Adaptive and Maladaptive Implications of Reinforcement Learning Processes: Fronto-Striatal Loops and Behavioural Correlates

That humans and animals learn from interaction with the environment is a foundational idea underlying nearly all theories of learning and intelligence. Learning that certain outcomes are associated with specific actions or stimuli (both internal and external), is at the very core of the capacity to adapt behaviour to environmental changes. In the present work, appetitive and aversive reinforcement learning paradigms have been used to investigate the fronto-striatal loops and behavioural correlates of adaptive and maladaptive reinforcement learning processes, aiming to a deeper understanding of how cortical and subcortical substrates interacts between them and with other brain systems to support learning. By combining a large variety of neuroscientific approaches, including behavioral and psychophysiological methods, EEG and neuroimaging techniques, these studies aim at clarifying and advancing the knowledge of the neural bases and computational mechanisms of reinforcement learning, both in normal and neurologically impaired population.

Cooperative communication and distributed detection in wireless sensor networks

Recent progress in microelectronic and wireless communications have enabled the development of low cost, low power, multifunctional sensors, which has allowed the birth of new type of networks named wireless sensor networks (WSNs). The main features of such networks are: the nodes can be positioned randomly over a given field with a high density; each node operates both like sensor (for collection of environmental data) as well as transceiver (for transmission of information to the data retrieval); the nodes have limited energy resources. The use of wireless communications and the small size of nodes, make this type of networks suitable for a large number of applications. For example, sensor nodes can be used to monitor a high risk region, as near a volcano; in a hospital they could be used to monitor physical conditions of patients. For each of these possible application scenarios, it is necessary to guarantee a trade-off between energy consumptions and communication reliability. The thesis investigates the use of WSNs in two possible scenarios and for each of them suggests a solution that permits to solve relating problems considering the trade-off introduced. The first scenario considers a network with a high number of nodes deployed in a given geographical area without detailed planning that have to transmit data toward a coordinator node, named sink, that we assume to be located onboard an unmanned aerial vehicle (UAV). This is a practical example of reachback communication, characterized by the high density of nodes that have to transmit data reliably and efficiently towards a far receiver. It is considered that each node transmits a common shared message directly to the receiver onboard the UAV whenever it receives a broadcast message (triggered for example by the vehicle). We assume that the communication channels between the local nodes and the receiver are subject to fading and noise. The receiver onboard the UAV must be able to fuse the weak and noisy signals in a coherent way to receive the data reliably. It is proposed a cooperative diversity concept as an effective solution to the reachback problem. In particular, it is considered a spread spectrum (SS) transmission scheme in conjunction with a fusion center that can exploit cooperative diversity, without requiring stringent synchronization between nodes. The idea consists of simultaneous transmission of the common message among the nodes and a Rake reception at the fusion center. The proposed solution is mainly motivated by two goals: the necessity to have simple nodes (to this aim we move the computational complexity to the receiver onboard the UAV), and the importance to guarantee high levels of energy efficiency of the network, thus increasing the network lifetime. The proposed scheme is analyzed in order to better understand the effectiveness of the approach presented. The performance metrics considered are both the theoretical limit on the maximum amount of data that can be collected by the receiver, as well as the error probability with a given modulation scheme. Since we deal with a WSN, both of these performance are evaluated taking into consideration the energy efficiency of the network. The second scenario considers the use of a chain network for the detection of fires by using nodes that have a double function of sensors and routers. The first one is relative to the monitoring of a temperature parameter that allows to take a local binary decision of target (fire) absent/present. The second one considers that each node receives a decision made by the previous node of the chain, compares this with that deriving by the observation of the phenomenon, and transmits the final result to the next node. The chain ends at the sink node that transmits the received decision to the user. In this network the goals are to limit throughput in each sensor-to-sensor link and minimize probability of error at the last stage of the chain. This is a typical scenario of distributed detection. To obtain good performance it is necessary to define some fusion rules for each node to summarize local observations and decisions of the previous nodes, to get a final decision that it is transmitted to the next node. WSNs have been studied also under a practical point of view, describing both the main characteristics of IEEE802:15:4 standard and two commercial WSN platforms. By using a commercial WSN platform it is realized an agricultural application that has been tested in a six months on-field experimentation.

