1

PhD Program in Bioengineering and Robotics

Curriculum Robotics and Autonomous Systems

Research themes

1. SENSOR BASED CONTROL OF ROBOTS FOR HUMAN-ROBOT COOPERATIVE OPERATIONS ......... 3 2. TACTILE SENSORS FOR ROBOT HANDS ......................................................................................... 5 3. MACHINE LEARNING MODELS FOR ROBOT TACTILE SENSING AND PERCEPTION ....................... 6 4. ROBOT MANIPULATION OF SOFT AND FLEXIBLE OBJECTS ........................................................... 8 5. CULTURE-AWARE ROBOTS AND ENVIRONMENTAL SENSOR SYSTEMS FOR ELDERLY SUPPORT 9 6. AUTONOMOUS DRONES AND UAVS FOR SEARCH & RESCUE .....................................................11 7. ADVANCED ROBOT MANIPULATION SKILLS ACQUIRED VIA HUMAN DEMONSTRATIONS .........13 8. MODELS FOR HUMAN-ROBOT COLLABORATION LEVERAGING HUMAN BIOMETRIC AND

PHYSIOLOGICAL DATA OBTAINED VIA WEARABLE SENSING SUITS ....................................................15 9. CONTROL AND PLANNING ARCHITECTURES FOR HUMAN ROBOT COLLABORATIVE TASKS .....17

The main goal of the Robotics and Autonomous Systems curriculum is to study, design and build robots, team of robots and, in general, autonomous systems able to exhibit a robust and predictable behavior while performing complex tasks in challenging indoor and outdoor environments. The focus is both on key methodologies and technologies (e.g., advanced robot control, robot coordination and cooperation, sensing, state estimation, knowledge representation, motion planning, real-time scheduling, human-robot interaction, design of macro/micro robot systems, design of sensors and actuators), as well as on specific robotic areas (e.g., aerial and space robotics, wheeled and legged robots, Industry 4.0) and on different application scenarios (e.g., search & rescue, surveillance and monitoring, material handling and transportation).

Furthermore, all the aspects outlined above are dealt with by focusing on the study and the adoption of theoretically sound methodologies and the design of experimentally verifiable solutions, with the goals of meeting robustness and predictability requirements even in unknown, dynamically changing, or even hazardous environments.

The themes offered in this intake are supported by the Department of Informatics, Bioengineering, Robotics and System Engineering, University of Genoa.

2

The ideal candidates are students with a higher-level university degree, with a strong desire for designing and developing robot systems able to have a huge impact on the society in the upcoming future. International applications are encouraged and will receive logistic support with visa issues, relocation, etc.

3

1. Sensor based control of robots for human-robot cooperative operations Tutor: Giorgio Cannata Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Innovative robot systems are expected to work and cooperate with other robots and humans for the execution of handling and manipulation tasks for industrial and service applications. To this aim highly sensorised robots are required to execute collaborative tasks safely side by side with humans and are expected to properly react to unexpected events as well as to estimate the human intentions or the changes within their working envelope to ensure the reliable and safe (in case of direct interaction with humans) execution of the planned tasks. Highly sensorised robot systems, or robot working cells (depending on context and application), typically feature multiple camera sensors, range sensors, tactile and force/torque sensors, and possibly task specific sensing devices. This doctorate project is part of the European Project Collaborate (https://collaborate-project.eu). The goal is to study robot control methods based on multisensory feedback (joint visuo-tactile feedback) for collaborative human-robot assembly operations for car assembly. These research topics have been part of relevant international research projects (Collaborate: https://collaborate-project.eu; CloPeMa: www.clopema.eu). Requirements: Applicants are expected to have good knowledge of: robot control, robot programming, mechatronics. Furthermore, good attitude for experimental work is mandatory. The candidates must have: very good programming skills with different languages (including C/C++, Python, Matlab/Simulink); confidence with electronic hardware and be capable to conduct experiments; attitude to problem solving, and be strongly motivated for team working. References:

G. Cannata, S. Denei, F. Mastrogiovanni. Towards the creation of tactile maps for robots and their use in robot contact motion control. Robotics and Autonomous Systems 63(3): 293–308, 2015.

S. Denei, F. Mastrogiovanni, G. Cannata. Parallel Force-Position Control Mediated by Tactile Maps for Robot Contact Tasks. Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2012), Vilamoura, Portugal, October 2012.

