Papers by Andrea Bonarini
The Design Journal, 2017
Present paper analyzes and compares two innovation models in robot design. The first part describ... more Present paper analyzes and compares two innovation models in robot design. The first part describes the incremental development process of Teo: a robotic tool for Autistic Spectrum Disorder (ASD) treatment for children. The second part describes the radical innovation model used for Riby: a robotic tool for ASD treatment for adults. At the end of this paper, authors discuss the results of each case and expose the conclusions as a possible new method of product development in social robotics.
Cornell University - arXiv, Feb 22, 2021
Faced with an ever-increasing complexity of their domains of application, artificial learning age... more Faced with an ever-increasing complexity of their domains of application, artificial learning agents are now able to scale up in their ability to process an overwhelming amount of data; However this comes at a cost of encoding and processing an increasing amount of redundant information. This work exploits the properties of learning systems, applied in partially observable domains, defined to selectively focus on the specific type of information that is more likely to express the causal interaction among the transitioning states of the environment. Experiments performed under a total of 32 different Atari game environments show that adaptive masking of the observation space based on the temporal difference displacement criterion enabled a significant improvement in convergence of temporal difference algorithms applied to partially observable Markov processes under identical reproducible settings.
Journal on Educational Technology, 2009
Robotics competitions have been introduced since the 80s in the community 'science, in order ... more Robotics competitions have been introduced since the 80s in the community 'science, in order to compare the results obtained by different researchers on common ground and shared. In races and 'required that robots play activities' as defined by the rules of the race and measure the quality' of performance in an objective and / or shared. Among the interesting aspects of this type of scientific debate we want to deliver us out some, also relevant to the races used for educational purposes
We study the benefit of sharing representations among tasks to enable the effective use of deep n... more We study the benefit of sharing representations among tasks to enable the effective use of deep neural networks in Multi-Task Reinforcement Learning. We leverage the assumption that learning from different tasks, sharing common properties, is helpful to generalize the knowledge of them resulting in a more effective feature extraction compared to learning a single task. Intuitively, the resulting set of features offers performance benefits when used by Reinforcement Learning algorithms. We prove this by providing theoretical guarantees that highlight the conditions for which is convenient to share representations among tasks, extending the well-known finite-time bounds of Approximate Value-Iteration to the multi-task setting. In addition, we complement our analysis by proposing multi-task extensions of three Reinforcement Learning algorithms that we empirically evaluate on widely used Reinforcement Learning benchmarks showing significant improvements over the single-task counterparts...
Stiamo assistendo all'invasione silenziosa dei robot nella vita di tutti i giorni. Dall'originale... more Stiamo assistendo all'invasione silenziosa dei robot nella vita di tutti i giorni. Dall'originale impiego in fabbrica, i robot vedono ormai una crescita esponenziale al di fuori dell'ambito della produzione industriale. È importante quindi capire come fa un robot autonomo a svolgere il proprio compito, quali sono le possibilità reali di applicazione in questo momento, e quali potranno essere in un prossimo futuro. Presentiamo una rapida rassegna dello stato della robotica ad oggi, e degli sviluppi in un futuro prossimo. Infine, mostriamo alcune implicazioni sociali ed etiche che stanno emergendo di conseguenza.
Journal on Educational Technology, 2009
INTRODUZIONE "Giochiamo con i robot" è un laboratorio interattivo per grandi e piccini realizzato... more INTRODUZIONE "Giochiamo con i robot" è un laboratorio interattivo per grandi e piccini realizzato per l'edizione 2007 del Festival della Scienza di Genova. Lungo un percorso che va dalla telerobotica alla robotica evolutiva, il laboratorio sviluppa il tema di dare intelligenza ai robot. Questo percorso, le cui tappe sono le varie installazioni, si conclude nella "bottega" dove è possibile costruire e programmare i propri robot o smontare e modificare quelli esposti durante il percorso didattico. I visitatori sono coinvolti in attività ludiche grazie alle quali possono entrare in contatto con alcune delle idee potenti della robotica, come "feedback" e "comportamenti emergenti". Chi viene al laboratorio è accolto all'ingresso da un "robot maggiordomo" che accompagna i visitatori alle varie installazioni fornendo una prima introduzione (figura 1). Le installazioni comprendono: • i fotovori, robot autonomi che si nutrono di luce grazie ad una cella solare; • SwarmBot, robot che si muovono su una piattaforma con 100 lampadine che pos-38 TD47 numero 2-2009 Giochiamo con i Robot Insegnare a un robot il gioco del calcio? Governare una colonia di automi? Questione di programmazione.
