fundamental human sensory modality, into robotic systems, focusing on bio-inspired approaches and neuromorphic computation.

The human tactile system, characterized by mechanoreceptors, neural pathways, and proprioceptive integration, provides a blueprint for designing tactile sensors. Drawing on these principles, recent developments in tactile sensing technologies are reviewed, from low-resolution designs to high-resolution sensors enabled by advanced manufacturing. This work highlights the advantages of event-driven processing for efficient tactile encoding, which mimics biological systems by transmitting data only upon input changes. Using these methods, a novel tactile Braille dataset was created, featuring the iCub fingertip sliding over 3D-printed Braille letters. Encoding strategies inspired by Fast Adaptive (FA) mechanoreceptors reduced data by up to 122 times. Neuromorphic hardware (Intel Loihi) demonstrated significant power savings during letter classification, achieving efficiency gains over traditional GPU-based methods, though at a cost to accuracy.

To further advance encoding methods, the WiN-GUI was developed, allowing researchers to visualize and optimize spike-pattern encoding in real-time. This tool integrates temporal, spatial, and spatio-temporal datasets to evaluate neuron behavior and classification performance, with applications extending beyond tactile sensing. Complementing this, human haptic object exploration was analyzed to uncover kinematic synergies and Exploratory Procedures (EPs). Using a multimodal dataset of blindfolded participants solving tactile discrimination tasks, this study identified recurring movement patterns and dimensionality reduction strategies that could inform robotic hand control.

By bridging insights from biology, neuromorphic engineering, and robotics, this work offers scalable solutions for touch-sensitive robots designed for dynamic and human-centric environments, paving the way for safer and more capable systems.

Publications

Müller-Cleve, S. F., Fra, V., Khacef, L., Pequeño-Zurro, A., Klepatsch, D., Forno, E., ... & Bartolozzi, C. (2022). Braille letter reading: A benchmark for spatio-temporal pattern recognition on neuromorphic hardware. Frontiers in Neuroscience, 16, 951164.

Di Domenico, D., Forsiuk, I., Müller-Cleve, S., Tanzarella, S., Garro, F., Marinelli, A., ... & Semprini, M. (2025). Reach&Grasp: a multimodal dataset of the whole upper-limb during simple and complex movements. Scientific Data, 12(1), 233.

Müller-Cleve, S. F., Quintana, F. M., Fra, V., Galindo, P. L., Perez-Peña, F., Urgese, G., & Bartolozzi, C. (2024). WiN-GUI: A graphical tool for neuron-based encoding. SoftwareX, 27, 101759.

Fra, V., Müller-Cleve, S. F., Urgese, G., & Bartolozzi, C. (2025). WiN-GUI Version 2: A graphical tool for neuron-based encoding. SoftwareX, 29, 102031.

Abstract

For humans, the sense of touch is of immeasurable importance. It starts with babies that hold on to fingers and explore their surroundings in a tactile fashion, using their hands and mouth. Touch supports seamless and safe physical interaction with the external world to accomplish tasks by grasping and confidently handling objects, detecting when they are about to slip and react accordingly. Touch information conveys the texture, shape and material of objects we interact with, complementing vision, especially in the dark, or when there are occlusions. Finally, social and affective touch, such as hugs, strokes, or handshakes, are crucial in development and influence well-being and social interactions.

With the increasing development of autonomous robots for laborious or tedious tasks, as well as for tasks that require high precision in industry, agriculture, surgery, healthcare or household, the main tasks the robots would have to fulfil require them to safely and robustly interact with objects and people. Interestingly, many of those (object manipulation, slip detection, grip strength modulation) often rely on vision-based algorithms. This is heavily related to the fact that many AI and machine learning approaches are based on visual inputs, as vision can count on large available standardised datasets and a long history of computer vision development.

However, this comes with high computational cost and low performance when the environment is dark or parts of the visual field are obstructed. Further, when purely visual information is processed, important cues such as the time and quality of contact are missed. As such, there is a need to equip robots with tactile sensors extensively. Not only robots but also prosthetic limbs benefit from the application of tactile sensors and the possibility of providing tactile feedback, to improve the sense of ownership and decrease the attentional burden of using vision to manipulate objects.

