We present a pipeline for fast object pose estimation using RGB-D images, which does not rely on image features or machine learning. We are interested in segmenting objects with large variety in app.earance, from lack of texture to presence of strong textures, with a focus on the task of robotic grasping. The proposed pipeline is divided into an object segmentation part and a pose estimation part. We first find candidate object clusters using a graph-based image segmentation technique. A modified Canny edge detector is introduced for extracting robust graph edges by fusing RGB and depth information. A suitable cost function is used for building the graph, which is then partitioned using the concept of internal and external differences between graph regions. The extracted object regions are then used to initialize the 3D position of a quaternion-based Particle Swarm Optimization algorithm (Q-PSO), that fits a 3D model of the object to the depth image. The fitness function is based on depth information only and the quaternion formulation avoids singularities and the need for conversions between rotation representations. In this work we focus on the details of the GPU implementation of Q-PSO, in order to fully exploit the highly parallelizable nature of the particular implementation of the particle swarm algorithm, and discuss critic implementation details. We then test the app.roach on different publicly available RGB-D object datasets, and provide numeric comparisons with other state-of-the-art methods, as well as a discussion on robustness and an extension to the case of articulated objects. We show how Q-PSO offers comparable performances to current learning-based app.roaches, while not suffering from the problems of lack of features in objects or issues related to training, such as the need for a large training set and long training times.

Object segmentation is an important capability for robotic systems, in particular for grasping. We present a graph- based approach for the segmentation of simple objects from RGB-D images. We are interested in segmenting objects with large variety in appearance, from lack of texture to strong textures, for the task of robotic grasping. The algorithm does not rely on image features or machine learning. We propose a modified Canny edge detector for extracting robust edges by using depth information and two simple cost functions for combining color and depth cues. The cost functions are used to build an undirected graph, which is partitioned using the concept of internal and external differences between graph regions. The partitioning is fast with O(NlogN) complexity. We also discuss ways to deal with missing depth information. We test the approach on different publicly available RGB-D object datasets, such as the Rutgers APC RGB-D dataset and the RGB-D Object Dataset, and compare the results with other existing methods.

This work a novel solutions that relies on autonomous robots with the aim of improving data centers power efficiency by providing an easy to use tool to perform environmental monitoring. Data center environmental monitoring has been highly explored in the last few years. In fact, due to the high power density managed by these buildings, precise monitoring is necessary to improve power efficiency without increasing the risk of hardware failure. This work proposes a robot system that is in charge of autonomously monitor temperature and humidity in the whole data center room environment. After that the robot have built a map of the data center room, it is able to localize in the map and performs localized measurements at different locations. Those measurements are stored and can be visualized by the user thanks to a web Graphical User Interface (GUI). The solution emulates an Environmental Sensor Network, a very popular solution for precise monitoring in these environments. The robot trajectories and the localized measurements can be easily reconfigured from the same GUI. For this reason, we name it a Virtual Sensor Network (VSN). In this work, we discuss the implementation choices and present some results collected during a use case experiment in a real environment.

Data-center monitoring has been a critical subject of research in recent years. Mobile robots have been successfully employed in the industrial field to efficiently perform common tasks. In this paper, we report some preliminary results on the study and development of a robotic system, in which a mobile robot equipped with a laser range sensor and an Inertial Measurement Unit (IMU) is able to autonomously navigate in a data-center room for accurate monitoring of critical measurements, such as servers’ external temperature, humidity and other physical quantities. The robot is able to autonomously create a map of a previously unknown room, localize therein and execute a list of measurements at different locations, which are provided by the user via a web graphical user interface (GUI). The robot is able to find the best trajectory to reach the given locations, while avoiding static and moving obstacles. The particular characteristics of the data-center scenario introduce specific problems related to map creation and localization using laser-based techniques (e.g., irregular surfaces as metal grids and high symmetry of the environment), which must be properly taken into account and are discussed throughout the paper. Preliminary experimental results show that the system is able to create a consistent map of the environment, to correctly localize itself therein and to follow a given path.