Energy-efficient Communication and Estimation in Wireless Sensor Networks

This thesis focuses on the energy efficiency in wireless networks under the transmission and information diffusion points of view. In particular, on one hand, the communication efficiency is investigated, attempting to reduce the consumption during transmissions, while on the other hand the energy efficiency of the procedures required to distribute the information among wireless nodes in complex networks is taken into account. For what concerns energy efficient communications, an innovative transmission scheme reusing source of opportunity signals is introduced. This kind of signals has never been previously studied in literature for communication purposes. The scope is to provide a way for transmitting information with energy consumption close to zero. On the theoretical side, starting from a general communication channel model subject to a limited input amplitude, the theme of low power transmission signals is tackled under the perspective of stating sufficient conditions for the capacity achieving input distribution to be discrete. Finally, the focus is shifted towards the design of energy efficient algorithms for the diffusion of information. In particular, the endeavours are aimed at solving an estimation problem distributed over a wireless sensor network. The proposed solutions are deeply analyzed both to ensure their energy efficiency and to guarantee their robustness against losses during the diffusion of information (against information diffusion truncation more in general).

Directly training spiking neural networks for cyber-physical systems: from supervised to reinforcement learning

Spiking Neural Networks (SNNs) are bio-inspired Artificial Neural Networks (ANNs) utilizing discrete spiking signals, akin to neuron communication in the brain, making them ideal for real-time and energy-efficient Cyber-Physical Systems (CPSs). This thesis explores their potential in Structural Health Monitoring (SHM), leveraging low-cost MEMS accelerometers for early damage detection in motorway bridges. The study focuses on Long Short-Term SNNs (LSNNs), although their complex learning processes pose challenges. Comparing LSNNs with other ANN models and training algorithms for SHM, findings indicate LSNNs' effectiveness in damage identification, comparable to ANNs trained using traditional methods. Additionally, an optimized embedded LSNN implementation demonstrates a 54% reduction in execution time, but with longer pre-processing due to spike-based encoding. Furthermore, SNNs are applied in UAV obstacle avoidance, trained directly using a Reinforcement Learning (RL) algorithm with event-based input from a Dynamic Vision Sensor (DVS). Performance evaluation against Convolutional Neural Networks (CNNs) highlights SNNs' superior energy efficiency, showing a 6x decrease in energy consumption. The study also investigates embedded SNN implementations' latency and throughput in real-world deployments, emphasizing their potential for energy-efficient monitoring systems. This research contributes to advancing SHM and UAV obstacle avoidance through SNNs' efficient information processing and decision-making capabilities within CPS domains.

Design of wireless sensor networks for fluid dynamic applications

In fluid dynamics research, pressure measurements are of great importance to define the flow field acting on aerodynamic surfaces. In fact the experimental approach is fundamental to avoid the complexity of the mathematical models for predicting the fluid phenomena. It’s important to note that, using in-situ sensor to monitor pressure on large domains with highly unsteady flows, several problems are encountered working with the classical techniques due to the transducer cost, the intrusiveness, the time response and the operating range. An interesting approach for satisfying the previously reported sensor requirements is to implement a sensor network capable of acquiring pressure data on aerodynamic surface using a wireless communication system able to collect the pressure data with the lowest environmental–invasion level possible. In this thesis a wireless sensor network for fluid fields pressure has been designed, built and tested. To develop the system, a capacitive pressure sensor, based on polymeric membrane, and read out circuitry, based on microcontroller, have been designed, built and tested. The wireless communication has been performed using the Zensys Z-WAVE platform, and network and data management have been implemented. Finally, the full embedded system with antenna has been created. As a proof of concept, the monitoring of pressure on the top of the mainsail in a sailboat has been chosen as working example.

Wireless Sensor Networks for Monitoring Applications

Wireless Sensor Networks (WSNs) are getting wide-spread attention since they became easily accessible with their low costs. One of the key elements of WSNs is distributed sensing. When the precise location of a signal of interest is unknown across the monitored region, distributing many sensors randomly/uniformly may yield with a better representation of the monitored random process than a traditional sensor deployment. In a typical WSN application the data sensed by nodes is usually sent to one (or more) central device, denoted as sink, which collects the information and can either act as a gateway towards other networks (e.g. Internet), where data can be stored, or be processed in order to command the actuators to perform special tasks. In such a scenario, a dense sensor deployment may create bottlenecks when many nodes competing to access the channel. Even though there are mitigation methods on the channel access, concurrent (parallel) transmissions may occur. In this study, always on the scope of monitoring applications, the involved development progress of two industrial projects with dense sensor deployments (eDIANA Project funded by European Commission and Centrale Adritica Project funded by Coop Italy) and the measurement results coming from several different test-beds evoked the necessity of a mathematical analysis on concurrent transmissions. To the best of our knowledge, in the literature there is no mathematical analysis of concurrent transmission in 2.4 GHz PHY of IEEE 802.15.4. In the thesis, experience stories of eDIANA and Centrale Adriatica Projects and a mathematical analysis of concurrent transmissions starting from O-QPSK chip demodulation to the packet reception rate with several different types of theoretical demodulators, are presented. There is a very good agreement between the measurements so far in the literature and the mathematical analysis.