Unlike artificial tactile sensors that periodically sample the pressure applied to them, the human sense of touch is based on a family of sensors embedded in the skin (mechanoreceptors), that only provide information to the brain when a stimulus is applied. Mechanoreceptors are grouped into different categories depending on their sensitivities and responses to stimulus types. Different mechanoreceptors families rely on physical transducers with different size, placement, adaptation and responsiveness, to capture the diverse information about sensory stimuli. This encoding substrate supports the rich, robust, accurate and efficient encoding of the information about contact needed for physical interaction.

The goal of this PhD thesis is to design and investigate circuits for the implementation of artificial tactile sensors that operate similarly to their biological counterparts. As such, it focusses on contact-driven encoding of pressure information from a variety of pressure transducers, such as capacitive and piezoelectric, and on capturing information about static and dynamic pressure at different bandwidths. The approach used in the circuit development is that of Neuromorphic Engineering, with a focus on mixed-mode subthreshold circuits, to achieve low power, low latency, compactness and adaptivity.

Publications

Panzeri, S., Janotte, E., Pequeño-Zurro, A., Bonato, J., & Bartolozzi, C. (2023). Constraints on the design of neuromorphic circuits set by the properties of neural population codes. Neuromorphic computing and engineering, 3(1), 012001.

D'Angelo, G., Janotte, E., Schoepe, T., O'Keeffe, J., Milde, M. B., Chicca, E., & Bartolozzi, C. (2020). Event-based eccentric motion detection exploiting time difference encoding. Frontiers in neuroscience, 14, 451.

Janotte, E., Bamford, S., Richter, O., Valle, M., & Bartolozzi, C. (2022, June). Neuromorphic capacitive tactile sensors inspired by slowly adaptive mechanoreceptors. In 2022 20th IEEE Interregional NEWCAS Conference (NEWCAS) (pp. 119-123). IEEE.

Janotte, E., Mastella, M., Chicca, E., & Bartolozzi, C. (2021). Touch in robots: A neuromorphic approach. ERCIM News, (125), 34-51.

Abstract

The human peripheral tactile system is responsible for the initial processing of tactile stimuli and is composed of the skin and various embedded mechanoreceptors innervated by afferents. Spiking models are widely used to characterize this system and infer how populations of afferents shape tactile perception. Leveraging existing models of tactile afferents and moved by their limitations, we present three studies designed to advance these essential tools in the investigation of the human peripheral tactile system.

Firstly, reconciling existing evidence, we quantitatively characterize the population of peripheral tactile afferents. We estimate that approximately 230,000 afferents cover the human body, provide innervation densities in different skin areas, and show the relation of these numbers with tactile acuity, hair follicle density, and somatosensory cortical representation.

Secondly, we ask how tactile afferents work together to encode information in complex ways. We find that information is spread across classes, and combining information from multiple classes improves transmission. We test the importance of temporal and spatial resolution in the population code, probing that destroying temporal information is more destructive than removing spatial information.

Finally, we use Optical Coherence Tomography to image the skin subsurface in vivo and dynamically and quantify the deformation of individual fingerprint ridges down to the type-1 mechanoreceptors' location. When scanning the skin with a flat surface, the ridge deforms as a single unit. Higher strains emerge from the stick-to-slip transition compared to plate movement reversal. When scanning the skin with small features, different ridge sub-units experience different strain patterns. Higher strains occur in the deepest layer imaged.

Overall, this research provides a better understanding of coding strategies of tactile afferents on a population level and of the link between skin mechanics and transduction mechanisms underlying tactile perception. Our findings will have implications for developing novel spiking models of the human peripheral tactile system.

Publications

Corniani, G., & Saal, H. P. (2020). Tactile innervation densities across the whole body. Journal of Neurophysiology, 124(4), 1229-1240.

Corniani, G., Casal, M. A., Panzeri, S., & Saal, H. P. (2022). Population coding strategies in human tactile afferents. PLOS Computational Biology, 18(12), e1010763.