This work presents a robotic application aimed at performing environmental monitoring in data centers. Due to the high energy density managed in data centers, environmental monitoring is crucial for controlling air temperature and humidity throughout the whole environment, in order to improve power efficiency, avoid hardware failures and maximize the life cycle of IT devices. State of the art solutions for data center monitoring are nowadays based on environmental sensor networks, which continuously collect temperature and humidity data. These solutions are still expensive and do not scale well in large environments. This paper presents an alternative to environmental sensor networks that relies on autonomous mobile robots equipped with environmental sensors. The robots are controlled by a centralized cloud robotics platform that enables autonomous navigation and provides a remote client user interface for system management. From the user point of view, our solution simulates an environmental sensor network. The system can easily be reconfigured in order to adapt to management requirements and changes in the layout of the data center. For this reason, it is called the virtual sensor network. This paper discusses the implementation choices with regards to the particular requirements of the application and presents and discusses data collected during a long-term experiment in a real scenario.

This paper describes a trajectory planning algorithm for mobile robot navigation in crowded environments; the aim is to solve the problem of planning a valid path through moving people. The proposed solution relies on an algorithm based on the Informed Optimal Rapidly-exploring Random Tree (InformedRRT*), where the planner continuously computes a valid path to navigate in crowded environments. While the robot executes the trajectory of the current path, this re-planning method always allows a feasible and optimal solution to be obtained. Compared to other state-of-the-art algorithms, this solution does not compute the entire path each time an obstacle is detected, instead it evaluating the current solution validity, i.e., the presence of moving obstacles on the current path; in this case the algorithm tries to repair the current solution. Only if the current path is completely unacceptable is a new path computed from scratch. Thanks to its reactivity, our solution always guarantees a valid path that brings the robot to the desired goal position. This dynamic approach is validated in a real case scenario where a mobile robot moves through a human crowd in a safe and reliable way.

Cloud robotics is a new approach to robotics that exploits the internet as a resource for parallel computing and data sharing. Robots are no more considered as isolated devices but now they can add new functionalities, communicate with the environment and share knowledge base. Exploiting this new technology a robot can also take advantages by off-loading heavy computations to the cloud, thus reducing hardware costs, power consumption. The Joint Open Lab on Connected Robotic Applications laB (JOL CRAB) is a research laboratory created by Telecom Italia in collaboration with Politecnico di Torino. The aim is to investigate technologies and develop concepts where the focus is mainly placed on the relationship between robots and the cloud computing, addressing issues that arise from the use of robotic services in public/enterprise environments not only of technological, legal, economical, sociological or psychological kind, but also related to ergonomics, cognitive perception, and relational experience. The collaboration between university and the industry led to different field trials where issues and feasibility of new services were evaluated and a community of stakeholders was created in the territory. We present these field trials, the different issues that arise and how that can be solved.

The aim of this paper is to present a sliding autonomy approach for Unmanned Aerial Vehicles (UAVs) in the context of the project Fly4SmartCity. The project consists in the implementation of a cloud robotics service in which small UAVs are employed for emergency management, monitoring and surveillance in a smart city scenario. Human-robot interaction is mediated by the cloud robotics platform. We imagine three main levels of autonomy for UAVs: full autonomy, mixed-initiative and teleoperation. Then we propose different scenarios in which we analyze the Level Of Autonomy and the sliding autonomy approach. All services use shared knowledge (crowdsourcing and other data sources available on the Internet) for the management and control of the UAVs.

Data center monitoring has been a critical subject of research in recent years. In this paper we present a robotic system, based on the Robot Operating System (ROS), in which a mobile robot equipped with a laser range sensor and an Inertial Motion Unit (IMU) is able to autonomously navigate in a data center room for accurate monitoring of critical measurements, such as servers' external temperature, humidity and other physical quantities. The robot is able to autonomously create a map of a previously unknown room, localize therein and execute a list of measurements at different locations, which are provided by the user via a web Graphical User Interface (GUI). The application is based on a cloud robotics infrastructure which encloses the ROS nodes and exposes REST APIs to the user. We discuss our implementation choices with regards to the particular requirements of the scenario, both in terms of robot navigation and software infrastructure, and present some preliminary results in a real scenario.

