Abstract:
Robots are progressively becoming more and more useful in bringing comfort to human life. One of the important new roles of robots is caring for elderly people who want to stay at home due to physical or cognitive difficulties. Navigation algorithm of a robot is a basic constituent that is to be programmed for its desired motion. Autonomous navigation of robots is tough to be planned in uncertain and dynamic environment. It becomes a complicated task to navigate a robot when the human‟s movement is uncertain. Partially Observable Markov Decision Process (POMDP) is one of the techniques used in navigation of robots in uncertain environment. POMDP is a complicated technique that requires more processing and computational power. In this work a Localized POMDP technique is introduced. This technique will enable robot to reduce computational power and allow it to calculate results in lesser time. Furthermore, in this work a robot learning phase is defined. In this phase a robot learns the positions of human with respect to time and creates a probability distribution function that help the robot to navigate human more accurately. An algorithm in MATLAB software with simulation is developed for its implementation and results.