A. Del Prete, S. Denei, L. Natale, F. Mastrogiovanni, F. Nori, G. Cannata, G. Metta. Skin spatial calibration using force/torque measurements.

4

Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2012), Vilamoura, Portugal, October 2012.

Contacts: Email: giorgio.cannata@unige.it

5

2. Tactile Sensors for robot hands Tutor: Giorgio Cannata Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Tactile sensing is the key element for the development of robot hands capable to perform handling and manipulation of objects as well as to perform complex interaction operations with humans (e.g., objects handover). Tactile sensors are embedded transduction and sensing device which must fit on human sized gripper or multi-fingered robot hands. The goal of this Doctorate project is to design a multimodal tactile sensor to be installed on a robot hand. The challenges are related to the study of mechatronic design solutions for the implementation of arrays of tactile sensors over curved surfaces, with embedded electronics. Design of tactile sensors for perception and control are challenging research topics and University of Genova has gained a solid reputation in this area in the past few years, and has been involved in important international research projects related to this topic (ROBOSKIN: www.roboskin.eu; CloPeMa: www.clopema.eu). Requirements: Applicants are expected to have strong motivation to study and design mechatronic systems. The candidates must have very good skills in at least one of the following topics: mechatronics, robotics, electronics. The candidates must have also a very good attitude for laboratory practice and must be capable to conduct experiments. Programming skills for firmware development are considered a major plus. Finally, the candidates must prove capability of team working. References:

G. Cannata, S. Denei, F. Mastrogiovanni. Towards the creation of tactile maps for robots and their use in robot contact motion control. Robotics and Autonomous Systems 63(3): 293–308, 2015.

P. Maiolino, M. Maggiali, G. Cannata, G. Metta, L. Natale. A Flexible and robust large scale capacitive tactile system for robots. IEEE Sensors Journal, vol. 13, no. 10, pp. 3910-3917, October 2013.

L. Muscari, F. Mastrogiovanni, L. Seminara, M. Capurro, G. Cannata, M Valle. Real-time reconstruction of contact shapes for large area robot skin. Proceedings of the 2013 IEEE International Conference on Robotics and Automation (ICRA 2013), Karlsruhe, Germany, May 2013.

Contacts: Email: giorgio.cannata@unige.it

6

3. Machine learning models for robot tactile sensing and perception Tutor: Giorgio Cannata Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Tactile sensing is one key topic for the development of future robots capable of complex interaction with humans and objects. Design of tactile sensors as well as tactile based sensing, perception and control are challenging research topics and University of Genova has gained a solid reputation in this area in the past few years, and has been involved in important international research projects related to this topic (ROBOSKIN: www.roboskin.eu; CloPeMa: www.clopema.eu). Tactile data convey information about the characteristics of the contact between the robot and the manipulated objects. This information is fundamental to proper control the grasping and manipulation; furthermore, controlled manipulation can in turn be used to extract further information useful to recognize and classify objects and contacts. The objective of the PhD program is to investigate techniques and methodologies for tactile data sensing, perception and interpretation. Furthermore, since tactile data are generated by active manipulation control another objective is the study of the mechanisms relating tactile based feedback robot control and tactile data perception. Requirements: Applicants are expected to have strong background and experience in at least one of the following topics: robot control, machine learning, system identification. The candidates must have: very good programming skills with different languages (including C/C++, Matlab/Simulink); experience with electronic hardware; be capable to conduct experiments and be strongly motivated for team working. References:

G. Cannata, S. Denei, F. Mastrogiovanni. Towards the creation of tactile maps for robots and their use in robot contact motion control. Robotics and Autonomous Systems 63(3): 293–308, 2015.

P. Maiolino, M. Maggiali, G. Cannata, G. Metta, L. Natale. A Flexible and robust large scale capacitive tactile system for robots. IEEE Sensors Journal, vol. 13, no. 10, pp. 3910-3917, October 2013.

L. Muscari, F. Mastrogiovanni, L. Seminara, M. Capurro, G. Cannata, M Valle. Real-time reconstruction of contact shapes for large area robot skin. Proceedings of the 2013 IEEE International Conference on Robotics and Automation (ICRA 2013), Karlsruhe, Germany, May 2013.