Corniani, G., Lee, Z. S., Carré, M. J., Lewis, R., Delhaye, B. P., & Saal, H. P. (2024). Sub-surface deformation of individual fingerprint ridges during tactile interactions. eLife, 13:RP93554.

Corniani, G. (2022, May). Imaging Sub-surface Skin Strain Patterns During Fingertip Sliding. In Haptics: Science, Technology, Applications: 13th International Conference on Human Haptic Sensing and Touch Enabled Computer Applications, EuroHaptics 2022, Hamburg, Germany, May 22–25, 2022, Proceedings (Vol. 13235, p. 358). Springer Nature.

Abstract

Tactile feedback is essential for object manipulation tasks for humans and robots, facilitating reactive control even without vision. This dissertation introduces several tactile-based methodologies to enhance the manipulation performance and safety of robotic systems. Before presenting these methods, the dissertation offers a high-level overview of tactile feedback in human manipulation tasks. Research indicates that humans divide manipulation tasks into distinct action phases, each defined by specific subgoals and guided by tactile cues. This concept is further refined in this thesis through the introduction of three manipulation task stages: grasping, manipulation, and placing. These stages provide a structured framework for categorizing the proposed methods and organization of the thesis. The first approach introduced in this thesis is a novel grasp force controller that is split into control phases, similar to human grasping, and implements several object safety features: It prevents undesired object motions during the grasp by halting finger movements based on tactile cues. Additionally, it maintains a given target force even under external disturbances while holding the object. Subsequently, a reinforcement learning scheme is introduced that enables zero-shot sim-to-real transfer of continuous grasp force control policies. A novel simulation environment is proposed to facilitate this, modeling a grasping scenario with realistically varied objects in size and stiffness. A comparison of these two methods facilitates a broader discussion on the advantages of learning-based methods while also highlighting the beneficial properties of classical controllers. For the manipulation stage, a simulation-based study is detailed, proposing and comparing different tactile sensor configurations for an anthropomorphic robot hand. The performance of each sensor configuration is evaluated by integrating its data into a policy trained on various in-hand manipulation tasks. This comparison of experimental results leads to the formulation of recommendations for developing future end-effector sensorizations. In the fourth and final approach, a supervised learning method for object placing is proposed. Given the challenges of occlusions in vision-based methods in placing scenarios, tactile data proves particularly valuable. The thesis concludes with a comparative analysis of the models, tasks, and sensors employed across the presented approaches.

Publications

Lach, L., Haschke, R., Tateo, D., Peters, J., Ritter, H., Borràs, J., & Torras, C. (2024, October). Zero-Shot Transfer of a Tactile-based Continuous Force Control Policy from Simulation to Robot. In 2024 IEE

 Lach, L., Haschke, R., Tateo, D., Peters, J., Ritter, H., Borràs, J., & Torras, C. (2023). Towards transferring tactile-based continuous force control policies from simulation to robot. arXiv preprint arXiv:2311.07245.

Lach, L., Lemaignan, S., Ferro, F., Ritter, H., & Haschke, R. (2022, October). Bio-inspired grasping controller for sensorized 2-dof grippers. In 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 11231-11237). IEEE.

Melnik, A., Lach, L., Plappert, M., Korthals, T., Haschke, R., & Ritter, H. (2021). Using tactile sensing to improve the sample efficiency and performance of deep deterministic policy gradients for simulated in-hand manipulation tasks. Frontiers in Robotics and AI, 8, 538773.

Melnik, A., Lach, L., Plappert, M., Korthals, T., Haschke, R., & Ritter, H. (2019, November). Tactile sensing and deep reinforcement learning for in-hand manipulation tasks. In IROS workshop on autonomous object manipulation (Vol. 39, pp. 3-20).

Lach, L., Funk, N., Haschke, R., Lemaignan, S., Ritter, H. J., Peters, J., & Chalvatzaki, G. (2023, October). Placing by Touching: An empirical study on the importance of tactile sensing for precise object placing. In 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 8964-8971). IEEE.