Thermal and air quality modeling of an electric vehicle cabin with low-cost sensors and reinforcement learning

The design process of any electric vehicle system has to be oriented towards the best energy efficiency, together with the constraint of maintaining comfort in the vehicle cabin. Main aim of this study is to research the best thermal management solution in terms of HVAC efficiency without compromising occupant’s comfort and internal air quality. An Arduino controlled Low Cost System of Sensors was developed and compared against reference instrumentation (average R-squared of 0.92) and then used to characterise the vehicle cabin in real parking and driving conditions trials. Data on the energy use of the HVAC was retrieved from the car On-Board Diagnostic port. Energy savings using recirculation can reach 30 %, but pollutants concentration in the cabin builds up in this operating mode. Moreover, the temperature profile appeared strongly nonuniform with air temperature differences up to 10° C. Optimisation methods often require a high number of runs to find the optimal configuration of the system. Fast models proved to be beneficial for these task, while CFD-1D model are usually slower despite the higher level of detail provided. In this work, the collected dataset was used to train a fast ML model of both cabin and HVAC using linear regression. Average scaled RMSE over all trials is 0.4 %, while computation time is 0.0077 ms for each second of simulated time on a laptop computer. Finally, a reinforcement learning environment was built in OpenAI and Stable-Baselines3 using the built-in Proximal Policy Optimisation algorithm to update the policy and seek for the best compromise between comfort, air quality and energy reward terms. The learning curves show an oscillating behaviour overall, with only 2 experiments behaving as expected even if too slow. This result leaves large room for improvement, ranging from the reward function engineering to the expansion of the ML model.

Unsupervised reinforcement learning via state entropy maximization

Reinforcement Learning (RL) provides a powerful framework to address sequential decision-making problems in which the transition dynamics is unknown or too complex to be represented. The RL approach is based on speculating what is the best decision to make given sample estimates obtained from previous interactions, a recipe that led to several breakthroughs in various domains, ranging from game playing to robotics. Despite their success, current RL methods hardly generalize from one task to another, and achieving the kind of generalization obtained through unsupervised pre-training in non-sequential problems seems unthinkable. Unsupervised RL has recently emerged as a way to improve generalization of RL methods. Just as its non-sequential counterpart, the unsupervised RL framework comprises two phases: An unsupervised pre-training phase, in which the agent interacts with the environment without external feedback, and a supervised fine-tuning phase, in which the agent aims to efficiently solve a task in the same environment by exploiting the knowledge acquired during pre-training. In this thesis, we study unsupervised RL via state entropy maximization, in which the agent makes use of the unsupervised interactions to pre-train a policy that maximizes the entropy of its induced state distribution. First, we provide a theoretical characterization of the learning problem by considering a convex RL formulation that subsumes state entropy maximization. Our analysis shows that maximizing the state entropy in finite trials is inherently harder than RL. Then, we study the state entropy maximization problem from an optimization perspective. Especially, we show that the primal formulation of the corresponding optimization problem can be (approximately) addressed through tractable linear programs. Finally, we provide the first practical methodologies for state entropy maximization in complex domains, both when the pre-training takes place in a single environment as well as multiple environments.

