The Prognostic and Health Management (PHM) is a process dealing with the predictive maintenance problems. It is based on three stages : Observation, Analysis, and Decision. In this context, we design and implement data driven solutions for diagnosis; prognostic and decision making problems based on artificial intelligence techniques and in particular machine learning. The PHM as a Service (PHMaaS) approach has been defined and adopted by introducing IoT, Edge computing and Cloud Computing techniques.

Today, With the development of information technology, the concept of smart healthcare became a trending research area. Smart healthcare uses a new generation of information technologies such as big data, cloud computing, and artificial intelligence(AI). These new techniques help to transform the traditional medical system to be more intelligent, efficient, convenient, and personalized. Computer-aided diagnosis (CAD) has become one of the major research subjects in medical computing and clinical diagnosis. However, how to efficiently and effectively make accurate diagnosis remains a challenging problem in data-driven models. therefore, we are interested in improving the performance of computer-aided diagnostic systems in the medical field by increasing the quality of medical data and the analytical techniques. To this end, several contributions have been proposed. First, we proposed an extension of Prognostic and Health Management (PHM) approaches in order to exploit its potential by adapting advanced industrial diagnostic models to medical diagnostics. Secondly, we focused on improving computer-assisted diagnosis, particularly in the dermatology field, using AI techniques as well as those of Big data. The proposed methods and the results obtained were validated by an extensive comparative analysis using benchmarks and private medical data.

Robot Operating System (ROS) is becoming a widely-used environment for devel- oping robot software systems. It provides unique features such as message-passing between processes and code reuse between robots. The new trend in ROS-based robotic systems is facing the development and delivery of effective services by com- bining the advantages of both cloud robotics and web services. Cloud robotics is the way that allows robots to overcome their limitations of pro- cessing and knowledge by boosting computational and cognitive capabilities. On the other hand, as an implementation of Service-Oriented Architecture (SOA), web Services allow mainly different ROS codes to be discovered over the internet for their reuse. However, the characterization, description, and discovery of the ROS service capability for the offered robotic functionality are still issues that are not fully ad- dressed. In this context, we focus in this thesis on developing an architecture for roboti software provisioning to both software developers and robots by exploiting the op- portunities of ROS, web services, and cloud robotics. We propose a complete SOA approach for cloud robotics, in which ROS-based robotic tasks are defined as web services. The approach focuses on defining the service cycle process of describing, discovering, and selecting services. Two characterizations for ROS web services are proposed. The service characterizations describe the semantic representation of the robot task from ROS itself. In each case, we present a strategy that allows users todiscover the relevant robotic service that can match their queries and robots.