Lach, L., Haschke, R., & Ferro, F. (2020). Leveraging Touch Sensors to Improve Mobile Manipulation. arXiv preprint arXiv:2010.10810.

Lach, L., Haschke, R., & Ferro, F. (2021). Improving Manipulation Performance of Mobile Robots Using Tactile Sensors. In 10th International IEEE EMBS Conference on Neural Engineering (NER’21).

Niemann, C., Leins, D., Lach, L., & Haschke, R. (2024, October). Learning When to Stop: Efficient Active Tactile Perception with Deep Reinforcement Learning. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 685-692). IEEE.

Lach, L., Ferro, F., & Haschke, R. (2023). TIAGo RL: Simulated Reinforcement Learning Environments with Tactile Data for Mobile Robots. arXiv preprint arXiv:2311.07260.

Abstract

Touch, one of our primary senses, allows us to explore and understand the physical world around us. It involves the detection of mechanical stimuli, such as pressure, vibration, and texture, through specialized receptors distributed across our skin. Robots and prostheses struggle to replicate the nuanced sense of touch found in humans. This limitation degrades their ability to perform tasks reliant on tactile perception, such as object manipulation and texture discrimination, hindering their overall functionality.

For this reason, there is a critical need to enhance these technologies with advanced tactile sensing capabilities to improve their versatility and enable more natural interactions.

In this thesis, we demonstrate how, using neuromorphic principles inspired by neuroscientific literature, we can address these limitations. In the first part of the thesis, we investigate how taking inspiration from mechanoreceptors can inform the design of novel sensors capable of encoding pressure into spike patterns. Following this, we explore the decoding of these signals using networks composed only of spiking neurons and synapses, drawing inspiration from biological findings. The resulting architectures show dynamics qualitatively resembling evidence from neuroscientific experiments. Finally, we demonstrate how the networks we designed can be translated onto CMOS hardware for future deployments in real-world agents.

Our results underscore the importance of leveraging neuroscientific literature to inform the design of future technologies for tactile perception in artificial agents.

This paradigm promises extremely powerful given the shared constraints between artificial systems and their biological counterparts.

This work lays down a compelling example of drawing inspiration from biology to enhance the design of artificial agents with tactile capabilities. As such, the results presented in this thesis pave the way for further research endeavors informed by the
blueprint of natural systems, encouraging neuromorphic engineers to explore the knowledge available in biology to inform their own designs. Embracing biological principles not only facilitates the development of more efficient and effective artificial agents but also fosters a deeper understanding of the natural world.

Publications

Mastella, M., Tiemens, T., & Chicca, E. (2024). Texture Recognition Using a Biologically Plausible Spiking Phase-Locked Loop Model for Spike Train Frequency Decomposition. arXiv preprint arXiv:2403.09723.

Mastella, M., & Chicca, E. (2021, May). A hardware-friendly neuromorphic spiking neural network for frequency detection and fine texture decoding. In 2021 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 1-5). IEEE.

Mastella, M., Greatorex, H., Tiemens, T., Cotteret, M., Richter, O., & Chicca, E. (2024, October). Event-Driven Frequency Decomposition with Spiking Phase-Locked Loops. In 2024 58th Asilomar Conference on Signals, Systems, and Computers (pp. 1517-1521). IEEE.

Greatorex, H., Richter, O., Mastella, M., Cotteret, M., Klein, P., Fabre, M., ... & Chicca, E. (2025). A neuromorphic processor with on-chip learning for beyond-CMOS device integration. Nature communications, 16(1), 6424.

Greatorex, H., Richter, O., Mastella, M., Cotteret, M., Klein, P., Fabre, M., ... & Chicca, E. (2024). TEXEL: A neuromorphic processor with on-chip learning for beyond-CMOS device integration. arXiv preprint arXiv:2410.15854.

Sengupta, D., Mastella, M., Chicca, E., & Kottapalli, A. G. P. (2022). Skin-inspired flexible and stretchable electrospun carbon nanofiber sensors for neuromorphic sensing. ACS applied electronic materials, 4(1), 308-315.