Concepts and methods for efficient decentralized learning in federated settings

Deep Neural Networks (DNNs) have revolutionized a wide range of applications beyond traditional machine learning and artificial intelligence fields, e.g., computer vision, healthcare, natural language processing and others. At the same time, edge devices have become central in our society, generating an unprecedented amount of data which could be used to train data-hungry models such as DNNs. However, the potentially sensitive or confidential nature of gathered data poses privacy concerns when storing and processing them in centralized locations. To this purpose, decentralized learning decouples model training from the need of directly accessing raw data, by alternating on-device training and periodic communications. The ability of distilling knowledge from decentralized data, however, comes at the cost of facing more challenging learning settings, such as coping with heterogeneous hardware and network connectivity, statistical diversity of data, and ensuring verifiable privacy guarantees. This Thesis proposes an extensive overview of decentralized learning literature, including a novel taxonomy and a detailed description of the most relevant system-level contributions in the related literature for privacy, communication efficiency, data and system heterogeneity, and poisoning defense. Next, this Thesis presents the design of an original solution to tackle communication efficiency and system heterogeneity, and empirically evaluates it on federated settings. For communication efficiency, an original method, specifically designed for Convolutional Neural Networks, is also described and evaluated against the state-of-the-art. Furthermore, this Thesis provides an in-depth review of recently proposed methods to tackle the performance degradation introduced by data heterogeneity, followed by empirical evaluations on challenging data distributions, highlighting strengths and possible weaknesses of the considered solutions. Finally, this Thesis presents a novel perspective on the usage of Knowledge Distillation as a mean for optimizing decentralized learning systems in settings characterized by data heterogeneity or system heterogeneity. Our vision on relevant future research directions close the manuscript.

The APSEL4D Monolithic Active Pixel Sensor and its Usage in a Single Electron Interference Experiment

We have realized a Data Acquisition chain for the use and characterization of APSEL4D, a 32 x 128 Monolithic Active Pixel Sensor, developed as a prototype for frontier experiments in high energy particle physics. In particular a transition board was realized for the conversion between the chip and the FPGA voltage levels and for the signal quality enhancing. A Xilinx Spartan-3 FPGA was used for real time data processing, for the chip control and the communication with a Personal Computer through a 2.0 USB port. For this purpose a firmware code, developed in VHDL language, was written. Finally a Graphical User Interface for the online system monitoring, hit display and chip control, based on windows and widgets, was realized developing a C++ code and using Qt and Qwt dedicated libraries. APSEL4D and the full acquisition chain were characterized for the first time with the electron beam of the transmission electron microscope and with 55Fe and 90Sr radioactive sources. In addition, a beam test was performed at the T9 station of the CERN PS, where hadrons of momentum of 12 GeV/c are available. The very high time resolution of APSEL4D (up to 2.5 Mfps, but used at 6 kfps) was fundamental in realizing a single electron Young experiment using nanometric double slits obtained by a FIB technique. On high statistical samples, it was possible to observe the interference and diffractions of single isolated electrons traveling inside a transmission electron microscope. For the first time, the information on the distribution of the arrival time of the single electrons has been extracted.

Preharvest factors affecting quality and antioxidant levels on Abate Fétel pears: study of superficial scald with a multivariate statistical approach and neural networks

There are only a few insights concerning the influence that agronomic and management variability may have on superficial scald (SS) in pears. Abate Fétel pears were picked during three seasons (2018, 2019 and 2020) from thirty commercial orchards in the Emilia Romagna region, Italy. Using a multivariate statistical approach, high heterogeneity between farms for SS development after cold storage with regular atmosphere was demonstrated. Indeed, some factors seem to affect SS in all growing seasons: high yields, soil texture, improper irrigation and Nitrogen management, use of plant growth regulators, late harvest, precipitations, Calcium and cow manure, presence of nets, orchard age, training system and rootstock. Afterwards, we explored the spatio/temporal variability of fruit attributes in two pear orchards. Environmental and physiological spatial variables were recorded by a portable RTK GPS. High spatial variability of the SS index was observed. Through a geostatistical approach, some characteristics, including soil electrical conductivity and fruit size, have been shown to be negatively correlated with SS. Moreover, regression tree analyses were applied suggesting the presence of threshold values of antioxidant capacity, total phenolic content, and acidity against SS. High pulp firmness and IAD values before storage, denoting a more immature fruit, appeared to be correlated with low SS. Finally, a convolution neural networks (CNN) was tested to detect SS and the starch pattern index (SPI) in pears for portable device applications. Preliminary statistics showed that the model for SS had low accuracy but good precision, and the CNN for SPI denoted good performances compared to the Ctifl and Laimburg scales. The major conclusion is that Abate Fétel pears can potentially be stored in different cold rooms, according to their origin and quality features, ensuring the best fruit quality for the final consumers. These results might lead to a substantial improvement in the Italian pear industry.