7

Contacts: Email: giorgio.cannata@unige.it

8

4. Robot manipulation of soft and flexible objects Tutor: Giorgio Cannata Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Highly sensorised robots are expected to execute highly coordinated motions to handle and manipulate soft and flexible objects (e.g. cables, fabrics etc.) autonomously or in cooperation with other agents (either robots or humans). These operations require a modelling of the manipulated objects along with sensors based (vision and touch) estimation algorithms to predict their deformation and to generate appropriate control strategies. Examples of relevant tasks of this class for industrial or service applications include, but are not limited to: handling of cables, assembly, packaging, co-operative human robot handling of non-rigid/deformable objects, etc. These research topics have been part of relevant international research projects (Collaborate: https://collaborate-project.eu/; CloPeMa: www.clopema.eu). Requirements: Applicants are expected to have a strong motivation for at least one of the following key topics in robotics: robot control, robot programming, system identification. Furthermore, good attitude for experimental work is mandatory. The candidates must have: very good programming skills with different languages (including C/C++, Python, Matlab/Simulink); confidence with electronic hardware and be capable to conduct experiments; attitude to problem solving, and be strongly motivated for team working. References:

G. Cannata, S. Denei, F. Mastrogiovanni. Towards the creation of tactile maps for robots and their use in robot contact motion control. Robotics and Autonomous Systems 63(3): 293–308, 2015.

S. Denei, F. Mastrogiovanni, G. Cannata. Parallel Force-Position Control Mediated by Tactile Maps for Robot Contact Tasks. Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2012), Vilamoura, Portugal, October 2012.

A. Del Prete, S. Denei, L. Natale, F. Mastrogiovanni, F. Nori, G. Cannata, G. Metta. Skin spatial calibration using force/torque measurements. Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2012), Vilamoura, Portugal, October 2012.

Contacts: Email: Giorgio.cannata@unige.it

9

5. Culture-aware robots and environmental sensor systems for elderly support

Tutor: Antonio Sgorbissa Co-Tutor(s): Barbara Bruno, Carmine Recchiuto Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it, caressesrobot.org Description: Research will be performed within the Horizon 2020 project CARESSES, a three years joint effort between EU and Japan aimed at designing culturally competent elder care robots, i.e., robots able to adapt how they behave and speak to the culture, customs and manners of the person they assist. In this general framework, the candidate will be allowed to choose between two different lines of research. In both lines of research, the system will be implemented using the robot Pepper developed by SoftBank Robotics (https://www.softbankrobotics.com/emea/en/index), using the functionalities for robot control that are made available by the NaoQi APIs, complemented with additional functionalities that have been developed in the context of CARESSES (as well as other tools). The objective of the first line of research will be to design a system for human-robot interaction that allows the robot and the human to purposefully communicate to each other, by taking into account the way in which people belonging to different cultures communicate through verbal and non-verbal interaction. Specifically, in the context of CARESSES, the robot must be able to understand the intentions and the emotions of the person by merging a priori knowledge about the her cultural identity, explicit information extracted from the speech as well as implicit information extracted from gestures, body postures and facial expressions.

Concerning technical solutions and tools to be adopted, the student will be encouraged to rely on Cloud-based services for Automatic Speech Recognition (e.g., Google ASR, https://cloud.google.com/speech/) as well as for Natural Language Processing (e.g., DialogFlow, https://dialogflow.com/), and finally for image recognition and categorization (e.g., CLARIFY: https://www.clarifai.com/). The objective of the second line of research will be to embed Pepper with advanced functionalities for autonomous navigation, obstacle avoidance, and localization, which constitute the basis for enabling the robot to exhibit a fully autonomous behavior. Even if – in principle – these functionalities are already provided by the NaoQi APIs, a deeper analysis shows that standard NaoQi functionalities are quite limited, and further developments are required to allow Pepper to show advanced capabilities in complex environment. Specifically, in the context of CARESSES, the system will take into account a priori knowledge about the cultural identity of the

10

person, which has an impact on the environment in which the person lives and the robot operates (in terms of furniture, decorations, etc.) and therefore provides hints to improve the navigation and localization process. Concerning technical solutions and tools to be adopted, the student will be encouraged to rely on Cloud-based services for the semantic labelling of images (e.g., CLARIFAI: https://www.clarifai.com/), as well as open source solutions for navigation, localization and mapping (e.g., OpenSLAM: https://openslam-org.github.io/). Requirements: The ideal candidate will have a strong motivation for working in an interdisciplinary and international group, a solid background in robotics and robot programming, and a specific interest in social robotics and human-robot interaction. Specifically, the student will be presumably required to learn and use the following programming languages and tools:

Python/Java/C++,

API to access cloud services for the semantic tagging of images as well as for Automatic Speech Recognition and Natural Language Processing,

NAOqi API for sensorimotor control of the Pepper's robot,

OWL2 API for building and querying ontologies. References:

Barbara Bruno, Nak Young Chong, Hiroko Kamide, Sanjeev Kanoria, Jaeryoung Lee, Yuto Lim, Amit Kumar Pandey, Chris Papadopoulos, Irena Papadopoulos, Federico Pecora, Alessandro Saffiotti and Antonio Sgorbissa, Paving the Way for Culturally Competent Robots: a Position Paper, 2017 IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN), Lisbon, Portugal, August 2017

Barbara Bruno, Carmine Tommaso Recchiuto, Irena Papadopoulos, Alessandro Saffiotti, Christina Koulouglioti, Roberto Menicatti, Fulvio Mastrogiovanni, Renato Zaccaria, Antonio Sgorbissa, Knowledge Representation for Culturally Competent Personal Robots: Requirements, Design Principles, Implementation, and Assessment, International Journal of Social Robotics, 2019

Cultural Robotics, First International Workshop, CR 2015, Held as Part of IEEE RO-MAN 2015, Kobe, Japan, August 31, 2015. Editors: Koh, J.T.K.V., Dunstan, B.J., Silvera-Tawil, D., Velonaki, M. (Eds.), Lecture Notes in Artificial Intelligence 9549, Springer.

Contacts: Email: antonio.sgorbissa@unige.it

11

6. Autonomous drones and UAVs for search & rescue Tutor: Antonio Sgorbissa Co-Tutor(s): Barbara Bruno, Carmine Recchiuto Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Research will be performed within the Italian project DIONISO, a three years effort aimed at designing autonomous systems (i.e., robots and smart wearable devices) that can support first responders in Search & Rescue operations. In this general framework, the candidate will be allowed to choose between two different lines of research. The objective of the first line of research will be to design and implement strategies that allow a team of drones embedded with sensors (e.g., cameras as well as additional devices for remote sensing) to monitor the environment in complete autonomy by coordinating their flight plans. Specifically, in the context of DIONISO, autonomous team coordination is the key to overcome the well-known limitations of multirotor drones in terms of energetic autonomy, and to allow them to repeatedly cover a large area in the most efficient and accurate way as possible (e.g., for Search & Rescue after a natural disaster, e.g., earthquake, fire, or flood). Also, the student will address the problem of merging and presenting the information returned by the team of drones during autonomous navigation, by providing the required situational awareness in a remote control room. The objective of the second line of research will be to design wearable system for Simultaneous Localization And Mapping (SLAM), i.e., based on wearable sensors to be embedded in rescuers clothes (e.g., Inertial Measurement Units, laser rangefinders, cameras - including RGBD and event cameras). Specifically, in the context of DIONISO, such wearable system allows rescuers to autonomously build a geometric map of the environment, augmented with additional visual, environmental and semantic information: images acquired by wearable cameras, environmental data returned by wireless sensors (temperature, acceleration, visibility), and finally symbolic labels assigned by the operator himself to significant locations (e.g., “church”, “school”, “crossroad”). The resulting map will then be used for Search & Rescue, to find an evacuation route, or even for assessing damages after a natural disaster (i.e., by comparing the environment before and after the event). Concerning technical solutions and tools, both lines of research will require the student to consider open source solutions for navigation, localization and mapping (e.g., OpenSLAM: https://openslam-org.github.io/). Additionally, the first line of research will require the student to consider collaborative strategies for coordinated

12

flight and exploration. The second line of research will require the student to consider Cloud-based solution for image-matching (e.g., CLARIFAI: https://www.clarifai.com/) as well as ontologies for the representation of semantic information associated to selected locations in the map (e.g., OWL2: http://www.owl2.it/). The Robot Operating System (ROS, http://www.ros.org/) will be considered in both cases as an option for implementation. Requirements: The ideal candidate will have a strong motivation for working in an interdisciplinary and international group, a solid background in robotics and robot programming, and a specific interest in autonomous robotic systems (including perception, reasoning, planning, action). Specifically, the student will be presumably required to learn and use the following programming languages and tools:

C++/Java,

ROS, the Robot Operating System,

OpenSLAM solutions,

API to access cloud services for the semantic tagging of images,

OWL2 ontologies.