Cipollini, D., Greatorex, H., Mastella, M., Chicca, E., & Schomaker, L. (2025). Fused-MemBrain: a spiking processor combining CMOS and self-assembled memristive networks. Neuromorphic Computing and Engineering, 5(2), 024002.

Mastella, M., Greatorex, H., Cotteret, M., Janotte, E., Soares Girao, W., Richter, O., & Chicca, E. (2023, August). Synaptic normalisation for on-chip learning in analog CMOS spiking neural networks. In Proceedings of the 2023 International Conference on Neuromorphic Systems (pp. 1-4).

Dabbous, A., Mastella, M., Natarajan, A., & Chicca, E. (2021, May). Artificial bio-inspired tactile receptive fields for edge orientation classification. In 2021 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 1-5). IEEE.

Dabbous, A., Mastella, M., Natarajan, A., & Chicca, E. (2021, May). Artificial bio-inspired tactile receptive fields for edge orientation classification. In 2021 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 1-5). IEEE.

Richter, O., Greatorex, H., Hucko, B., Cotteret, M., Soares Girao, W., Janotte, E., ... & Chicca, E. (2023, August). A subthreshold second-order integration circuit for versatile synaptic alpha kernel and trace generation. In Proceedings of the 2023 International Conference on Neuromorphic Systems (pp. 1-4).

Cotteret, M., Richter, O., Mastella, M., Greatorex, H., Janotte, E., Girão, W. S., ... & Chicca, E. (2023, May). Robust spiking attractor networks with a hard winner-take-all neuron circuit. In 2023 IEEE International Symposium on Circuits and Systems (ISCAS) (pp. 1-5). IEEE.

Greatorex, H., Mastella, M., Cotteret, M., Richter, O., & Chicca, E. (2025). Event-based vision for egomotion estimation using precise event timing. arXiv preprint arXiv:2501.11554.

Greatorex, H., Mastella, M., Richter, O., Cotteret, M., Girão, W. S., Janotte, E., & Chicca, E. (2025). A scalable event-driven spatiotemporal feature extraction circuit. arXiv preprint arXiv:2501.10155.

Janotte, E., Mastella, M., Chicca, E., & Bartolozzi, C. (2021). Touch in robots: A neuromorphic approach. ERCIM News, (125), 34-51.

Abstract

Electronic skins (e-skins) are systems designed to provide robots and prosthetics with sensing capabilities similar to human skin’s. To replicate human skin’s high sensing acuity, e-skin must feature dense sensor arrays, continuously generating large datasets. This leads to high computational and energy costs in conventional computing architectures. An efficient alternative is offered by neuromorphic computing paradigms of localised, asynchronous, and parallel processing. The structural properties of e-skin require the neuromorphic circuits to be flexible and large-area. To satisfy these requirements, this thesis focuses on the development of the fundamental unit devices for neuromorphic e-skin flexible circuits via printed electronics-based fabrication processes.

As active electronic materials, ZnO nanowires (NWs) are selected for their versatile functionality (allowing both sensing and resistive switching), flexibility, and integrability over large area. To develop dense printed circuits, NW integration must also be site-selective. Hence, two suitable NW integration processes were developed: (i) selective NW removal, a lithography-free technique that allows high fidelity patterning of in-plane NW arrays, and (ii) selective hydrothermal synthesis, which allows the out-of-plane growth of NWs only in areas defined by the high-resolution electrohydrodynamic printing technique. These two complementary integration methods were employed for the fabrication of e-skin sensing and processing devices.

As sensing elements, flexible UV photodetectors are developed via a fully printed process, showing high performance (responsivity of 1.4×10⁷ A W⁻¹, record-high among ZnO-based photodetectors) with optimal performance retention upon bending. Then, digital memristors were fabricated via a roll-to-roll compatible process. These devices demonstrated high resistance switching (ON/OFF ratio ~10⁴) at ultralow biocompatible voltages (0.07 V SET, −0.06 V RESET) and robust performance under mechanical bending. Lastly, optoelectronic synapses are created by printing ZnO NWs over nanogap-separated electrodes. These devices demonstrated high bio-plausibility under both electrical and optical stimulation, exhibiting short-term synaptic plasticity functions, spike-rate-dependent plasticity, and transition from short- to long-term memory.