Gestures are signals that contain important spatiotemporal information. Understanding gestures is a trivial task for humans, but for machines it is a challenging task involving thousands of computations per video frame. This paper investigates an efficient hand gesture recognition technique which is based on time projections of the hand location. For recognition, non-linear classifiers, namely Support Vector Machines and Artificial Neural Networks, are tested. The proposed method performs much faster than the conventional Markov Model based gesture recognition techniques while achieving comparable recognition results.

Mobile robots can be employed in the logistic field to efficiently perform common tasks, such as building and updating maps of indoor and outdoor logistic spaces, locating specific goods on the map, tracing the product flow in the area, while preserving situational awareness and safety of the environment. This paper reports and discusses the main results of the MACP4Log (Mobile Autonomous and Cooperating robotic Platforms for supervision and monitoring of large LOGistic surfaces) research project, aimed at the study and development of a set of algorithms and services, enabling autonomous navigation of a team of mobile robots in large logistic spaces, and exploiting cooperation, through communication with a supervisor and among the robotic platforms. Although the main services required for the robots coincide with the most common issues of mobile robotics (i.e., localization, mapping, SLAM and exploration), the particular characteristics of the logistic spaces introduce specific problems (e.g., related to a high symmetry of the environment and/or to its variability), which must be properly taken into account. The paper discusses in detail such problems, summarizing the main results achieved both from the methodological and the experimental standpoint, and is completed by the description of the general functional architecture of the whole system, including navigation, logistic, and monitoring services.

This paper presents a new and efficient algorithm for complex human activity recognition using depth videos recorded from a single Microsoft Kinect camera. The algorithm has been implemented on videos recorded from Kinect camera in OpenNI video file format (.oni). OpenNI file format provides a combined video with both RGB and depth information. An OpenNI specific dataset of such videos has been created containing 200 videos of 8 different activities being performed by different individuals. This dataset should serve as a reference for future research involving OpenNI skeleton tracker. The algorithm is based on skeleton tracking using state of the art OpenNI skeleton tracker. Various joints and body parts in human skeleton have been tracked and the selection of these joints is made based on the nature of the activity being performed. The change in position of the selected joints and body parts during the activity has been used to construct feature vectors for each activity. Support vector machine (SVM) multi-class classifier has been used to classify and recognize the activities being performed. Experimental results show the algorithm is able to successfully classify the set of activities irrespective of the individual performing the activities and the position of the individual in front of the camera.

The paper addresses and solves the problem of multirobot collaborative localization in highly symmetrical 2D environments, such as the ones encountered in logistic applications. Because of the environment symmetry, the most common localization algorithms may fail to provide a correct estimate of the position and orientation of the robot, if its initial position is not known, no specific landmark is introduced, and no absolute information (e.g., GPS) is available: the robot can estimate its position with respect to the walls of the corridor, but it could be critical to determine in which corridor it is actually moving. The proposed algorithm is based upon a particle filter cooperative Monte Carlo Localization (MCL) and implements a three-stage procedure for the global localization and the accurate position tracking of each robot of the team. Online simulations and experimental tests, which investigate different situations with respect to the number of robots involved and their initial positions, show how the proposed solution can lead to the global localization of each robot, with a precision sufficient to be used as starting point for the subsequent robot tracking.

The paper presents a method aiming at improving the reliability of Simultaneous Localization And Mapping (SLAM) approaches based on vision systems. Classical SLAM approaches treat camera capturing time as negligible, and the recorded frames as sharp and well-defined, but this hypothesis does not hold true when the camera is moving too fast. In such cases, in fact, frames may be severely degraded by motion blur, making features matching task a difficult operation. The method here presented is based on a novel approach that combines the benefits of a fully probabilistic SLAM algorithm with the basic ideas behind modern motion blur handling algorithms. Whereby the Kalman Filter, the new approach predicts the best possible blur Point Spread Function (PSF) for each feature and performs matching using also this information.