Design and testing of innovative, artificial intelligence-based techniques to model and control the combustion process and to reduce the Co2 emissions in modern, high-performance, gasoline direct injection engines

This work deals with the development of calibration procedures and control systems to improve the performance and efficiency of modern spark ignition turbocharged engines. The algorithms developed are used to optimize and manage the spark advance and the air-to-fuel ratio to control the knock and the exhaust gas temperature at the turbine inlet. The described work falls within the activity that the research group started in the previous years with the industrial partner Ferrari S.p.a. . The first chapter deals with the development of a control-oriented engine simulator based on a neural network approach, with which the main combustion indexes can be simulated. The second chapter deals with the development of a procedure to calibrate offline the spark advance and the air-to-fuel ratio to run the engine under knock-limited conditions and with the maximum admissible exhaust gas temperature at the turbine inlet. This procedure is then converted into a model-based control system and validated with a Software in the Loop approach using the engine simulator developed in the first chapter. Finally, it is implemented in a rapid control prototyping hardware to manage the combustion in steady-state and transient operating conditions at the test bench. The third chapter deals with the study of an innovative and cheap sensor for the in-cylinder pressure measurement, which is a piezoelectric washer that can be installed between the spark plug and the engine head. The signal generated by this kind of sensor is studied, developing a specific algorithm to adjust the value of the knock index in real-time. Finally, with the engine simulator developed in the first chapter, it is demonstrated that the innovative sensor can be coupled with the control system described in the second chapter and that the performance obtained could be the same reachable with the standard in-cylinder pressure sensors.

How to explain: from theory to practice

Today we live in an age where the internet and artificial intelligence allow us to search for information through impressive amounts of data, opening up revolutionary new ways to make sense of reality and understand our world. However, it is still an area of improvement to exploit the full potential of large amounts of explainable information by distilling it automatically in an intuitive and user-centred explanation. For instance, different people (or artificial agents) may search for and request different types of information in a different order, so it is unlikely that a short explanation can suffice for all needs in the most generic case. Moreover, dumping a large portion of explainable information in a one-size-fits-all representation may also be sub-optimal, as the needed information may be scarce and dispersed across hundreds of pages. The aim of this work is to investigate how to automatically generate (user-centred) explanations from heterogeneous and large collections of data, with a focus on the concept of explanation in a broad sense, as a critical artefact for intelligence, regardless of whether it is human or robotic. Our approach builds on and extends Achinstein’s philosophical theory of explanations, where explaining is an illocutionary (i.e., broad but relevant) act of usefully answering questions. Specifically, we provide the theoretical foundations of Explanatory Artificial Intelligence (YAI), formally defining a user-centred explanatory tool and the space of all possible explanations, or explanatory space, generated by it. We present empirical results in support of our theory, showcasing the implementation of YAI tools and strategies for assessing explainability. To justify and evaluate the proposed theories and models, we considered case studies at the intersection of artificial intelligence and law, particularly European legislation. Our tools helped produce better explanations of software documentation and legal texts for humans and complex regulations for reinforcement learning agents.

Bioactive and toxic compounds in different foodstuffs: analytical techniques and sensory analysis

Foods that provide medical and health benefits or have a role in disease risk prevention are termed functional foods. The functionality of functional foods is derived from bioactive compounds that are extranutritional constituents present in small quantities in food. Bioactive components include a range of chemical compounds with varying structures such as carotenoids, flavonoids, plant sterols, omega-3 fatty acids (n-3), allyl and diallyl sulfides, indoles (benzopyrroles), and phenolic acids. The increasing consumer interest in natural bioactive compounds has brought about a rise in demand for these kinds of compounds and, in parallel, an increasing number of scientific studies have this type of substance as main topic. The principal aim of this PhD research project was the study of different bioactive and toxic compounds in several natural matrices. To achieve this goal, chromatographic, spectroscopic and sensorial analysis were performed. This manuscript reports the main results obtained in the six activities briefly summarized as follows: • SECTION I: the influence of conventional packaging on lipid oxidation of pasta was evaluated in egg spaghetti. • SECTION II: the effect of the storage at different temperatures of virgin olive oil was monitored by peroxide value, fatty acid activity, OSI test and sensory analysis. • SECTION III: the glucosinolate and phenolic content of 37 rocket salad accessions were evaluated, comparing Eruca sativa and Diplotaxis tenuifolia species. Sensory analysis and the influence of the phenolic and glucosinolate composition on sensory attributes of rocket salads has been also studied. • SECTION IV: ten buckwheat honeys were characterised on the basis of their pollen, physicochemical, phenolic and volatile composition. • SECTION V: the polyphenolic fraction, anthocyanins and other polar compounds, the antioxidant capacity and the anty-hyperlipemic action of the aqueous extract of Hibiscus sabdariffa were achieved. • SECTION VI: the optimization of a normal phase high pressure liquid chromatography–fluorescence detection method for the quantitation of flavanols and procyanidins in cocoa powder and chocolate samples was performed.