Publications

Luca De Pamphilis, Abhishek Singh Dahiya, Yuxin Xia, Evangelos Moutoulas, Dimitra G. Georgiadou, and Ravinder Dahiya, Coplanar Electrodes based Dual-Modulated Optoelectronic Memristive Synaptic Devices, NPJ Flex Electronics, 2025 (under review).

Luca De Pamphilis, Sihang Ma, Abhishek Singh Dahiya, Adamos Christou, and Ravinder Dahiya, Site-Selective Nanowire Synthesis and Fabrication of Printed Memristor Arrays with Ultralow Switching Voltages on Flexible Substrate, ACS Applied Materials & Interfaces 2024 16 (44), 60394-60403. DOI: 10.1021/acsami.4c07172

F. Liu, A. Christou, R. Chirila, L. De Pamphilis, R. Dahiya, Stochastic Nature of Large-Scale Contact Printed ZnO Nanowires Based Transistors. Adv. Funct. Mater. 2025, 35, 2412299. https://doi.org/10.1002/adfm.202412299

De Pamphilis, L., Dahiya, A. S., Christou, A., Ma, S., & Dahiya, R. (2022). Patterned assembly of inorganic semiconducting nanowires using lithography-free technique. IEEE Journal on Flexible Electronics, 2(2), 223-232.

Ma, S., Dahiya, A. S., Christou, A., De Pamphilis, L., & Dahiya, R. (2023). All-Printed ZnO Nanowire-Based High Performance Flexible Ultraviolet Photodetectors. IEEE Journal on Flexible Electronics, 2(2), 216-222.

De Pamphilis, L., Dahiya, A. S., Ma, S., & Dahiya, R. (2023, July). Printed Memristors Using Hydrothermally Grown Zinc Oxide Nanowires. In 2023 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS) (pp. 1-4). IEEE.

De Pamphilis, L., Christou, A., Dahiya, A. S., & Dahiya, R. (2022, July). Selective removal of contact printed nanowires for lithography-free patterning. In 2022 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS) (pp. 1-4). IEEE.

Neto, J., Dahiya, A. S., Christou, A., Zumeit, A., De Pamphilis, L., & Dahiya, R. (2023, July). Dual-Gate Transistors Using Contact Printed ZnO Nanowires. In 2023 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS) (pp. 1-4). IEEE.

Luca De Pamphilis, Abhishek S. Dahiya, Ravinder Dahiya, ZnO Nanowire Based Flexible Transient Ultraviolet Photodetectors, Editor(s): A.S.M.A. Haseeb, Encyclopedia of Materials: Electronics, Academic Press, 2023, Pages 85-96, ISBN 9780128197356, https://doi.org/10.1016/B978-0-12-819728-8.00124-8.

Abstract

As the world of electronics quickly expands, emerging materials and novel electronic devices nurture the expansion of more-than-Moore technologies, with special developments made in areas such as flexible electronics and artificial electronic skins (e-skins). Such developments are promising when it comes to deliver natural touch sensation to prosthetic users, or next-generation robotic applications. Nevertheless, replicating the human skin is not an easy feat as it contains thousands of specialized receptors that instantly responding to pressures, vibrations, temperature and noxious cues. In this context, artificial skins are systems that replicate the sense of touch by means of soft electronic sensors. These require flexibility in its nature, to conform to rigid parts such as robotic arms, while meeting requirements such as the miniaturization and densification of devices over large areas. Evidently, processing artificial skins packed with such sensing arrays comes at with high power consumption and high computational cost. To this end, emerging concepts such as neuromorphic computing carry great promise, as they are capable of providing event-driven, parallel and energy efficient computation. Following such motivations, this thesis focusses on the coalition of flexible nanomaterial-based electronics and emerging neuromorphic devices, for the development of building blocks of next generation neuromorphic electronic skins. The demonstrated flexible devices consist in highly crystalline inorganic nanowires (NWs) with high aspect ratio (>1000), enabling superior mechanical properties. The functional NWs are deployed at precise locations over flexible substrates through dielectrophoretic (DEP) solution processable assembly. The versatility of the setup is shown through the assembly of different materials such as Vanadium Pentoxide NWs (V2O5) and Zinc Oxide (ZnO) over large areas. The aligned V2O5 NWs are used as active channel for highly sensitive thermal sensors, fabricated through the novel high-resolution electrohydrodynamic (EHD) jet printing. The ultra-small thermoreceptor exhibits high sensitivity (-1.1 ± 0.3%ºC-1), fast response (≈1s) and exceptional stability. The device capabilities are demonstrated through the reflex to thermal pain on a robotic arm.

