Application of Metaheuristic Search Techniques for the Service Composition Problem in Internet of Things

Abstract

The Internet of Things (IoT) is an advancement of the Internet that has swept the world off its feet with its vast applications. It has given new wings to further improve our daily lives to be more comfortable, streamlined and optimized. We live in an intellectual and interconnected world and IoT has supplemented it with new communicative, interactive systems and innovative applications. It has vast applications across industries owing to its unique flexibility and ability, making it suitable for innovative applications such as Smart home, Smart power, e-health, etc. newlineThe main component of the IoT based systems is IoT services. These are standard web services provided by IoT devices to represent functionality on the Internet. A single service cannot fulfil all the requirements of the user. So, an appropriate set of services/tasks has to be composed by selecting the most suitable candidate for each service. Service composition is a set of these services. The candidate is an instance of the service with the same functionality but possesses different Quality of Service (QoS) parameters. newline As the number of IoT services and the candidates for each service facilitating the same functional domain are increasing tremendously, identifying an optimal composition of services that fulfils user requirements with the best optimized QoS parameters is of utmost importance. Hence, the service composition in IoT based on QoS parameters has become a decision problem. The Service Composition problem is an NP-hard problem as it would be extremely difficult to search for all the combinations of services to determine the best solution. Conventional optimization algorithms are not appropriate in such a search space. For such problems, efficient and high-performance evolutionary algorithms have proved to provide promising solutions within rational timeframe newlineIn this thesis Service composition Problem in IoT has been addressed using the metaheuristic search algorithms. First and foremost, this problem has been addressed using single objecti