Perspectives and research on play for children with disabilities, 2020
For creating play opportunities for children with disabilities toys, games, apps, robots, and oth... more For creating play opportunities for children with disabilities toys, games, apps, robots, and other technological products are as important as for typically developing children. Above all the products have great potential for inclusive play. However, many anecdotes from clinical practice and data from research show the challenges in finding and choosing a suitable toy or technology, in evaluating these play objects on their usability and accessibility for given children, in designing and producing a toy usable for all children. This paper describes the scoping review carried out to investigate: (1) which guidelines and tools regarding usability and accessibility of toys and technologies for play for children with disabilities exist, (2) what is their possible use for different stakeholders involved in play for children with disabilities, (3) what are the strengths and the weaknesses of the guidelines and tools. For this review, sources identified by experts, different databases, and handmade search results were considered, which yielded to a final set of 15 guidelines on usability and accessibility of toys and technologies for play for children with disabilities that was explored in detail. Each guideline was reviewed by two reviewers using the adapted AGREE II instrument. The review resulted in the selection of 10 guidelines on usability or accessibility of toys and technologies, only 5 had a specific focus on play. For most of the guidelines the rigour of the development and the supporting evidence were not described. Further research and development is needed, as adults involved in play for children with disabilities need support in handling or creating the appropriate toys and technologies. Summary of achievements Achieved effects References to the intervention or research project List of published materials referring to the specific entry of the database
2017 Joint IEEE International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob), Sep 1, 2017
The use of biologically realistic (brain-like) control systems in autonomous robots offers two po... more The use of biologically realistic (brain-like) control systems in autonomous robots offers two potential benefits. For neuroscience, it may provide important insights into normal and abnormal control and decision-making in the brain, by testing whether the computational learning and decision rules proposed on the basis of simple laboratory experiments lead to effective and coherent behaviour in complex environments. For robotics, it may offer new insights into control system designs, for example in the context of threat avoidance and self-preservation. In the brain, learning and decision-making for rewards and punishments (such as pain) are thought to involve integrated systems for innate (Pavlovian) responding, habit-based learning, and goal-directed learning, and these systems have been shown to be well-described by RL models. Here, we simulated this 3-system control hierarchy (in which the innate system is derived from an evolutionary learning model), and show that it reliably achieves successful performance in a dynamic predatoravoidance task. Furthermore, we show situations in which a 3system architecture provides clear advantages over single or dual system architectures. Finally, we show that simulating a computational model of obsessive compulsive disorder, an example of a disease thought to involve a specific deficit in the integration of habit-based and goal-directed systems, can reproduce the results of human clinical experiments. The results illustrate how robotics can provide a valuable platform to test the validity and utility of computational models of human behaviour, in both health and disease. They also illustrate how bio-inspired control systems might usefully inform selfpreservative behaviour in autonomous robots, both in normal and malfunctioning situations.
Proceedings of the Companion of the 2017 ACM/IEEE International Conference on Human-Robot Interaction, 2017
Emotions are considered by many researchers as beneficial in social robotics, since they can enri... more Emotions are considered by many researchers as beneficial in social robotics, since they can enrich human-robot interaction with non-verbal clues. Although there have been works that have studied emotion expression in robotics, the mechanisms created to project emotion are usually highly integrated in each solution. This limits the possibility to develop a general approach. This paper presents a system that has been initially created for a theatrical robot to enrich its actions with emotions, but it has been designed to be adaptable to other fields. The emotional enrichment system has been envisioned to be used with any action decision system.
Adaptive Behavior, 2016
In many robotic applications, a robot body should have a functional shape that cannot include bio... more In many robotic applications, a robot body should have a functional shape that cannot include bio-inspired elements, but it would still be important that the robot can express emotions, moods, or a character, to make it acceptable, and to involve its users. Dynamic signals from movement can be exploited to provide this expression while the robot is acting to perform its task. A research effort has been started to find general emotion expression models for actions that could be applied to any kind of robot to obtain believable and easily detectable emotional expressions. The need for a unified representation of emotional expression emerged. A framework to define action characteristics that could be used to represent emotions is proposed in this paper. Guidelines are provided to identify quantitative models and numerical values for parameters, which can be used to design and engineer emotional robot actions. A set of robots having different shapes, movement possibilities, and goals ha...
IEEE Transactions on Education, 2013
The authors designed and ran a crash course on emotional robotics involving students from both th... more The authors designed and ran a crash course on emotional robotics involving students from both the Information Engineering School and the Design School of Politecnico di Milano. The course consisted of two intensive days of short introductory lessons and lab activity, done in interdisciplinary groups and supported by a well-equipped prototyping and modeling lab. People from very different backgrounds had to work efficiently together, going from problem setting through the demonstration of the physical implementation of an object able to show four different emotional states. Both teacher evaluation, and questionnaire-based feedback from the students, show that it was successful and useful to set up this type of intensive experience in which students share their abilities to achieve a common goal. Key aspects for the success of the course were the short time the students had to reach a well-defined, yet general, goal, the students' ability to find efficient ways of cooperating and sharing their competences, students' motivation to arrive at a working prototype, and the strong support from teachers and lab personnel.
