Routine learning using statistics for robot’s tasks planning software

Abstract

Nowadays, the need to optimize time has caused a recent growth in the interest for home robots. Therefore, in this thesis the main household tasks feasible by a home robot are described and their usefulness is analysed by means of a questionnaire. This questionnaire is also required in order to determine the most interesting applications for the present society and focus on them. Home robots usually share their workspace with people. Therefore, there exists the need to take into account the presence of humans when planning their actions and it is indispensable to have knowledge of robots’ environments. It means knowing when (time and events duration) and where (workspace) robot's tasks can be performed. This thesis deals with the obtaining of the information required to execute a software to plan tasks to be performed by a robot. With this aim, two programs are developed. The first one is able to analyse data from real events using statistics, create histograms, determine the probability that a certain event takes places in a certain time and estimate a certain event duration. The second one is capable to define meaningful areas or zones in the robot workspace by the use of a Machine Learning technique called clustering. Finally, these programs are tested using real data obtained from different cameras located along the corridors of CSE department of University of Oulu.