References:

Carmine Tommaso Recchiuto, Antonio Sgorbissa, Post-disaster assessment with unmanned aerial vehicles: A survey on practical implementations and research approaches, Journal of Field Robotics Volume 35, Issue 4, Pages: 419-640, June 2018

Carmine Tommaso Recchiuto, Antonello Scalmato, Antonio Sgorbissa, A dataset for human localization and mapping with wearable sensors, Robotics and Autonomous Systems, Volume 97, November 2017, Pages 136-143.

Contacts: Email: antonio.sgorbissa@unige.it

13

7. Advanced robot manipulation skills acquired via human demonstrations Tutor: Fulvio Mastrogiovanni Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Humans excel in the manipulation of everyday objects or tools, especially when learning new manipulation skills, or when adapting already acquired skills to different tasks. Manipulation is a key ability in a world full of human-made artefacts and human-oriented objects and tools. Such an expert ability to use the hands for manipulation results from a life-long learning process that draws upon the observation of other humans as well as ourselves, as we discover how to handle objects first hand. Unfortunately, today's manipulation capabilities in robots are unable to achieve such high level of dexterity in comparison to humans, and acceptable results are obtained only in very specific application scenarios and use cases. In order for robots to operate reliably in environments made by humans and for humans, robots must be capable of manipulating a wide variety of unknown objects modulating different parameters, such as contact strength, motion dexterity, and grasp stability. Furthermore, an object or tool may be considered (and therefore perceived) not only for its physical features, but also for the possibilities it entails, i.e., its affordances. The need arises to provide robot hands (in terms of the software architectures controlling them) with high-end cognitive capabilities, and specifically to deal with real-world uncertainties in perception and action, to represent heterogenous, multi-modal object features, and to generalise previously acquired manipulation skills to new objects and tools, and for new tasks. This PhD proposal aims at (i) understanding how humans perform in-hand object manipulation, and at (ii) replicating the observed skilled movements with dexterous artificial hands. The work to be done will merge the concepts of reinforcement and transfer learning to generalise in-hand manipulation skills to previously unknown objects, tools, and tasks. An abstract representation of previously acquired knowledge will be fundamental for reproducing acquired (learned) skills to different robot hand mechanical shapes. The learning process will use heterogeneous, multi-modal data that will be collected, annotated and assembled into a large dataset. The data and the methods developed throughout the PhD period will be shared with the whole research community to allow for testing against open benchmarks and reproduction of results. More concretely, the core objectives of the PhD proposal are:

1. to build a multi-modal robot perception architecture that extracts data about object manipulation via human demonstration;

14

2. the creation of a multi-modal dataset of in-hand manipulation tasks including re-grasping, reorienting and fine repositioning;

3. the development of an advanced object modelling and recognition method, including the characterisation of object’s physical properties and their affordances;

4. to autonomously learn and precisely imitate human strategies in manipulation tasks.

This work will be done in the context of the H2020 CHIST-ERA collaborative project InDex. The student will have the opportunity to spend research periods abroad at Aston University (U.K.), Sorbonne University (France), TU Wien (Austria), and the University of Tartu (Estonia). Requirements:

Notions related to machine learning, data analysis, trajectory and motion planning.

Software development in C/C++. References:

1. A. Carfì et al. A multi-sensor dataset for human-human handover. DiB 22, 2019.

2. A. Capitanelli et al. On the manipulation of articulated objects in human-robot cooperation scenarios. RAS 109, 2018.

3. A. Carfì et al. Online human gesture recognition using recurrent neural networks and wearable sensors. Proc. IEEE RO-MAN 2018.

Contacts: Email: fulvio.mastrogiovanni@unige.it

15

8. Models for human-robot collaboration leveraging human biometric and physiological data obtained via wearable sensing suits