Cloud robotics is revolutionizing not only the robotics industry but also the ICT world, giving robots more storage and computing capacity, opening new scenarios that blend the physical to the digital world. In this vision new IT architectures are required to manage robots, retrieve data from them and create services to interact with users. In this paper a possible implementation of a cloud robotics architecture for the interaction between users and UAVs is described. Using the latter as monitoring agents, a service for fighting crime in urban environment is proposed, making one step forward towards the idea of smart city

Robotics education at M. Sc. courses is quite challenging, especially in the organization of intriguing laboratory activities about industrial manipulators, to accompany and complete traditional material and lessons on kinematics, dynamics, motion planning and control issues. Time and space constraints, the number of students attending the labs, and the available software and equipments (and their cost) can seriously limit the lab practices that can be actually offered in a Robotics course. On the basis of the experience matured at Politecnico di Torino, this paper describes how lab practices about manipulators can be organized by using low-cost structures, built up from the standard LEGO Mindstorms Kit, so to let the students become familiar with manipulators in different ways. Such practices are then well completed by activities with an industrial robot, available at Politecnico.

This article investigates the problem of SimultaneousLocalization and Mapping (SLAM) from the perspective of linear estimation theory.The problem is first formulated in terms of graph embedding: a graph describing robot poses at subsequent instants of time needs be embedded in a three-dimensional space, assuring that the estimated configuration maximizes measurement likelihood.Combining tools belonging to linear estimation and graph theory, a closed-form approximation to the full SLAM problem is proposed, under the assumption that the relative position and the relative orientation measurements are independent.The approach needs no initial guess for optimization and is formally proven to admit solution under the SLAM setup.The resulting estimate can be used as an approximation of the actual nonlinear solution or can be further refined by using it as an initial guess for nonlinear optimization techniques.Finally, the experimental analysis demonstrates that such refinement is often unnecessary, since the linear estimate is already accurate.

We present real-time, stereo vision based autonomous landing system for small Unmanned Aerial Vehicles (UAV) onto an unknown landing target. The paper describes the algorithms and design of FPGA based co-processor implementing Artificial Neural Network (ANN) to implement real time object tracking, 3D position estimation using Visual Odometry(VO), Horizontal displacement and Euclidean distance from landing target. This approach doesn't require any explicit marker or landing target, it estimates attitude, track safe landing area, and compute distance and horizontal displacement form landing target. Experimental results show suitability of the real-time stereo vision landing approach using FPGA for tracking, that doesn't require any explicit landing marker.

We present an academically developed framework for real time computations in Micro UAVs and Ground Robotics. Dynamic Partial Reconfiguration is used as a hardware accelerator in a heterogeneous environment which enables computationally demanding applications to take effective advantage of adaptive hardware resources while maintaining a high flexibility of software. We demonstrate our hypothesis by prototyping a workable application of aerial image acquisitioning and processing using partially reconfigurable hardware. We present an in-depth evaluation of proposed architecture in terms of its benefits in area, power consumption and timings.

The human hand is an extremely complex system due to its large number of degrees of freedom (DoF) within a significantly reduced space. Moreover, it is required for most of the tasks performed by humans. That is why it is necessary to understand deeply all the characteristics of the human hand in order to develop devices interacting with it: to support it, to substitute injured parts, to help the recovery from injuries, or to enhance the performances while preserving its natural level of dexterity. The aim of this paper is to provide a complete and exhaustive summary of the kinematic, static and dynamic characteristics of the human hand as a preliminary step towards the development of hand devices such as prosthetic/robotic hands and exoskeletons. Both fields provide promising opportunities in research and space applications; the former through humanoid robotic helpers (e.g., Eurobot, Robonaut), the latter through the rising necessity to help the astronauts during Extravehicular Activity (EVA). In literature, several papers can be found analyzing kinematics, workspace, constraints and forces of the human hand 2,4 . However this information is scattered among several papers, regarding in particular the exerted forces and the dependencies of joint forces and velocities from the angular values of the same joint or the adjacent one. Direct and inverse kinematics are presented for all the fingers and the data related to maximum forces, velocities, acceleration for each joint of each finger has been collected and is presented in this work.