Playing a game that requires physical interaction with people is a challenging human-robot intera... more Playing a game that requires physical interaction with people is a challenging human-robot interaction application. Besides the usual abilities expected by autonomous robots (such as: localization, planning, navigation), here the robot should behave in a safe and engaging way, possibly showing to be purposeful and fair, so to make the human player comfortably involved. Among the relevant issues we have faced in this area, we present in this paper how the robot could model the specific player and adapt its behavior to his/her perceived ability, with the aim of obtaining an even game that optimizes the involvement of the player in a flow situation. We implemented our approach in a physically interactive robotic game, and tested it with 52 subjects. The adapting robot was appreciated more than the non-adaptive one.
ArXiv, 2020
MushroomRL is an open-source Python library developed to simplify the process of implementing and... more MushroomRL is an open-source Python library developed to simplify the process of implementing and running Reinforcement Learning (RL) experiments. Compared to other available libraries, MushroomRL has been created with the purpose of providing a comprehensive and flexible framework to minimize the effort in implementing and testing novel RL methodologies. Indeed, the architecture of MushroomRL is built in such a way that every component of an RL problem is already provided, and most of the time users can only focus on the implementation of their own algorithms and experiments. The result is a library from which RL researchers can significantly benefit in the critical phase of the empirical analysis of their works. MushroomRL stable code, tutorials and documentation can be found at this https URL.
The fusion of multisensorial data is a common practice to identify a world model when data coming... more The fusion of multisensorial data is a common practice to identify a world model when data coming from a single sensor is unreliable. A possible approach to multisensor data fusion consists in identifying common interpretations of data coming from different sensors, and to integrate the information represented as symbols related to the possible interpretations. In this paper, we present an architecture of integrated instances of Fuzzy ARTMAP, a fuzzy neural network system, able to learn to classify numerical input vectors into fuzzy classes, i.e., to build symbolic interpretations. We have successfully applied this architecture on a navigation task in an unstructured, unknown environment for a mobile robot, integrating data from sensors, omnatidia, bumpers, and the odometric system.
We present SMILe (Self-Motivated Incremental Learning), a new learning framework where an agent l... more We present SMILe (Self-Motivated Incremental Learning), a new learning framework where an agent learns a set of abilities needed to face several tasks in its environm ent, by following a biologically inspired, self–motivated approach that loops over three main phases. In thebabbling phase, the agent randomly explores the environment, in a way similar to what animal puppies do. This provides information about the effects of action on the environment. In themotivating phase, the agent identifies what is interesting in the environment and develo ps an intrinsic motivation in achieving situations with highest interest. In the skill acquisition phase, the agent learns the skills needed to reach the most interesting state, guided by a selfgenerated reinforcement function. Once a new skill is avail able the babbling phase can start again with the enlarged set of abilities, and learning continues all the life long. We pres ent results on a gridworld abstraction of a robotic environment to s...
The implementation of behaviors for embodied autonomous agents by means of Fuzzy Logic Controller... more The implementation of behaviors for embodied autonomous agents by means of Fuzzy Logic Controllers (FLC) has natural and engineering motivations. Fuzzy logic is recognized as a powerful mean to represent approximation intrinsic in human (and animal) reasoning and reacting. On the other side, fuzzy logic shows flexibility and robustness, important in the implementation of artificial devices. Two aspects of the development of autonomous agents may be faced by learning FLCs: the adaptation of the agent to the environment, and the reduction of the design time and efforts. In this paper, we present issues related to learn behaviors implemented as FLCs, and we propose our approach implemented in ELF (Evolutionary Learning for Fuzzy rules). We are using ELF to support the development of different types of agents. Finally, we present the results that we have obtained both in simulated and real environments.
... Modeling issues in multimedia car-driver interaction. Author: Andrea Bonarini, Published in: ... more ... Modeling issues in multimedia car-driver interaction. Author: Andrea Bonarini, Published in: · Book. Intelligent multimedia interfaces. American Association for Artificial Intelligence Menlo Park, CA, USA ©1993 table of contents ISBN:0-262-63150-4. 1993 Article. Bibliometrics. ...
International Journal of Social Robotics
Play is a common activity, providing not only pleasure but also physical and cognitive developmen... more Play is a common activity, providing not only pleasure but also physical and cognitive development. In the quest for new playing experiences, there is an increasing tendency to develop robots playing with people. Making believable playing robots able to keep human players engaged and satisfied by the playing experience is the main challenge. In this work, we investigate the possibilities of a playful interaction between a human player and a mobile robot. In particular, this paper focuses on the applicability of deception as a means to support engagement and the attribution of rationality to playing robotic agents. By analyzing the interaction situation between the human and robot players, by identifying the need for deception, and by deciding whether and how to deceive, we aim at increasing self-reported engagement and fun, which are also related to the perception of the robotic opponent as smart enough to compete at an appropriate level. Experiments were conducted on a sample of 78 subjects facing two different deceptive behaviors and a basic behavior without any deception. All participants responded to a post-interaction questionnaire from which it was possible to observe a positive acceptance of the perception of the robot as a rational agent aimed at winning. In general, deception was perceived by most of the players as one of the robot's abilities, when actuated, and contributed to the reported fun.
Uploads
Papers by Andrea Bonarini