Tutor: Fulvio Mastrogiovanni Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Nowadays, industrial manufacturers are faced with increasing cost pressures, growing product diversity, and fluctuating demands. Even where human shop-floor operators are considered affordable, this new generation of complex products requires assembly adaptability, precision, and reliability beyond the skills of human workers alone. To be prepared for this high-mix, low-volume era, manufacturing methods must be flexible and automated. Traditional industrial robots continue to evolve to meet today’s automation needs, with a broad mix of form factors, payloads, and capabilities. However, new automation opportunities are also emerging for variable and semi-structured environments, especially in small and mid-sized businesses, where collaborative robots can be easily trained and reprogrammed to solve new tasks due to their ability to work alongside humans. These user-friendly, lightweight, flexible and affordable tools are lowering the automation barrier tremendously, while helping remove human intervention from labour intensive, hazardous work environments, thus enabling automation in areas previously considered too complex or costly. Unfortunately, a natural, genuine collaborative operation where robots work in synergy with human operators within a defined workspace is far from being a reality. In the past few years, University of Genoa has conducted research and development activities to design novel models and methods allowing collaborative robots to work effectively with human workers in industrial environments, which span cognition, perception, knowledge representation, learning, motion and manipulation skills. These modules allow collaborative robots to make sense of the environment, understand relevant perceptual traits, localise, efficiently represent information and reason upon sensory perceptions and high-level goals, manipulate objects avoiding static and dynamic obstacles, and coordinating with human operators and other collaborative robots. Based on such a work, the main goal of this PhD proposal is twofold. On the one hand, to advance state-of-the-art methods and techniques to integrate complementary technologies, including a wide range of wearable devices to be worn by human operators, e.g., wearable sensing suits, smart t-shirts, IMUs, EMGs, HMDs for virtual reality or augmented reality applications, to allow for an inherently-safe, human-centred, human-robot collaboration. On the other hand, to design novel methods for robot-based estimation of a human operator’s workload, fatigue and stress statuses, framing them according to the overall human-robot cooperation’s goals, as well as adapting robot behaviours and proving easy-to-understand

16

explanations about robot actions in order to enforce in the human operator feelings of robot acceptance and a sense of self-achievement. This work will be done in the contest of the TeamUp project, funded internally by University of Genoa. Furthermore, it will include activities done in cooperation with NTT Data Italia, specifically for the use of the HITOE smart t-shirt. Requirements:

Notions related to machine learning, data analysis, knowledge representation, reasoning and planning.

Software development in C/C++/Python. References:

1. A. Capitanelli et al. On the manipulation of articulated objects in human-robot cooperation scenarios. RAS 109, 2018.

2. A. Carfì et al. Online human gesture recognition using recurrent neural networks and wearable sensors. Proc. IEEE RO-MAN 2018.

3. K. Darvish et al. Flexible human-robot cooperation models for assisted shop-floor tasks. Mechatronics 51, 2018.

Contacts: Email: fulvio.mastrogiovanni@unige.it

17

9. Control and planning architectures for human robot collaborative tasks Tutor: Enrico Simetti Co-Tutor: Fulvio Mastrogiovanni, Giuseppe Casalino Department: Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa Web: www.dibris.unige.it Description: Collaborative robots in manufacturing have been proposed to work alongside humans to perform a series of tasks traditionally considered stressful, tiring or difficult. Recently, we have proposed FlexHRC [1], showing interesting results on the Baxter dual-arm robot and on a single mobile manipulator, while maintaining a uniform control architecture based on a task priority approach [2]. The goal of the research is to continue the work on FlexHRC, along two main trends. On the one hand, one trend would be extending the capabilities of the task-priority control framework, including force regulation and admittance schemes and integrating motion planning techniques. On the other hand, the research should focus on further increasing the cohesion with the high-level planning. The goal would be to show rich coordination capabilities between the robotic systems (composed by dual arm robots and mobile manipulators) at the action sequencing and execution levels, and to show integrated robot capabilities in executing high-level (i.e., discrete) actions via low-level (i.e., continuous) behaviours, monitoring the execution of such low-level behaviours and, in case of faults or unsuccessful executions, assign the high-level layer with the task of finding an alternative course of action. Requirements:

Good rigid body kinematics and dynamics knowledge Good C/C++ and MATLAB coding skills

References: [1] Darvish, K., Wanderlingh, F., Bruno, B., Simetti, E., Mastrogiovanni, F., & Casalino, G. (2018). Flexible human–robot cooperation models for assisted shop-floor tasks. Mechatronics, 51, 97-114. doi:10.1016/j.mechatronics.2018.03.006 [2] Simetti, E., & Casalino, G. (2016). A novel practical technique to integrate inequality control objectives and task transitions in priority based control. Journal of Intelligent and Robotic Systems: Theory and Applications, 84(1-4), 877-902. doi:10.1007/s10846-016-0368-6

Contacts: enrico.simetti@unige.it fulvio.mastrogiovanni@unige.it giuseppe.casalino@unige.it