Moreover, DEP assembled ZnO nanowires and transfer printed high-mobility Si nanoribbons (Si NRs) are used as active channel for flexible top-gated transistors. Given the high mobility of doped silicon NRs, high-performance printed n- and p-type FETs are developed through EHD jet printing of metal and encapsulation layers. The printed Si NR transistors show effective peak mobilities of 15 cm2/Vs (n-channel) and 5 cm2/Vs (p-channel) at low 1 V drain voltage, with good stability after 10000 bending cycles at different bending radius (40, 25, and 15 mm). Lastly, neuron-like transistors are developed using the DEP aligned ZnO NWs. The device incorporates a floating gate (FG) which capacitively couples with multiple top-gates. Such coupling enables voltage-mode summation at the FG, where more than one parallel input modulates the output of the device. Given the observed charge trapping and multiple input configuration, the device exhibits spatiotemporal summation of spike-based inputs, demonstrating the efficacy of the developed neural FET for synaptic applications.

Publications

Neto, J., Dahiya, A.S. & Dahiya, R. Multi-gate neuron-like transistors based on ensembles of aligned nanowires on flexible substrates. Nano Convergence 12, 2 (2025). https://doi.org/10.1186/s40580-024-00472-z

Neto, J., Chirila, R., Dahiya, A. S., Christou, A., Shakthivel, D., & Dahiya, R. (2022). Skin‐inspired thermoreceptors‐based electronic skin for biomimicking thermal pain reflexes. Advanced Science, 9(27), 2201525.

Neto, J., Dahiya, A. S., Zumeit, A., Christou, A., Ma, S., & Dahiya, R. (2023). Printed n-and p-channel transistors using silicon nanoribbons enduring electrical, thermal, and mechanical stress. ACS Applied Materials & Interfaces, 15(7), 9618-9628.

Dahiya, A. S., Christou, A., Neto, J., Zumeit, A., Shakthivel, D., & Dahiya, R. (2022). In Tandem Contact‐Transfer Printing for High‐Performance Transient Electronics. Advanced Electronic Materials, 8(9), 2200170.

Neto, J., Dahiya, A. S., Kumaresan, Y., Shakthivel, D., & Dahiya, R. (2021, June). V 2 O 5 nanowires-based flexible temperature sensor. In 2021 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS) (pp. 1-4). IEEE.

Neto, J., Dahiya, A. S., & Dahiya, R. (2023). Influence of Printed Encapsulation Layer on the Mechanical Reliability and Performance of V₂O₅ Nanowires-Based Flexible Temperature Sensors. IEEE Journal on Flexible Electronics, 2(2), 168-174.

Neto, J., Dahiya, A. S., Christou, A., Zumeit, A., De Pamphilis, L., & Dahiya, R. (2023, July). Dual-Gate Transistors Using Contact Printed ZnO Nanowires. In 2023 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS) (pp. 1-4). IEEE.

Neto, J., Dahiya, A. S., & Dahiya, R. (2022, July). Influence of Encapsulation on the Performance of V 2 O 5 Nanowires-Based Temperature Sensors. In 2022 IEEE International Conference on Flexible and Printable Sensors and Systems (FLEPS) (pp. 1-4). IEEE.